Closed Bug 1089183 Opened 10 years ago Closed 9 years ago

Enabling OMTC slows down the entire browser

Categories

(Core :: Graphics: Layers, defect)

33 Branch
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE
mozilla36
Tracking Status
firefox33 + fixed
firefox34 + fixed
firefox35 + fixed
firefox36 + fixed
relnote-firefox --- 33+

People

(Reporter: rafa.oliveirap, Assigned: nical, NeedInfo)

References

Details

(Keywords: perf, Whiteboard: Workaround (not recommended): See comment #6 or #7 in bug 1083071)

Attachments

(7 files, 1 obsolete file)

Attached file Graphics.txt (obsolete) —
User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:33.0) Gecko/20100101 Firefox/33.0
Build ID: 20141023194920

Steps to reproduce:

Enabled OMTC


Actual results:

Firefox got slower


Expected results:

Firefox should've gotten faster
It gets slow as in there's a long time between my action and the browser's response. It appears to also have low FPS when scrolling, but that might also be result of said lag. Window resizing is also slow, including turning full screen mode on. The slowness goes away when OMTC is turned off.
Component: Untriaged → Graphics: Layers
Product: Firefox → Core
Status: UNCONFIRMED → NEW
Ever confirmed: true
Summary: OMTC slowness → Enabling OMTC slows down the entire browser
Blocks: 899785
I have a friend (who works in IT too) who had the same issue. Disabling OMTC fixed his issue.
Rafael, could you provide the information in about:support?
Flags: needinfo?(rafa.oliveirap)
Attachment #8511537 - Attachment is obsolete: true
Flags: needinfo?(rafa.oliveirap)
Keywords: perf
Attached file aboutsupport.txt
Here is the information from my friend.
He confirms that it is still happening with 33.0.2 (omtc = off fixes his issue)
And disabling all his extensions does not change anything.
Whiteboard: Workaround (not recommended): See comment #6 or #7 in bug 1083071
I experience the same problem (Windows 7). The only thing which helps is disabling OMTC.
Do you all continue to see this on Fx 33.0.2?  Please update and let us know.  Thank you.
(In reply to [:tracy] Tracy Walker - QA Mentor from comment #11)
> Do you all continue to see this on Fx 33.0.2?  Please update and let us
> know.  Thank you.

Still the same problem with 33.0.2.
To anyone that can reproduce, please paste the graphics section of about:support in this bug so that we can see if there is a correlation between hardware/drivers.
Can you try to disable Direct2D, but leave everything else at default settings (go to about:support and set gfx.direct2d.disabled to true)?
(In reply to Nicolas Silva [:nical] from comment #13)
> To anyone that can reproduce, please paste the graphics section of
> about:support in this bug so that we can see if there is a correlation
> between hardware/drivers.

Direct2D aktiviert: true
DirectWrite aktiviert: true (6.1.7600.16385)
Geräte-ID: 0x68f9
GPU #2 aktiv: false
GPU-beschleunigte Fenster: 1/1 Direct3D 11 (OMTC)
Karten-Beschreibung: ATI Radeon HD 5450
Karten-RAM: 1024
Karten-Treiber: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Treiber-Datum: 12-9-2009
Treiber-Version: 8.682.1.0
Vendor-ID: 0x1002
WebGL-Renderer: Google Inc. -- ANGLE (ATI Radeon HD 5450 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d
AzureContentBackend: direct2d
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0
(In reply to Bas Schouten (:bas.schouten) from comment #14)
> Can you try to disable Direct2D, but leave everything else at default
> settings (go to about:support and set gfx.direct2d.disabled to true)?

Problem still occurs.
Useful info from the duplicate bugs:

Adapter Description	ATI Radeon HD 5450
Vendor ID	0x1002
Device ID	0x68f9
Driver Date	12-9-2009
Driver Version	8.682.1.0
Direct2D Enabled
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)

--

ATI Mobility Radeon HD 5165
Vendor ID	0x1002
Device ID	0x9480
Driver Date	9-8-2009
Driver Version	8.653.0.0
Direct2D Enabled	true
GPU Accelerated Windows	1/1 Direct3D 10
(for this user using D3D9 instead of D3D11 fixes the issue)

---

- ATI Mobility Radeon HD5650 (driver: 8.702.5.0 latest version)
- ATI Mobility Radeon HD 5145 - driver 8.653.0.0 (latest version)
- ATI Radeon HD5670 graphics card
- Intel HD Graphics (driver: 8.702.5.0 latest version)
https://support.mozilla.org/en-US/questions/1026328
Bug 1089875 comment 5:

> ATI Mobility Radeon HD 5145 - driver 8.653.0.0 :
> 
> hardware acceleration ON
> layers.offmainthreadcomposition.enabled = true
> layers.prefer-d3d9 = true
> gfx.direct2d.disabled = true
> 
> ==> OK, no lagging.
So does that mean that the problem might be not so much with D3D11 compositing, as it is with surface-sharing (e.g. sharing d2d-drawn surfaces)?

It would be interesting in this respect to know how WebGL FPS are affected, e.g. on a page like
http://alteredqualia.com/three/examples/webgl_pasta.html
(as WebGL relies on surface sharing if possible, but is unrelated to D2D).
For info, my friend is going to the office on Friday to lend us his laptop.
Added to the release notes with
"Windows: Interface may be slow when typing or selecting text (1087674)" as wording.
Of course, the bug number is "1089183"
This is a breakdown of ATI driver versions in our crash data, so we can map the affected version numbers ranges above to numbers of users likely to be affected, and can evaluate the impact of tentative blacklisting rules if we choose to get there.

The first column is ATI driver version, the second column is number of crash reports with this exact ATI driver version, the third column is number of crash reports with ATI driver version <= that. (Total 37,463 crash reports, data from October 25).

The above reports seem all to be on quite old (late 2009, early 2010) driver versions around 8.650 to 8.682.

I'm not factoring in the information that this bug might be Windows 8.1 only, because we don't seem to have enough data about that at this point.

From this table, we see that if we blacklisted driver versions from comment 17, so versions 8.653 to version 8.682.1, we would be blacklisting (11004 - 10377) / 37643 == 1.6% of Windows ATI users, so about 0.3% of all Windows users.

Of course, we can't blacklist anything until we know what feature exactly needs to be blacklisted :)

Generated by:

$ cat 20141025-pub-crashdata.csv | grep Windows.NT | grep 'AdapterVendorID: 0x1
002' | grep 'AdapterDriverVersion\:\ [0-9]\+\.[0-9]\+\.[0-9]\+\.[0-9]\+' |sed '
s/.*AdapterDriverVersion\:\ \([0-9]\+\.[0-9]\+\.[0-9]\+\)\.\([0-9]\+\).*/\1 \2/
g' | awk '{$2 = sprintf("%04d", $2); print}' | sed 's/\([0-9]\+\)\ \([0-9]\+\)/
\1.\2/g' | sort -n | uniq -c | awk '{sum += $1; printf("%-24s %-8d %d\n", $2, $
1, sum);}' > ati-win-driver-versions
Attachment #8513590 - Attachment description: ati-win-driver-versions → Breakdown of ATI driver versions. Third column is number of users with <= that driver version. Using Oct. 25 crashdata.
Here's the breakdown for Windows 8.1 just FYI. There, these old driver versions are very rare --- there was not a single crash with a driver version in this range, on Oct. 25, on Windows 8.1.
(In reply to Benoit Jacob [:bjacob] from comment #23)
> I'm not factoring in the information that this bug might be Windows 8.1
> only, because we don't seem to have enough data about that at this point.

I experience the problem on Windows 7.
Thanks testit, that's very helpful information! :-)
OS: Windows 8.1 → All
Hardware: x86_64 → All
It would be very interesting to understand this bug, so I'm still looking forward to us being able to reproduce it and come up with a good solution. Meanwhile though, as a short-term mitigation, it seems that we can easily avoid making Firefox slow for some users, by just blacklisting D3D11 layers and D2D on old ATI drivers from > 4.5 years ago, which seems acceptable. Note that ATI Catalyst 10.6 from June 2010 had the DLL version number 8.741 to would pass this >= 8.700 requirement. Users on older drivers still get D3D9 layers and (slower) WebGL.
Attachment #8513665 - Flags: review?(jmuizelaar)
This tells us that on Oct. 25, among Firefox 32.0 crash reports, 1652 were from people on Windows, with D3D10 layers (so on D3D10 capable hardware), with a driver version older than 8.700.

This is out of 32534 crash reports that day, on Firefox 32.0, on Windows, with D3D10 layers.

1652/32523 == 5%

So this blacklist rule would cause 5% of the D3D10+ enabled population, to fall back to D3D9. That's quite a lot. Maybe we need to wait until we know for sure that many of these people are actually affected by the bug, before we take this step.
We can reproduce (Toronto office)! On the first ATI laptop that we could find. It happens to have a ATI Mobility Radeon HD 3650.

The driver version that we reproduced with is Catalyst 10.1 (January 2010), which has the internal DLL version number 8.641.1.1000. It's just the first version that we tried.

So the STR seems to be: just take any ATI D3D10 hardware, and install an old driver from early 2010. The input lag is immediately noticeable (and very, very annoying indeed).
I've also had Firefox UI lagging problems with H/W accel ON (checked in Options) and OMTC enabled = true.  Win 7, ATI Radeon HD5670, driver 8.690.0.0 (Dell's latest).  These settings were working fine through 32.0.3, but problems have been consistent from 33.0 to present (33.0.2).  Details of various testing (including most recently d3d9 and D2D) in Mozilla Support Firefox forum thread 1026328.
https://support.mozilla.org/en-US/questions/1026328#answer-646966 :

ATI Radeon HD 5670 -  8.690.0.0

just updated to 33.0.2, and redid all above tests with same results.
I expanded my tests with hardware acceleration ON and OMTC enabled = true to cover all combinations of:

layers.prefer-d3d9:    T T F F  (T = True ; F = False)
gfx.direct2d.disabled: T F T F  (T = True ; F = False) 
test result:           C C L L  (C = no lag, but contrast problem ; L = lag, no contrast problem)

the color contrast was muted compared to before (most obvious on the minimize/maximize/close buttons)

so again, I've reverted to hardware acceleration ON, OMTC enabled = false, others default (both false)
Confirming problem with laggin for FF 33.0.2:
ATI Mobility Radeon HD 5145 - 8.653.0.0 (latest version for Toshiba)

Web-GL-renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5145 Direct3D9Ex vs_3_0 ps_3_0)
Hardware acceleration = on
layers.offmainthreadcomposition.enabled = true
(In reply to bigstijn from comment #32)
> Confirming problem with laggin for FF 33.0.2:
> ATI Mobility Radeon HD 5145 - 8.653.0.0 (latest version for Toshiba)
> 
> Web-GL-renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5145
> Direct3D9Ex vs_3_0 ps_3_0)
> Hardware acceleration = on
> layers.offmainthreadcomposition.enabled = true
device id = 0x9553
Updates:

1) we've confirmed that on the machine that reproduced the bug with Catalyst 10.1, upgrading to the current driver update solves the problem. So, this is totally confirmed to be driver-dependent.

...but:

2) BenWa/Jeff realized that this lag/slowness is related to the Windows cursor. In the control center, adjust the cursor blink rate --> this affects the amount of lag experienced in Firefox. It's as if Firefox were waiting for some invalidation/presentation/repainting event of some kind, but I don't know the details.

Bas, can you make sense of that? - if you want to reproduce locally, it seems that all you need is to downgrade ATI drivers from something around late 2009/early 2010, such as Catalyst 10.1.
Flags: needinfo?(bas)
The about:support info attached to comment 0 (or the full text in comment 4) means the reported can reproduce this on an Nvidia card.
Good point, Reuben! That's getting even more mysterious now --- because at this point we know for sure that at least on ATI is depends on the driver version.
It doesn't seem to depend on the driver version in the reporter's case. Between comment 0 and comment 4, he updated to the latest Nvidia drivers.
Tried disabling Direct2D and the problem seemed to be solved for me, despite OMTC being enabled. Will give more info when I go back home. Also, in my case, I'm not sure the Nvidia driver is even used since about:support said that GPU#2 is not enabled and the WebGL Renderer is the Intel's integrated GPU.
(In reply to bigstijn from comment #32)
> Confirming problem with laggin for FF 33.0.2:
> ATI Mobility Radeon HD 5145 - 8.653.0.0 (latest version for Toshiba)
> 
> Web-GL-renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5145
> Direct3D9Ex vs_3_0 ps_3_0)
> Hardware acceleration = on
> layers.offmainthreadcomposition.enabled = true

I posted this to bug 1084440 but didn't get a response back on whether the OP tried it.

A quick forum search yielded this info: http://goo.gl/BD21ni

Apparently those driver will work on *any* laptop with kludged drivers such as the Toshiba. Can you give them a try? They're 2011 but at least that's newer than 2009. The instructions are pretty detailed and the forum from which I obtained this info said to be exacting in following the instructions. I'd also suggest having the installer for the drivers you currently have.
(In reply to Arthur K. from comment #39)
> (In reply to bigstijn from comment #32)
> > Confirming problem with laggin for FF 33.0.2:
> > ATI Mobility Radeon HD 5145 - 8.653.0.0 (latest version for Toshiba)
> > 
> > Web-GL-renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5145
> > Direct3D9Ex vs_3_0 ps_3_0)
> > Hardware acceleration = on
> > layers.offmainthreadcomposition.enabled = true
> 
> I posted this to bug 1084440 but didn't get a response back on whether the
> OP tried it.
> 
> A quick forum search yielded this info: http://goo.gl/BD21ni
> 
> Apparently those driver will work on *any* laptop with kludged drivers such
> as the Toshiba. Can you give them a try? They're 2011 but at least that's
> newer than 2009. The instructions are pretty detailed and the forum from
> which I obtained this info said to be exacting in following the
> instructions. I'd also suggest having the installer for the drivers you
> currently have.

Apparently the link to the updated Catalyst drivers was broken. Users in that forum reported success with 13-1-legacy_vista_win7_win8_64_dd_ccc.exe on kludged Toshiba laptops. Updated link to those drivers here: http://goo.gl/gFkhPO
Comment on attachment 8513665 [details] [diff] [review]
Temporary band-aid: Blacklist D3D11 and D2D on old ATI drivers (early 2010 and older)

I wasn't sure if we still wanted to do this or not...
Attachment #8513665 - Flags: review?(jmuizelaar) → review+
(In reply to Jeff Muizelaar [:jrmuizel] from comment #41)
> Comment on attachment 8513665 [details] [diff] [review]
> Temporary band-aid: Blacklist D3D11 and D2D on old ATI drivers (early 2010
> and older)
> 
> I wasn't sure if we still wanted to do this or not...

Question relevant to end-user testing:
When blacklisting certain capabilities that are about:config parameters, which takes priority?  Or do the blacklisted capabilities' parameters get forced (or the defaults changed) so that the user is aware that those capabilities are blacklisted?
As noted in comment 31, I've reverted to hardware acceleration ON, OMTC enabled = false, layers.prefer-d3d9 = false (default), and gfx.direct2d.disabled = false (default).  Is there any graphics speedup (or other) benefit of this over just turning hardware acceleration OFF?  (My system details are in comment 30; running FF 33.0.2).  Sorry if this pollutes the thread.
Hello everybody,

I'm just a standard user, on Win7, but this bug is really annoying ! The browser is not usable ! 

Can you resolve it ? This may concern thousand of users...

Didn't have the slow before the upadate today

Thanks
I confirm that omtc turned off solves the problem ^^

(working on standard laptop i5, hd5470, win7 family premium)
(In reply to alex from comment #44)
> Hello everybody,
> 
> I'm just a standard user, on Win7, but this bug is really annoying ! The
> browser is not usable ! 
> 
> Can you resolve it ? This may concern thousand of users...
> 
> Didn't have the slow before the upadate today
> 
> Thanks

The resolution for many people is to update their video drivers (if possible). Have you tried this option?
(In reply to Arthur K. from comment #46)

> The resolution for many people is to update their video drivers (if
> possible). Have you tried this option?

Just said before your answer, turning off omtc (in about:config) solved it

Even so, how are you going to tell all windows users to update their video drivers or turn off omtc ?
(In reply to alex from comment #47)
> (In reply to Arthur K. from comment #46)
> 
> > The resolution for many people is to update their video drivers (if
> > possible). Have you tried this option?
> 
> Just said before your answer, turning off omtc (in about:config) solved it
> 
> Even so, how are you going to tell all windows users to update their video
> drivers or turn off omtc ?

It's not an easy thing to accomplish nor is there a one-size-fits-all solution. Blacklisting seems to be working but degrades performance in case where a driver update, if available, could fix things. I suppose the hard way to do it would be to fire up some notification upon (blacklisted) version check and verification and then load some notification or link to the updated driver. There's a narrow group who is affected (Intel) and some ATI people. The second reality is that we're seeing a generation of VGA cards that were validated pre-DX11 and are just showing their age in terms of feature support and there's not too much we can do about that.
(In reply to Arthur K. from comment #48)
> Blacklisting seems to be working but degrades performance in case
> where a driver update, if available, could fix things.

But is there really a performance problem? In my opinion it's the best solution just to blacklist the drivers which cause problems. Additionally I'm sure there are many people which don't like to upgrade the graphics driver (including me) and prefer switching to another browser.
(In reply to testit from comment #49)
> (In reply to Arthur K. from comment #48)
> > Blacklisting seems to be working but degrades performance in case
> > where a driver update, if available, could fix things.
> 
> But is there really a performance problem? In my opinion it's the best
> solution just to blacklist the drivers which cause problems. Additionally
> I'm sure there are many people which don't like to upgrade the graphics
> driver (including me) and prefer switching to another browser.

I suppose it's for the benchmarks to decide but from the standpoint of usability, there's no problem. Users on blacklisted drivers just won't get all the benefits of OMTC.
Let's clarify things a bit.
We encourage people to update their drivers because on average, recent driver versions work better than older ones. This happens to fix the issue but we know it is not THE solution to get a good (heck, usable) web experience to all of the users. Deactivating OMTC works around some problems but we also know that turning OMTC off causes other bugs so I wouldn't encourage people to do it too much.
We are working on finding solutions to this bug, even for older drivers. It will probably involve some blacklisting but to do this we need to understand exactly what feature we blacklist on what hardware and what driver versions, so it takes a bit of investigating. The issue will be fixed, and sooner rather than later.
Now, if you read this comment and have old drivers, you should know that updating them has a good chance of helping you while we are working on a solution on our end.
I hope that this makes things a bit clearer about we are going about this and the other driver-related issues we are having with this release.
(In reply to Nicolas Silva [:nical] from comment #51)
> ...Deactivating OMTC works around some problems
> but we also know that turning OMTC off causes other bugs so I wouldn't
> encourage people to do it too much.

What sorts of other bugs?  I haven't seen any that I'm aware of, in a few days of browsing.

Also, if you don't mind, can you address my comments 42 and 43?

Thanks.
Bruce, please avoid off topic discussions/questions. The bug is about Firefox being slow on some hardware.

> What sorts of other bugs?
Some graphic glitches.
I have now have a laptop (intel 0x0116 driver 8.15.10.2455 + nvidia 0x0df4 driver 8.17.12.8577) That seems to have this problem (or something that looks like this problem). Turning D2D off (while still using D3D11) fixes the issue.
It feels like there's a problem with the synchronization of shared DXG surfaces. Hard to tell because it looks like there is a whole bunch of issues with this laptop (text disappearing ect.)
(In reply to Nicolas Silva [:nical] from comment #54)
> I have now have a laptop (intel 0x0116 driver 8.15.10.2455 + nvidia 0x0df4
> driver 8.17.12.8577) That seems to have this problem (or something that
> looks like this problem). Turning D2D off (while still using D3D11) fixes
> the issue.
> It feels like there's a problem with the synchronization of shared DXG
> surfaces. Hard to tell because it looks like there is a whole bunch of
> issues with this laptop (text disappearing ect.)

What happens when you update the Intel driver to 15.28.22.64.3517 and enable D2D or 
what happens when you update the NVidia driver to 344.48 and enable D2D?
Toshiba notebook:
I updated the graphics driver using http://support.amd.com/en-us/download/auto-detect-tool  (instead of using the toshiba website or the windows 7 search for updates) .
=> is ok.

ATI Mobility Radeon HD 4500/5100 Series
driver 8.970.100.1100  (before: 8.653.0.0)
driver date: 4-29-2013

hardware acceleration ON
gfx.direct2d.disabled = false
layers.prefer-d3d9    = false
layers.offmainthreadcomposition.enabled = true
(In reply to bigstijn from comment #56)
> Toshiba notebook:
> I updated the graphics driver using
> http://support.amd.com/en-us/download/auto-detect-tool  (instead of using
> the toshiba website or the windows 7 search for updates) .
You must have seen my post on the AMD tool in another bug.

> => is ok.
This resolved it for you?

> 
> ATI Mobility Radeon HD 4500/5100 Series
> driver 8.970.100.1100  (before: 8.653.0.0)
> driver date: 4-29-2013
> 
> hardware acceleration ON
> gfx.direct2d.disabled = false
> layers.prefer-d3d9    = false
> layers.offmainthreadcomposition.enabled = true
Yes, for me this is resolved, by updating drivers using http://support.amd.com/en-us/download/auto-detect-tool.

But I did not use the link provided in comment 40 as I could not determine if that site was trustable.
(In reply to bigstijn from comment #58)
> Yes, for me this is resolved, by updating drivers using
> http://support.amd.com/en-us/download/auto-detect-tool.
> 
> But I did not use the link provided in comment 40 as I could not determine
> if that site was trustable.

Understood. I was referring to bug 1083071 comment 194. Glad it resolved.
Finally got a fresh build in a debugger running on the affected machine I have.
I placed a break point in every single place where HRESULTs are checked and failing in DrawTargetD2D, TextureD3D11 and CompositorD3D11, and also in all of the code paths where we early return in the middle of a drawing or compositing operation, as well as all of the code paths where creating a DrawTarget, a TextureClient or a D3D10 texture returns nullptr. None of the breakpoints are hit.

I also activated the D3D debug layer (all log levels activated except info which just dumps all api calls), and the only log that I get is:

D3D11 WARNING: ID3D11DeviceContext::CopySubResourceRegion1: Empty pSrcBox means nothing will get copied [...].

CopySubRegion is called in CompositorD3D11::CreateRenderTargetFromSource.
(In reply to Nicolas Silva [:nical] from comment #60)
> Finally got a fresh build in a debugger running on the affected machine I
> have.
> I placed a break point in every single place where HRESULTs are checked and
> failing in DrawTargetD2D, TextureD3D11 and CompositorD3D11, and also in all
> of the code paths where we early return in the middle of a drawing or
> compositing operation, as well as all of the code paths where creating a
> DrawTarget, a TextureClient or a D3D10 texture returns nullptr. None of the
> breakpoints are hit.
> 
> I also activated the D3D debug layer (all log levels activated except info
> which just dumps all api calls), and the only log that I get is:
> 
> D3D11 WARNING: ID3D11DeviceContext::CopySubResourceRegion1: Empty pSrcBox
> means nothing will get copied [...].
> 
> CopySubRegion is called in CompositorD3D11::CreateRenderTargetFromSource.

I believe these CopySubRegion calls are coming from Direct2D. We are very strict about checking our usage of CopySubResourceRegion now. If you turn off D2D do you still get them?
Unless I missed something (data being scattered in a different places), all reports coming from ATI drivers are in the versions ranging from 8.653 to 8.690.0.0

Let's blacklist D2D on these configuration for now.

There's also the reports with dual nvidia+intel hardware that I need to go though.
Attached patch ati-blacklistSplinter Review
Approval Request Comment
[Feature/regressing bug #]:
[User impact if declined]: unusable browser with some ATI drivers (0.3% of windows users)
[Describe test coverage new/current, TBPL]:
[Risks and why]: low
[String/UUID change made/needed]:
Attachment #8518307 - Flags: review?(bjacob)
Attachment #8518307 - Flags: approval-mozilla-release?
Attachment #8518307 - Flags: approval-mozilla-beta?
Attachment #8518307 - Flags: approval-mozilla-aurora?
Comment on attachment 8518307 [details] [diff] [review]
ati-blacklist

Review of attachment 8518307 [details] [diff] [review]:
-----------------------------------------------------------------

::: widget/windows/GfxInfo.cpp
@@ +1023,5 @@
> +    // Disable D2D on some ATI drivers which don't support dxgi keyed mutex correctly (bug 1089183)
> +    APPEND_TO_DRIVER_BLOCKLIST_RANGE( DRIVER_OS_ALL,
> +        (nsAString&) GfxDriverInfo::GetDeviceVendor(VendorATI), GfxDriverInfo::allDevices,
> +      nsIGfxInfo::FEATURE_DIRECT2D, nsIGfxInfo::FEATURE_BLOCKED_DRIVER_VERSION,
> +      DRIVER_BETWEEN_INCLUSIVE_START, V(8,653,0,0), V(8,691,0,0), "ATI Catalyst 14.6+");

Can you add a comment about where these numbers come from.
(In reply to Nicolas Silva [:nical] from comment #62)
> Unless I missed something (data being scattered in a different places), all
> reports coming from ATI drivers are in the versions ranging from 8.653 to
> 8.690.0.0
> 
> Let's blacklist D2D on these configuration for now.
> 
> There's also the reports with dual nvidia+intel hardware that I need to go
> though.

But without having some of these users with combo setups update one or both of the drivers, we'll never know which driver is causing it and might be blacklisting unnecessarily. Such as in the case of bug 1093863 comment 3. It *could* be that the reason that both device are present but one is not registered, for whatever reason, is because it might be just a case of a driver needing a reinstall or updated to re-register itself.
Comment on attachment 8518307 [details] [diff] [review]
ati-blacklist

Review of attachment 8518307 [details] [diff] [review]:
-----------------------------------------------------------------

Please address Jeff's comment.
Attachment #8518307 - Flags: review?(bjacob) → review+
(In reply to Benoit Jacob [:bjacob] from comment #66)
> Please address Jeff's comment.

the numbers come from this bug already mentioned in the comment
Comment on attachment 8518307 [details] [diff] [review]
ati-blacklist

thanks
Attachment #8518307 - Flags: approval-mozilla-release?
Attachment #8518307 - Flags: approval-mozilla-release+
Attachment #8518307 - Flags: approval-mozilla-beta?
Attachment #8518307 - Flags: approval-mozilla-beta+
Attachment #8518307 - Flags: approval-mozilla-aurora?
Attachment #8518307 - Flags: approval-mozilla-aurora+
(In reply to Arthur K. from comment #65)
> (In reply to Nicolas Silva [:nical] from comment #62)
> > Unless I missed something (data being scattered in a different places), all
> > reports coming from ATI drivers are in the versions ranging from 8.653 to
> > 8.690.0.0
> > 
> > Let's blacklist D2D on these configuration for now.
> > 
> > There's also the reports with dual nvidia+intel hardware that I need to go
> > though.
> 
> But without having some of these users with combo setups update one or both
> of the drivers, we'll never know which driver is causing it and might be
> blacklisting unnecessarily. Such as in the case of bug 1093863 comment 3. It
> *could* be that the reason that both device are present but one is not
> registered, for whatever reason, is because it might be just a case of a
> driver needing a reinstall or updated to re-register itself.

33.1 is going to build tonight and we need to unthrottle its release completely so at this point, better safe than sorry. I have a laptop that reproduces this issue so we'll continue investigating afterwards. There is a patch that solves the issue on the affected machine I have but the patch is way too risky to land in release (we aren't sure we want it at all).
https://hg.mozilla.org/mozilla-central/rev/e228bf3b5f02
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla36
I just tried the new Firefox 33.1 build (http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/33.1-candidates/build3/). On my computer the problem still occurs.
(In reply to testit from comment #72)
> I just tried the new Firefox 33.1 build
> (http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/33.1-candidates/
> build3/). On my computer the problem still occurs.

Could you please post the Graphics section of about:support?
This has only been blocked on some computers.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
(In reply to Arthur K. from comment #73)
> Could you please post the Graphics section of about:support?

Adapter Description: ATI Radeon HD 5450
Adapter Drivers: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM: 1024
Device ID: 0x68f9
Direct2D Enabled: Blocked for your graphics driver version. Try updating your graphics driver to version ATI Catalyst 14.6+ or newer.
DirectWrite Enabled: false (6.1.7600.16385)
Driver Date: 12-9-2009
Driver Version: 8.682.1.0
GPU #2 Active: false
GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
Vendor ID: 0x1002
WebGL Renderer: Google Inc. -- ANGLE (ATI Radeon HD 5450 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote: true
AzureCanvasBackend: skia
AzureContentBackend: cairo
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0
(In reply to testit from comment #75)
> (In reply to Arthur K. from comment #73)
> > Could you please post the Graphics section of about:support?
> 
> Adapter Description: ATI Radeon HD 5450
> Adapter Drivers: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32
> atiumdva atiumd6a atitmm64
> Adapter RAM: 1024
> Device ID: 0x68f9
> Direct2D Enabled: Blocked for your graphics driver version. Try updating
> your graphics driver to version ATI Catalyst 14.6+ or newer.
> DirectWrite Enabled: false (6.1.7600.16385)
> Driver Date: 12-9-2009
> Driver Version: 8.682.1.0
> GPU #2 Active: false
> GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
> Vendor ID: 0x1002
> WebGL Renderer: Google Inc. -- ANGLE (ATI Radeon HD 5450 Direct3D9Ex vs_3_0
> ps_3_0)
> windowLayerManagerRemote: true
> AzureCanvasBackend: skia
> AzureContentBackend: cairo
> AzureFallbackCanvasBackend: cairo
> AzureSkiaAccelerated: 0

Seeing as how your driver is from 2009, that likely explains why you are having this issue. Could you please try AMD's Autodetect tool and see if it helps update you to a newer version: http://support.amd.com/en-us/download/auto-detect-tool
(In reply to Arthur K. from comment #76)
> Seeing as how your driver is from 2009, that likely explains why you are
> having this issue. Could you please try AMD's Autodetect tool and see if it
> helps update you to a newer version:
> http://support.amd.com/en-us/download/auto-detect-tool

No sorry, I don't like to update the driver. I were surprised, that only the direct2d feature was disabled, but OMTC is still enabled. The latter causes the problem on my computer.
(In reply to testit from comment #77)
> (In reply to Arthur K. from comment #76)
> > Seeing as how your driver is from 2009, that likely explains why you are
> > having this issue. Could you please try AMD's Autodetect tool and see if it
> > helps update you to a newer version:
> > http://support.amd.com/en-us/download/auto-detect-tool
> 
> No sorry, I don't like to update the driver. I were surprised, that only the
> direct2d feature was disabled, but OMTC is still enabled. The latter causes
> the problem on my computer.

Then there is no problem.
(In reply to Arthur K. from comment #78)
> (In reply to testit from comment #77)
> > (In reply to Arthur K. from comment #76)
> > > Seeing as how your driver is from 2009, that likely explains why you are
> > > having this issue. Could you please try AMD's Autodetect tool and see if it
> > > helps update you to a newer version:
> > > http://support.amd.com/en-us/download/auto-detect-tool
> > 
> > No sorry, I don't like to update the driver. I were surprised, that only the
> > direct2d feature was disabled, but OMTC is still enabled. The latter causes
> > the problem on my computer.
> 
> Then there is no problem.

What do you mean by this?
(In reply to testit from comment #77)
> (In reply to Arthur K. from comment #76)
> > Seeing as how your driver is from 2009, that likely explains why you are
> > having this issue. Could you please try AMD's Autodetect tool and see if it
> > helps update you to a newer version:
> > http://support.amd.com/en-us/download/auto-detect-tool
> 
> No sorry, I don't like to update the driver. I were surprised, that only the
> direct2d feature was disabled, but OMTC is still enabled. The latter causes
> the problem on my computer.

The last drivers for your GPU are AMD Catalyst™ 14.9 from 9/29/2014. Don't be surprised to hit graphics issues with 5-year-old drivers.
(In reply to Loic from comment #80)
> (In reply to testit from comment #77)
> > (In reply to Arthur K. from comment #76)
> > > Seeing as how your driver is from 2009, that likely explains why you are
> > > having this issue. Could you please try AMD's Autodetect tool and see if it
> > > helps update you to a newer version:
> > > http://support.amd.com/en-us/download/auto-detect-tool
> > 
> > No sorry, I don't like to update the driver. I were surprised, that only the
> > direct2d feature was disabled, but OMTC is still enabled. The latter causes
> > the problem on my computer.
> 
> The last drivers for your GPU are AMD Catalyst™ 14.9 from 9/29/2014. Don't
> be surprised to hit graphics issues with 5-year-old drivers.

Judging by the line "6.1.7600.16385" in his about:support output, he's also on Win7 RTM, not SP1.
(In reply to Loic from comment #80)
> The last drivers for your GPU are AMD Catalyst™ 14.9 from 9/29/2014. Don't
> be surprised to hit graphics issues with 5-year-old drivers.

I'm not surprised about this. But I'm surprised about the way the bug was fixed. The title of the bug is "Enabling OMTC slows down the entire browser", so I don't understand, why the solution is disabling the direct2d feature but not OMTC.
I started getting this issue with the release of 33.0.0. I use Windows 7 64 bit and have a Radeon 5450 graphics card. I originally set OMTC to off and it did resolve the problem. I am writing this to hopefully help others with the same issue. There is a big difference between updating the graphics card and updating the Catalyst software. I tried to update my Radeon 5450 through device manager and it said there were no newer drivers available. I tried to download and install the Catalyst 14.9 update and ended up with a bsod and then had to roll back to the previous driver. I then found an article that explains how to update the Catalyst software. You must go to add/remove programs and uninstall the existing Catalyst install manager. You do not have to uninstall your graphics card driver. I was able to successfully update to Catalyst 14.9 at this point and then turn OMTC back on and everything works fine.
(In reply to Dimitris from comment #85)
> I also want to confirm this behavior. It started with version 33.0.x and is
> really annoying especially the long time gap between typing and the
> browser's response.

In you URL bar, please type about:support and kindly post the Graphics section.
(In reply to Arthur K. from comment #86)
> (In reply to Dimitris from comment #85)
> > I also want to confirm this behavior. It started with version 33.0.x and is
> > really annoying especially the long time gap between typing and the
> > browser's response.
> 
> In you URL bar, please type about:support and kindly post the Graphics
> section.

Graphics
--------

Adapter Description: ATI Mobility Radeon HD 5450 Series
Adapter Drivers: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM: 1024
Device ID: 0x68e0
Direct2D Enabled: true
DirectWrite Enabled: true (6.2.9200.16571)
Driver Date: 12-17-2009
Driver Version: 8.683.2.0
GPU #2 Active: false
GPU Accelerated Windows: 1/1 Direct3D 10
Vendor ID: 0x1002
WebGL Renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5450 Series Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote: false
AzureCanvasBackend: direct2d
AzureContentBackend: direct2d
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0
I've just come across this thread from clicking on Known Issue Windows: Interface may be slow when typing or selecting text (1089183) in Release Notes. Unfortunately I don't have any information on versions, but I have been having an intermittent problem which may be related for several years. I notice it specifically when editing a long Wikipedia article: typing becomes extremely slow, displaying character ... by ... character. It happened several times months and years ago, and was always resolved for a time by reinstalling the same version of Firefox, no change to extensions, just an update to the same version. The problem went away for a long time, and surfaced again in the past few days. I keep Firefox up-to-date, but don't always upgrade for things that don't affect me "resolved black screen starting Firefox with some cards", so it will be a 33.0.x, x=1 or 2, probably. My purpose for this "me-too" post is slightly different information: Win7/32, ATI RadeonXpress 1150 (R485M) card (not TurboXpress, for which links are given here to upgrade a Sony driver not upgradable from the ATI Web site), driver 8.593. I've tried a lot to get a later driver (modded or not), and this is the best I can do (it's a 2006 Dell Inspiron 1501 laptop, but works very well with more memory etc.).
P.S to mine (Michael Salem) from comment #90
Re a very old Radeon Xpress: the slowness was happening with 33.0.3; I just installed 33.1 and find it is also slow, immediately. This is unlike my previous experience: the slowness is of exactly the same nature, but happens on a newly-installed Firefox upgrade; in the past it developed over time. Using Wikipedia editing again, editing a long article is slow, but a short one is not. I haven't yet tried the fixes suggested here. I posted largely because drivers 8.650 - 8.682 were implicated, 8.593 is too.
(In reply to Dimitris from comment #89)
> (In reply to Arthur K. from comment #86)
> > (In reply to Dimitris from comment #85)
> > > I also want to confirm this behavior. It started with version 33.0.x and is
> > > really annoying especially the long time gap between typing and the
> > > browser's response.
> > 
> > In you URL bar, please type about:support and kindly post the Graphics
> > section.
> 
> Graphics
> --------
> 
> Adapter Description: ATI Mobility Radeon HD 5450 Series
> Adapter Drivers: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32
> atiumdva atiumd6a atitmm64
> Adapter RAM: 1024
> Device ID: 0x68e0
> Direct2D Enabled: true
> DirectWrite Enabled: true (6.2.9200.16571)
> Driver Date: 12-17-2009
> Driver Version: 8.683.2.0
> GPU #2 Active: false
> GPU Accelerated Windows: 1/1 Direct3D 10
> Vendor ID: 0x1002
> WebGL Renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5450 Series
> Direct3D9Ex vs_3_0 ps_3_0)
> windowLayerManagerRemote: false
> AzureCanvasBackend: direct2d
> AzureContentBackend: direct2d
> AzureFallbackCanvasBackend: cairo
> AzureSkiaAccelerated: 0

Can you try using AMD's Autodetect tool from bug 1083071 comment 194 and see if it helps you?
(In reply to Michael Salem from comment #90)
> I've just come across this thread from clicking on Known Issue Windows:
> Interface may be slow when typing or selecting text (1089183) in Release
> Notes. Unfortunately I don't have any information on versions, but I have
> been having an intermittent problem which may be related for several years.
> I notice it specifically when editing a long Wikipedia article: typing
> becomes extremely slow, displaying character ... by ... character. It
> happened several times months and years ago, and was always resolved for a
> time by reinstalling the same version of Firefox, no change to extensions,
> just an update to the same version. The problem went away for a long time,
> and surfaced again in the past few days. I keep Firefox up-to-date, but
> don't always upgrade for things that don't affect me "resolved black screen
> starting Firefox with some cards", so it will be a 33.0.x, x=1 or 2,
> probably. My purpose for this "me-too" post is slightly different
> information: Win7/32, ATI RadeonXpress 1150 (R485M) card (not TurboXpress,
> for which links are given here to upgrade a Sony driver not upgradable from
> the ATI Web site), driver 8.593. I've tried a lot to get a later driver
> (modded or not), and this is the best I can do (it's a 2006 Dell Inspiron
> 1501 laptop, but works very well with more memory etc.).

Can you try using AMD's Autodetect tool from bug 1083071 comment 194 and see if it helps you? Being that it is a 2006 laptop and the ATI card is in no way validated to run on Win 7, I don't have high hopes. It wasn't really meant to run on Win 7, modded or not.
(In reply to Arthur K. from comment #93)
> (In reply to Michael Salem from comment #90)
> > ...Win7/32, ATI RadeonXpress 1150 (R485M) card
> > ... driver 8.593 ... 2006 Dell Inspiron 1501 laptop

> Can you try using AMD's Autodetect tool ... I don't have high hopes. It wasn't really
> meant to run on Win 7, modded or not.

Thanks for response. Had already done this, several times I think, over the years. Did it again now: "you have the latest driver". Graphics hardware identified as 0x1002 - 0x5975PCI. Dell does not provide any support for Win7 on this machine, many drivers unofficial. But it works a treat...

BTW, just a one-off so far, but I've just had FF33.1 (with layers.prefer-d3d9 true, just changed; layers.offmainthreadcomposition.enabled unchanged for now) lock up VERY badly; responding to each down-arrow keypress browsing (not editing Web site text) took many seconds. Restart fixed it, for now. I've had occasional unresponsive FF over the years, requiring restart. Mentioned for completeness, only relevant if it's a common problem.
(In reply to Michael Salem from comment #94)
> (In reply to Arthur K. from comment #93)
> > (In reply to Michael Salem from comment #90)
> > > ...Win7/32, ATI RadeonXpress 1150 (R485M) card
> > > ... driver 8.593 ... 2006 Dell Inspiron 1501 laptop
> 
> > Can you try using AMD's Autodetect tool ... I don't have high hopes. It wasn't really
> > meant to run on Win 7, modded or not.
> 
> Thanks for response. Had already done this, several times I think, over the
> years. Did it again now: "you have the latest driver". Graphics hardware
> identified as 0x1002 - 0x5975PCI. Dell does not provide any support for Win7
> on this machine, many drivers unofficial. But it works a treat...
> 
> BTW, just a one-off so far, but I've just had FF33.1 (with
> layers.prefer-d3d9 true, just changed;
> layers.offmainthreadcomposition.enabled unchanged for now) lock up VERY
> badly; responding to each down-arrow keypress browsing (not editing Web site
> text) took many seconds. Restart fixed it, for now. I've had occasional
> unresponsive FF over the years, requiring restart. Mentioned for
> completeness, only relevant if it's a common problem.

What it boils down to is that we're now revealing the limitations of your hardware, esp. with the OMTC work, and these old drivers that don't support all the functions needed to make things work properly. I hesitate to ask but did you try the FLEM 2.0 Modding tool to roll your own Modded drivers? Might be your only option.
(In reply to Arthur K. from comment #95)
> (In reply to Michael Salem from comment #94)
> > (In reply to Arthur K. from comment #93)
> > > (In reply to Michael Salem from comment #90)
> > > > ...Win7/32, ATI RadeonXpress 1150 (R485M) card
> > > > ... driver 8.593 ... 2006 Dell Inspiron 1501 laptop
> > 
> > > Can you try using AMD's Autodetect tool ... I don't have high hopes. It wasn't really
> > > meant to run on Win 7, modded or not.
> > 
> > Thanks for response. Had already done this, several times I think, over the
> > years. Did it again now: "you have the latest driver". Graphics hardware
> > identified as 0x1002 - 0x5975PCI. Dell does not provide any support for Win7
> > on this machine, many drivers unofficial. But it works a treat...
> > 
> > BTW, just a one-off so far, but I've just had FF33.1 (with
> > layers.prefer-d3d9 true, just changed;
> > layers.offmainthreadcomposition.enabled unchanged for now) lock up VERY
> > badly; responding to each down-arrow keypress browsing (not editing Web site
> > text) took many seconds. Restart fixed it, for now. I've had occasional
> > unresponsive FF over the years, requiring restart. Mentioned for
> > completeness, only relevant if it's a common problem.
> 
> What it boils down to is that we're now revealing the limitations of your
> hardware, esp. with the OMTC work, and these old drivers that don't support
> all the functions needed to make things work properly. I hesitate to ask but
> did you try the FLEM 2.0 Modding tool to roll your own Modded drivers? Might
> be your only option.

Thanks again. I was having some issues with video, but they were resolved by downloading a ready-modded driver. Basically, everything is working as well as I expect (the processor is a dual-core 1.6GHz Turion X2, not too bad; SSD helps). I can do everything I need perfectly well (I have a faster desktop machine for processor-intensive work); and I can edit long Wikipedia articles quite happily with Opera and Chrome (according to a single test today); I have just discovered that IE has the same text-entry sluggishness as FF33.1. So I don't want, or need, to do anything; if FF doesn't work, I can use another browser, or try the fixes suggested. I'm surprised that the machine still works so well, I'd have expected to need a change after 4-5 years (I bought it exactly 8 years ago). Essentially I was trying to contribute to this discussion (there weren't many reports about Win7, and none about a driver this old), rather than seek help.

I had looked at FLEM a while ago, but decided not to make changes to a basically working system.
(In reply to Nicolas Silva [:nical] from comment #51)
> Let's clarify things a bit.
> We encourage people to update their drivers because on average, recent
> driver versions work better than older ones. This happens to fix the issue
> but we know it is not THE solution to get a good (heck, usable) web
> experience to all of the users. Deactivating OMTC works around some problems
> but we also know that turning OMTC off causes other bugs so I wouldn't
> encourage people to do it too much.
> We are working on finding solutions to this bug, even for older drivers. It
> will probably involve some blacklisting but to do this we need to understand
> exactly what feature we blacklist on what hardware and what driver versions,
> so it takes a bit of investigating. The issue will be fixed, and sooner
> rather than later.
> Now, if you read this comment and have old drivers, you should know that
> updating them has a good chance of helping you while we are working on a
> solution on our end.
> I hope that this makes things a bit clearer about we are going about this
> and the other driver-related issues we are having with this release.

All the last comments are about updating graphics drivers but for me this goes into the wrong direction. Is someone working on really SOLVING the problem?
(In reply to testit from comment #98)
> (In reply to Nicolas Silva [:nical] from comment #51)
> > Let's clarify things a bit.
> > We encourage people to update their drivers because on average, recent
> > driver versions work better than older ones. This happens to fix the issue
> > but we know it is not THE solution to get a good (heck, usable) web
> > experience to all of the users. Deactivating OMTC works around some problems
> > but we also know that turning OMTC off causes other bugs so I wouldn't
> > encourage people to do it too much.
> > We are working on finding solutions to this bug, even for older drivers. It
> > will probably involve some blacklisting but to do this we need to understand
> > exactly what feature we blacklist on what hardware and what driver versions,
> > so it takes a bit of investigating. The issue will be fixed, and sooner
> > rather than later.
> > Now, if you read this comment and have old drivers, you should know that
> > updating them has a good chance of helping you while we are working on a
> > solution on our end.
> > I hope that this makes things a bit clearer about we are going about this
> > and the other driver-related issues we are having with this release.
> 
> All the last comments are about updating graphics drivers but for me this
> goes into the wrong direction. Is someone working on really SOLVING the
> problem?

Old graphics drivers that don't support current functions appears to *be* the problem.
(In reply to Arthur K. from comment #99)
> (In reply to testit from comment #98)
> > (In reply to Nicolas Silva [:nical] from comment #51)
> > > Let's clarify things a bit.
> > > We encourage people to update their drivers because on average, recent
> > > driver versions work better than older ones. This happens to fix the issue
> > > but we know it is not THE solution to get a good (heck, usable) web
> > > experience to all of the users. Deactivating OMTC works around some problems
> > > but we also know that turning OMTC off causes other bugs so I wouldn't
> > > encourage people to do it too much.
> > > We are working on finding solutions to this bug, even for older drivers. It
> > > will probably involve some blacklisting but to do this we need to understand
> > > exactly what feature we blacklist on what hardware and what driver versions,
> > > so it takes a bit of investigating. The issue will be fixed, and sooner
> > > rather than later.
> > > Now, if you read this comment and have old drivers, you should know that
> > > updating them has a good chance of helping you while we are working on a
> > > solution on our end.
> > > I hope that this makes things a bit clearer about we are going about this
> > > and the other driver-related issues we are having with this release.
> > 
> > All the last comments are about updating graphics drivers but for me this
> > goes into the wrong direction. Is someone working on really SOLVING the
> > problem?
> 
> Old graphics drivers that don't support current functions appears to *be*
> the problem.

The question is not what the problem is but what the solution is. 

> We are working on finding solutions to this bug, even for older drivers.

As stated by Nicolas the solution is NOT that everybody which has got problems updates the drivers.
(In reply to Arthur K. from comment #92)
> (In reply to Dimitris from comment #89)
> > (In reply to Arthur K. from comment #86)
> > > (In reply to Dimitris from comment #85)
> > > > I also want to confirm this behavior. It started with version 33.0.x and is
> > > > really annoying especially the long time gap between typing and the
> > > > browser's response.
> > > 
> > > In you URL bar, please type about:support and kindly post the Graphics
> > > section.
> > 
> > Graphics
> > --------
> > 
> > Adapter Description: ATI Mobility Radeon HD 5450 Series
> > Adapter Drivers: atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32
> > atiumdva atiumd6a atitmm64
> > Adapter RAM: 1024
> > Device ID: 0x68e0
> > Direct2D Enabled: true
> > DirectWrite Enabled: true (6.2.9200.16571)
> > Driver Date: 12-17-2009
> > Driver Version: 8.683.2.0
> > GPU #2 Active: false
> > GPU Accelerated Windows: 1/1 Direct3D 10
> > Vendor ID: 0x1002
> > WebGL Renderer: Google Inc. -- ANGLE (ATI Mobility Radeon HD 5450 Series
> > Direct3D9Ex vs_3_0 ps_3_0)
> > windowLayerManagerRemote: false
> > AzureCanvasBackend: direct2d
> > AzureContentBackend: direct2d
> > AzureFallbackCanvasBackend: cairo
> > AzureSkiaAccelerated: 0
> 
> Can you try using AMD's Autodetect tool from bug 1083071 comment 194 and see
> if it helps you?

If I were you I would be very careful about suggesting users to update their AMD drivers with this tool especially to the latest AMD Catalyst version 14. The AHCI driver (SATA driver) is updated too and there is a good chance users are forced to revert back to the old version since the new one is buggy and causes Windows to stuck at the loading screen. My personal experience, I spent half a day to revert back and have Windows boot again.
I find amusing that every now and then, we have to deal with GPU issues with new versions of Firefox... Not to compare, but hrome and Opera seem to handle this pretty well... Why not Firefox?...

But well, to the possible solutions, I feel like you are going for a very generic thing... why not relying on something a bit more specific and even faster, like OpenGL? OpenGL is used even on Windows and works fast as hell... if OMTC can be based on OpenGL this bugs would be history in my opinion.

Why I ask this? Playing with the values for this thing under the layers category, I find that enabling .prefer.opengl or .prefer.d3d9 fixes all the slowdown you get with the default OMTC on... Now, it has bugs though... if you enable opengl, you have bugs with aero (Windows 7) and transparencies (Title bar and tab edges mainly), if you enable d3d9, works better, but the window buttons (minimize, maximize/restore and close buttons) seem to be overlapped by the title bar grey color, but is nothing mayor. If any of this bugs can be fixed (specially for OpenGL), I bet all this drama will be gone.

Upgrading my graphics driver is not an option, unless is ABSOLUTELY necessary, just because my vendor customized some setings to keep the system stable (tends to overheat with generic AMD drivers)... By the other hand, is kinda funny that you have to upgrade your drivers with every Firefox update, when Chrome, Opera or even Internet Explorer don't need that...

I hope you take this in a good way, Firefox is such a nice browser, and this "driver" problems, as the old D2D text rendering we got before have better solutions that just "update your drivers and let me go on"... If you need time, take your time, at least we got a workaround to survive while you fix this, it's not a big deal. But please, stop forcing people to upgrade drivers because of this optimizaions that, I hope, are for making the browser faster, or for supporting something critical, because for cosmetical duties, this is kinda pricey...

Anyway, thank you for your work and new features, but take your time, good stuff needs it, and Firefox is quality, not quantity.

My Specs:
AMD Athlon X2-QL 65 2.1 Ghz
ATI (Mobility) Radeon HD 3200 (256 MB) Driver Ver. 8.653
4 GB RAM Chipset AMD SB7xx

Hope you can solve this with success, again, take your time and don't rush yourselves people, just do it right, Firefox deserves it ;)
I'm not on my computer right now, but my nvidia driver is updated to the last version and for some reason I'm unable to update my intel driver, which seems to be the one Firefox is using. Windows (8.1) installs an outdated version of it and after I try to update, the updater shows I already have the last version while the Intel control center, Firefox and even Windows control panel shows the outdated one. So, updating is not an option for me also.
(In reply to Rafael from comment #103)
> I'm not on my computer right now, but my nvidia driver is updated to the
> last version and for some reason I'm unable to update my intel driver, which
> seems to be the one Firefox is using. Windows (8.1) installs an outdated
> version of it and after I try to update, the updater shows I already have
> the last version while the Intel control center, Firefox and even Windows
> control panel shows the outdated one. So, updating is not an option for me
> also.

You should be getting your drivers from NVidia and Intel and not from Microsoft. Could you please type about:support into the URL bar and past the Graphics portion here?
Yes, there are people working on this. Yes, we want to find a workaround so that firefox works on buggy drivers, and yes it's a high priority. It's also not an easy problem to fix. Can we please stop spamming bugzilla with these discussions, it makes bugs like this one hard to navigate for the devs who need to go through the many comments to find the information that is needed to investigate the bug.
Other browsers also occasionally have issues like this one. It's just not the same people who run into problems on all browsers. It would be helpful if we keep bugzilla comments limited to stuff that helps with fixing bugs. I am sorry that this bug is still around and that it is so frustrating for affected users.
(In reply to Nicolas Silva [:nical] from comment #105)
> Yes, there are people working on this. Yes, we want to find a workaround so
> that firefox works on buggy drivers, and yes it's a high priority. It's also
> not an easy problem to fix. Can we please stop spamming bugzilla with these
> discussions, it makes bugs like this one hard to navigate for the devs who
> need to go through the many comments to find the information that is needed
> to investigate the bug.
> Other browsers also occasionally have issues like this one. It's just not
> the same people who run into problems on all browsers. It would be helpful
> if we keep bugzilla comments limited to stuff that helps with fixing bugs. I
> am sorry that this bug is still around and that it is so frustrating for
> affected users.

I guess you're right... what about the graphic "engine"?... I don't know how are you doing it right now but, maybe focusing on OpenGL could help you get rid of all the bugs?... Just wondering.

Anyway, thanks for the hard work, and again, take your time, I'm aware this are things that cannot be fixed in 5 minutes, and that are complex as hell XD.
(In reply to ferline2000mx from comment #102)
> I find amusing that every now and then, we have to deal with GPU issues with
> new versions of Firefox... Not to compare, but Chrome and Opera seem to
> handle this pretty well... Why not Firefox?...

I've been using firefox for 10 years ;) Never see that kind of trouble before. Hope it will be resolve for Cristmas :) Good Luck Firefox team, I stay behind you.

Now, the textArea scrolling turns black when I disable my OMTC...
(In reply to Arthur K. from comment #104)
> (In reply to Rafael from comment #103)
> > I'm not on my computer right now, but my nvidia driver is updated to the
> > last version and for some reason I'm unable to update my intel driver, which
> > seems to be the one Firefox is using. Windows (8.1) installs an outdated
> > version of it and after I try to update, the updater shows I already have
> > the last version while the Intel control center, Firefox and even Windows
> > control panel shows the outdated one. So, updating is not an option for me
> > also.
> 
> You should be getting your drivers from NVidia and Intel and not from
> Microsoft. Could you please type about:support into the URL bar and past the
> Graphics portion here?

I have attached this information to the bug already, but here you go:

Adapter Description: Intel(R) HD Graphics 4600
Adapter Description (GPU #2): NVIDIA GeForce GTX 765M
Adapter Drivers: igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter Drivers (GPU #2): nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter RAM: Unknown
Adapter RAM (GPU #2): 2048
ClearType Parameters: Gamma: 2200 Pixel Structure: R
Device ID: 0x0416
Device ID (GPU #2): 0x11e2
Direct2D Enabled: true
DirectWrite Enabled: true (6.3.9600.17111)
Driver Date: 9-29-2014
Driver Date (GPU #2): 11-3-2014
Driver Version: 10.18.10.3960
Driver Version (GPU #2): 9.18.13.4465
GPU #2 Active: false
GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
Vendor ID: 0x8086
Vendor ID (GPU #2): 0x10de
WebGL Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics 4600 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d
AzureContentBackend: direct2d
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0

Also, I do update my drivers from nvidia and intel, but the intel one just refuses to be updated. I have Intel Driver Update Utility 2.0 showing me that I have the last version (15.36.764.3960) while the rest shows the opposite (10.18.10.3960). Maybe it really is updated and it just shows the wrong number, but still, the driver is recent, from 29/September/2014.
(In reply to alex from comment #107)
> I've been using firefox for 10 years ;) Never see that kind of trouble
> before. Hope it will be resolve for Cristmas :) Good Luck Firefox team, I
> stay behind you.
> 
> Now, the textArea scrolling turns black when I disable my OMTC...

Do you already use Firefox 33.1? If so, I guess you hit bug 1090338. The only thing which helped for me was disabling hardware acceleration.
I'm also a big Firefox fan, but starting with version 33.0 it's really not much fun using Firefox, it's getting worse from release to release.
I'd repeat something I said before just in case it's significant and was missed (I didn't emphasise it). Environment: very old ATI Xpress (Directx 9.0c, OpenGL 2.1) on laptop, AMD CPU, Win7/32; full details in earlier posting. The various about:config fixes suggested didn't help. I am consistently having very slow text entry with FF, all recent versions, don't remember exactly with which it started (a 32.0.x), which seems to match what others say.

In much earlier versions (I don't remember which) I occasionally got EXACTLY the same symptoms some considerable time after I updated. They persisted, but went away when I reinstalled the same version, and usually stayed away until a few versions later. If this fact is of any use, please use it; but there's no need either to respond to me or to discuss it if it is not useful. Best wishes
Is it possible to show the slowness in a screencapture? i.e. Perhaps before and after FF 33?
Nical has a laptop with the bug. Maybe he can help here.
Flags: needinfo?(nical.bugzilla)
(In reply to Michael Salem from comment #110)
> I'd repeat something I said before just in case it's significant and was
> missed (I didn't emphasise it). Environment: very old ATI Xpress (Directx
> 9.0c, OpenGL 2.1) on laptop, AMD CPU, Win7/32; full details in earlier
> posting. The various about:config fixes suggested didn't help. I am
> consistently having very slow text entry with FF, all recent versions, don't
> remember exactly with which it started (a 32.0.x), which seems to match what
> others say.
> 
> In much earlier versions (I don't remember which) I occasionally got EXACTLY
> the same symptoms some considerable time after I updated. They persisted,
> but went away when I reinstalled the same version, and usually stayed away
> until a few versions later. If this fact is of any use, please use it; but
> there's no need either to respond to me or to discuss it if it is not
> useful. Best wishes

This is likely a different problem if you saw it in 32. Perhaps bug 1092260.
Can everyone who's experiencing this post the make and model of the computer they're seeing it on?
Graphics info posted previously in this thread, so here's my computer info:

Dell Studio XPS 7100 (Win7 Home Premium SP1 64-bit)

Can provide additional info if desired.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #114)
> Can everyone who's experiencing this post the make and model of the computer
> they're seeing it on?

Dell Inspiron 580.
Dell Inspiron 1564 (early 2010). AMD ATI Mobility Radeon HD driver version 8.692.1.0000. FYI, BSOD when tried the suggested catalyst update to current version 14. 

To cut a long story short, FF is inoperable.
I confirm hardly operable browser since version 33.0. The latest version added brief appearances of black areas during page scroll. Tried to update the driver via device manager. Got the message "latest is already installed".


Version 	33.1
User Agent 	Mozilla/5.0 (Windows NT 6.1; WOW64; rv:33.0) Gecko/20100101 Firefox/33.0

Toshiba Satellite L505

Adapter Description	ATI Mobility Radeon HD 5165
Adapter Drivers	atiumd64 atidxx64 atiumdag atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM	1024
Device ID	0x9480
Direct2D Enabled	Blocked for your graphics driver version. Try updating your graphics driver to version 10.6 or newer.
DirectWrite Enabled	false (6.2.9200.16571)
Driver Date	9-8-2009
Driver Version	8.653.0.0
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Vendor ID	0x1002
WebGL Renderer	Google Inc. -- ANGLE (ATI Mobility Radeon HD 5165 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote	true
AzureCanvasBackend	skia
AzureContentBackend	cairo
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0
We have a build that we hope fixes this issue available here:
http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-inbound-win32/1416005448/

If everyone who's experience the bug could test that out to see if it resolves the issue that would be much appreciated.
(In reply to tinfire from comment #118)
> Direct2D Enabled	Blocked for your graphics driver version. Try updating your
> graphics driver to version 10.6 or newer.

Could you please try the suggestion mention in the about:support report and see if ti reolves?
(In reply to tinfire from comment #118)
> Direct2D Enabled	Blocked for your graphics driver version. Try updating your
> graphics driver to version 10.6 or newer.

Could you please try the suggestion mention in the about:support report and see if it resolves?(In reply to Arthur K. from comment #120)
> (In reply to tinfire from comment #118)
> > Direct2D Enabled	Blocked for your graphics driver version. Try updating your
> > graphics driver to version 10.6 or newer.
> 
> Could you please try the suggestion mention in the about:support report and
> see if ti reolves?

If it resolves.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #119)
> We have a build that we hope fixes this issue available here:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-
> inbound-win32/1416005448/
> 
> If everyone who's experience the bug could test that out to see if it
> resolves the issue that would be much appreciated.

Installed the recommended Nightly. After brief testing found that the problems still persist. Namely luggish keyboard typing and mouse hover responses and command actions usually not shown until next command is performed.This build crashes yahoo.com.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #119)
> We have a build that we hope fixes this issue available here:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-
> inbound-win32/1416005448/
> 
> If everyone who's experience the bug could test that out to see if it
> resolves the issue that would be much appreciated.

Installed the nightly. It crashes immediately after start.
Can you post the crash report url from about:crashes?
(In reply to Jeff Muizelaar [:jrmuizel] from comment #114)
> Can everyone who's experiencing this post the make and model of the computer
> they're seeing it on?

My notebook is not from a big brand, it's Avell G1511 NEW, but I mentioned it's graphics properties in comment #108.

(In reply to Jeff Muizelaar [:jrmuizel] from comment #119)
> We have a build that we hope fixes this issue available here:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-
> inbound-win32/1416005448/
> 
> If everyone who's experience the bug could test that out to see if it
> resolves the issue that would be much appreciated.

That solved it for me, apparently. For the ones it didn't, this is basically firefox nightly build with some changes, so disabled e10s (File -> New Non-e10s Window) and try again.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #124)
> Can you post the crash report url from about:crashes?

https://crash-stats.mozilla.com/report/index/ce06be62-ca78-4d7d-9068-70bd12141115
Comment on attachment 8512080 [details]
Information in about:support

>Application Basics
>------------------
>
>Name: Firefox
>Version: 33.0.1
>User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:33.0) Gecko/20100101 Firefox/33.0
>Multiprocess Windows: 0/1
>
>Crash Reports for the Last 3 Days
>---------------------------------
>
>All Crash Reports (including 1 pending crash in the given time range)
>
>Extensions
>----------
>
>Name: Reddit Enhancement Suite
>Version: 4.5.0.2
>Enabled: true
>ID: jid1-xUfzOsOFlzSOXg@jetpack
>
>Name: Online Accounts Extension
>Version: 1.0
>Enabled: false
>ID: {41ecbc0b-34d5-4cd4-935f-253a30e2cb7e}
>
>Name: Online Accounts Extension
>Version: 1.1
>Enabled: false
>ID: {d4da7309-b89a-45ec-8ebb-cfb2ae13618b}
>
>Name: Vertical Tabs
>Version: 0.10b1
>Enabled: false
>ID: verticaltabs@philikon.de
>
>Graphics
>--------
>
>Adapter Description: Intel(R) HD Graphics 4600
>Adapter Description (GPU #2): NVIDIA GeForce GTX 765M
>Adapter Drivers: igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
>Adapter Drivers (GPU #2): nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
>Adapter RAM: Unknown
>Adapter RAM (GPU #2): 2048
>ClearType Parameters: Gamma: 2200 Pixel Structure: R
>Device ID: 0x0416
>Device ID (GPU #2): 0x11e2
>Direct2D Enabled: true
>DirectWrite Enabled: true (6.3.9600.17111)
>Driver Date: 1-22-2014
>Driver Date (GPU #2): 10-16-2014
>Driver Version: 10.18.10.3412
>Driver Version (GPU #2): 9.18.13.4448
>GPU #2 Active: false
>GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
>Vendor ID: 0x8086
>Vendor ID (GPU #2): 0x10de
>WebGL Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics 4600 Direct3D9Ex vs_3_0 ps_3_0)
>windowLayerManagerRemote: true
>AzureCanvasBackend: direct2d
>AzureContentBackend: direct2d
>AzureFallbackCanvasBackend: cairo
>AzureSkiaAccelerated: 0
>
>Important Modified Preferences
>------------------------------
>
>accessibility.typeaheadfind.flashBar: 0
>browser.cache.disk.capacity: 358400
>browser.cache.disk.smart_size.first_run: false
>browser.cache.disk.smart_size.use_old_max: false
>browser.cache.frecency_experiment: 3
>browser.places.smartBookmarksVersion: 7
>browser.sessionstore.upgradeBackup.latestBuildID: 20141023194920
>browser.startup.homepage_override.buildID: 20141023194920
>browser.startup.homepage_override.mstone: 33.0.1
>dom.mozApps.used: true
>extensions.lastAppVersion: 33.0.1
>gfx.direct3d.last_used_feature_level_idx: 0
>media.gmp-gmpopenh264.lastUpdate: 1414250652
>media.gmp-gmpopenh264.version: 1.1
>media.gmp-manager.lastCheck: 1414403895
>network.cookie.prefsMigrated: true
>places.database.lastMaintenance: 1414420934
>places.history.expiration.transient_current_max_pages: 104858
>plugin.disable_full_page_plugin_for_types: application/pdf
>plugin.importedState: true
>privacy.sanitize.migrateFx3Prefs: true
>storage.vacuum.last.index: 1
>storage.vacuum.last.places.sqlite: 1413735297
>
>Important Locked Preferences
>----------------------------
>
>JavaScript
>----------
>
>Incremental GC: true
>
>Accessibility
>-------------
>
>Activated: false
>Prevent Accessibility: 0
>
>Library Versions
>----------------
>
>NSPR
>Expected minimum version: 4.10.7
>Version in use: 4.10.7
>
>NSS
>Expected minimum version: 3.17.1 Basic ECC
>Version in use: 3.17.1 Basic ECC
>
>NSSSMIME
>Expected minimum version: 3.17.1 Basic ECC
>Version in use: 3.17.1 Basic ECC
>
>NSSSSL
>Expected minimum version: 3.17.1 Basic ECC
>Version in use: 3.17.1 Basic ECC
>
>NSSUTIL
>Expected minimum version: 3.17.1
>Version in use: 3.17.1
>
>Experimental Features
>---------------------
(In reply to Rafael from comment #125)
> That solved it for me, apparently. For the ones it didn't, this is basically
> firefox nightly build with some changes, so disabled e10s (File -> New
> Non-e10s Window) and try again.

I'm not able to disable e10 because the browser crashes immediately. There is way to change the configuration using a configuration file. Any hints how this works?
(In reply to testit from comment #126)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #124)
> > Can you post the crash report url from about:crashes?
> 
> https://crash-stats.mozilla.com/report/index/ce06be62-ca78-4d7d-9068-
> 70bd12141115

This actually doesn't have symbols. Can you try regular firefox nightly and see if that crashes for you? If it does the url for that crash report will be more useful.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #124)
> Can you post the crash report url from about:crashes?

I do not find a related crash report in the log. Nevertheless - yahoo.com page loads and after scrolling down to the bottom of the page it goes for refreshment and never returns.The browser dims the page and blocks the UI. If it is of any help I could post the Windows report generated after force-closing FF.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #129)
> (In reply to testit from comment #126)
> > (In reply to Jeff Muizelaar [:jrmuizel] from comment #124)
> > > Can you post the crash report url from about:crashes?
> > 
> > https://crash-stats.mozilla.com/report/index/ce06be62-ca78-4d7d-9068-
> > 70bd12141115
> 
> This actually doesn't have symbols. Can you try regular firefox nightly and
> see if that crashes for you? If it does the url for that crash report will
> be more useful.

The regular nightly doesn't crash.
(In reply to tinfire from comment #130)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #124)
> > Can you post the crash report url from about:crashes?
> 
> I do not find a related crash report in the log. Nevertheless - yahoo.com
> page loads and after scrolling down to the bottom of the page it goes for
> refreshment and never returns.The browser dims the page and blocks the UI.
> If it is of any help I could post the Windows report generated after
> force-closing FF.

I'm wondering if you're perhaps experiencing bug 1092260. Does the problem happen in IE 11 for example after you see it Firefox?
(In reply to Jeff Muizelaar [:jrmuizel] from comment #119)
> We have a build that we hope fixes this issue available here:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-
> inbound-win32/1416005448/
Downloaded and installed firefox-36.0a1.en-US.win32.installer.exe from there in addition to my en-GB 33.1; no change, exactly the same form of slowness. My details are in other posts: very old Radeon Xpress, Win7/32, Dell Inspiron 1501 (AMD CPU), no newer driver available. I don't have an image of the slowness, but when editing the middle of a long Wikipedia article I type characters; maybe a few appear; I keep typing, nothing, suddenly a few more characters appear, always lagging my typing. After I finish typing the remainder always eventually appears, nothing waits forever. I'm definitely running the nightly version (per About); it has picked up all the add-ons, history, etc. HTH, no response required.
Further to my comment of a moment ago: v33.1.1 has just appeared, and I have installed it. (en-GB). Same equipment always. I think it is slightly WORSE than 33.1: the slowness is unchanged as far as I can tell, but the text area (editing always the same Wikipedia article) suddenly blanks after typing every few characters, then reappears. Good news: making progress. Bad news: backwards.
(In reply to Michael Salem from comment #134)
> Further to my comment of a moment ago: v33.1.1 has just appeared, and I have
> installed it. (en-GB). Same equipment always. I think it is slightly WORSE
> than 33.1: the slowness is unchanged as far as I can tell, but the text area
> (editing always the same Wikipedia article) suddenly blanks after typing
> every few characters, then reappears. Good news: making progress. Bad news:
> backwards.

Just to confirm, do you have the same problem with FF32? If you do then your bug is not this one and we can continue looking into the problem in bug 1092260.
Flags: needinfo?(mgs)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #135)
> do you have the same problem with FF32?
I will downgrade temporarily (presumably this won't leave any old settings that will confuse the issue? From others' reports add-ons don't seem relevant). Which specific version should I use? There are several 32.x; also, does the specific language version matter, and which would you prefer me to use (not traditional Chinese please, I don't know what buttons to press)?
Flags: needinfo?(mgs)
(In reply to Michael Salem from comment #136)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #135)
> > do you have the same problem with FF32?
> I will downgrade temporarily (presumably this won't leave any old settings
> that will confuse the issue? From others' reports add-ons don't seem
> relevant). Which specific version should I use? There are several 32.x;
> also, does the specific language version matter, and which would you prefer
> me to use (not traditional Chinese please, I don't know what buttons to
> press)?

Running any release from here https://ftp.mozilla.org/pub/mozilla.org/firefox/releases/32.0.3/win32/ should be fine. You should be able to just extract the a zip file and run the binary from there.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #137)
> (In reply to Michael Salem from comment #136)
> > (In reply to Jeff Muizelaar [:jrmuizel] from comment #135)

> Running any release from here
> https://ftp.mozilla.org/pub/mozilla.org/firefox/releases/32.0.3/win32/
> should be fine. You should be able to just extract the a zip file and run
> the binary from there.

I ran 32.0.3 en-GB, that I already had stored, from a RAM drive. Confirmed version with About. Problem the same as before, except that 33.x also blanks edit area of screen (I'll recheck that 32.0.3 doesn't , will report if it does, otherwise assume it doesn't). I mentioned problem with 32.x my much earlier contributions to this thread, I think, though how you'd keep track I don't know. Best wishes
I also installed v.32.0.3. On a stick. And I see no problems with the responsiveness with this version.
As there are differences in the reports concerning the bug I would suggest a simple test for what I see (one of the many effect).
1. Open a suitable page and doubleclick a word in the text paying attention not move the mouse.
2. nothing happens the pointer still being idle over the clicked word.
3. Move the pointer away from the word.
4. The words highlights with some delay thus showing the default action. (for 2.)
Can anyone who can reproduce this problem on AMD hardware check to see if they see a similar in IE11?
For context, I've been running FF 33.1 (just saw need to update to 33.1.1) with h/w accel turned OFF in options, since the 33.x lag problem affects me otherwise (as described in earlier posts).  I had tried IE before and had no problems, but I just tried it again (ver 11.0.9600.17420), also no problems.  My computer/graphics info is listed in my earlier posts, but to recap it's an ATI Radeon HD 5670, driver ver 8.690.0.0.
Flags: needinfo?(nical.bugzilla)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #140)
> Can anyone who can reproduce this problem on AMD hardware check to see if
> they see a similar in IE11?

My test shows there is no such problem in IE 11.0 with Radeon HD 5156, driver ver.8.653.0.0
I would reccomend to people testing the modified Nightly build, to disable addons or to try it on a clean install with no addons if possible, backup your infor and all with MozBackup, uninstall entirely your current build including your user preferences and install the nightly build... Remember that, betas can have issues with addons and stuff not yet tested, that is not target to fix currently.

Then, try the Nightly build and see if you still have issues with it.

By my side, I contacted Acer for my particular driver updates, and after pushing them from their confort/lazy zone, they finally said that drivers from 2012 and beyond are compatible with Acer hardware (ATI Radeons HD 3200 and up, probably other models, from 2008 and up), and that AMD optimized some modules to keep them running cool and smooth on a mobile system. So, I installed Catalyst 12.4 I had that one saved from 2012 just in case...), version 8.971 (IIRC), all bugs are gone, obviously. If you are on an Acer system, and need to upgrade, but Acer is LAZY as hell to release a more up-to-date driver, try out the driver detector from AMD and install, it will most likely work on your system wit no issues.

For Intel Based mobile PC's, it would be the same, as Intel is VERY generic driver-wise...

For those that cannot upgrade, try the Nightly without any addon, pure clean install and see if the problem persists... let's just hope that a good solution comes from OpenGL and we can get rid of all this DX **** nonsense...
Here's a build that I think might fix the issue, especially for those with AMD drivers:
http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/jmuizelaar@mozilla.com-815f50bb7696/try-win32/

Let me know if it works.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #144)
> Here's a build that I think might fix the issue, especially for those with
> AMD drivers:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/jmuizelaar@mozilla.
> com-815f50bb7696/try-win32/
> 
> Let me know if it works.

My Firefox browser shines again. Thank you, Jeff. You are the man!

I installed the nightly version and found non of the issues I experienced with 33.0. Namely sluggish scrolling with appearing black areas, retarded mouse hover and select actions and partial text selection. 

The test was done with the nightly started at a new clean profile.

May be my list of add-ons has some other problem with this nightly as, if started from my usual profile, it still crashes yahoo.com (this was an issue I reported in an earlier post).

Nevertheless I greatly appreciate your persistence in resolving the issue.
(In reply to Jeff Muizelaar [:jrmuizel] from comment #144)
> Here's a build that I think might fix the issue, especially for those with
> AMD drivers:
> http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/jmuizelaar@mozilla.
> com-815f50bb7696/try-win32/
> 
> Let me know if it works.

Jeff, I'd like to test to contribute to the solution, but am concerned about risk to my system/data.  Do you consider this build to be safe for a non-developer to install and test on his primary working computer?
Response to #144 (and relevant to #146).
I downloaded the Windows ZIP file from http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/jmuizelaar@mozilla.com-815f50bb7696/try-win32/ and extracted it into a temporary directory, then clicked on firefox.exe to run it, as advised previously,, without installing, or deleting my previous installation. After testing my original installation worked fine (answer tp Bruce Lott).

I have posted several times, won't repeat all details to reduce clutter (but can repost if requested). Basically, very old machine, Radeon Xpress, DirectX 9c, OpenGL 2.1, no newer driver available. Intermittently have had the slowness I describe here over the years; until v32.0.x(2?) cured by "upgrading" Firefox to the same version. Now the problem is permanent, possibly slightly worse with 815f50bb7696. Editing a long Wikipedia article (it's OK on short ones) I typed some characters; they appeared suddenly about 10" (an eternity) after I typed them. I also noted some hanging with v33.1 (and possibly other recent versions) for a few seconds when clicking on a different tab, which In the past never happened. I tried changing some about:config settings, which made no difference. This may not be the problem mainly discussed in this thread (is it a single issue?). If my comments are not useful, please say; I'm trying to contribute, not confuse.
I just installed FF 34b10 (http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/34.0b10-candidates/). The problem seems to be solved.
I am running FF 34.0.5 as of today. The browser is really slow when typing in a URL or a search term. 
After typing a 3-letter word, the into the URL bar, the search function only seemed to recognize the first letter. I had to append a space or move the mouse in order for all 3 letter to appear. Typing something in a text field inside the browser is also very slow.
After setting layers.offmainthreadcomposition.enabled to false and restarting FF the browser is back to its former speed.
I forgot to list Graphics section from about:support. Here it is:

Adapter Description	NVIDIA NVS 4200M
Adapter Description (GPU #2)	Intel(R) HD Graphics Family
Adapter Drivers	nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter Drivers (GPU #2)	igdumd64 igd10umd64 igd10umd64 igdumdx32 igd10umd32 igd10umd32
Adapter RAM	1024
Adapter RAM (GPU #2)	Unknown
Device ID	0x1057
Device ID (GPU #2)	0x0116
Direct2D Enabled	true
DirectWrite Enabled	true (6.2.9200.16492)
Driver Date	5-25-2011
Driver Date (GPU #2)	3-6-2011
Driver Version	8.17.12.6871
Driver Version (GPU #2)	8.15.10.2321
GPU #2 Active	false
GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
Subsys ID	21d117aa
Subsys ID (GPU #2)	21d117aa
Vendor ID	0x10de
Vendor ID (GPU #2)	0x8086
WebGL Renderer	Google Inc. -- ANGLE (NVIDIA NVS 4200M Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d
AzureContentBackend	direct2d
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0
(In reply to André Weidemann from comment #150)
> I forgot to list Graphics section from about:support. Here it is:
> 
> Adapter Description	NVIDIA NVS 4200M
> Adapter Description (GPU #2)	Intel(R) HD Graphics Family
> Adapter Drivers	nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
> Adapter Drivers (GPU #2)	igdumd64 igd10umd64 igd10umd64 igdumdx32 igd10umd32
> igd10umd32
> Adapter RAM	1024
> Adapter RAM (GPU #2)	Unknown
> Device ID	0x1057
> Device ID (GPU #2)	0x0116
> Direct2D Enabled	true
> DirectWrite Enabled	true (6.2.9200.16492)
> Driver Date	5-25-2011
> Driver Date (GPU #2)	3-6-2011
> Driver Version	8.17.12.6871
> Driver Version (GPU #2)	8.15.10.2321
> GPU #2 Active	false
> GPU Accelerated Windows	1/1 Direct3D 11 (OMTC)
> Subsys ID	21d117aa
> Subsys ID (GPU #2)	21d117aa
> Vendor ID	0x10de
> Vendor ID (GPU #2)	0x8086
> WebGL Renderer	Google Inc. -- ANGLE (NVIDIA NVS 4200M Direct3D9Ex vs_3_0
> ps_3_0)
> windowLayerManagerRemote	true
> AzureCanvasBackend	direct2d
> AzureContentBackend	direct2d
> AzureFallbackCanvasBackend	cairo
> AzureSkiaAccelerated	0


Can you try with a current driver: http://www.nvidia.com/download/driverResults.aspx/80141/en-us
This slowness still persists even in firefox36.
Running this demo http://ie.microsoft.com/testdrive/performance/fishbowl/ with omtc enabled results in 1-2fps (beside the general browser slowness, slow scrolling, slow tab switching, etc. which happens even with blank tabs open). After disabling omtc, the demo runs fine with 60fps, browser behaves normally(no lags).
The weird thing i noticed, is that this only happens when starting firefox with the nvidia card.
So, to summarize:
nvidia+omtc=poor performance
nvidia+no omtc=good performance
intel+omtc=good performance

Graphics
--------

Adapter Description: Intel(R) HD Graphics 4600
Adapter Description (GPU #2): NVIDIA GeForce GT 740M
Adapter Drivers: igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32
Adapter Drivers (GPU #2): nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter RAM: Unknown
Adapter RAM (GPU #2): 2048
Device ID: 0x0416
Device ID (GPU #2): 0x1292
Direct2D Enabled: true
DirectWrite Enabled: true (6.2.9200.16571)
Driver Date: 1-5-2015
Driver Date (GPU #2): 2-5-2015
Driver Version: 10.18.14.4080
Driver Version (GPU #2): 9.18.13.4752
GPU #2 Active: false
GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC)
Subsys ID: 502317aa
Subsys ID (GPU #2): 502317aa
Vendor ID: 0x8086
Vendor ID (GPU #2): 0x10de
WebGL Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics 4600 Direct3D9Ex vs_3_0 ps_3_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d
AzureContentBackend: direct2d
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0


One weird thing is that firefox says "GPU #2 Active: false" when it should be true. "Nvidia gpu activity" reports firefox as running, and it is clearly(for me) running on the nvidia card.
This bugs contains a bunch of noise and reporters that experienced different problems. Can anyone still experiencing this please try out FF 37 (beta) and report the results and the graphics section of your about:support in bug 1137705
Status: REOPENED → RESOLVED
Closed: 10 years ago9 years ago
Resolution: --- → INCOMPLETE
I'm not sure if it's the same problem because my Firefox is functioned so-so (the stuttering in the UI and slowness in scrolling are not outright obvious) but no videos were smooth. I couldn't watch 1080p videos at all. After disabling OMTC I can watch 1080p at 60 FPS without any major stuttering! I've just disabled OMTC, didn't touch the hardware acc setting.*

*Disabling hardware acceleration while leaving OMTC did improve performance for me, though not to acceptable levels (still video performance used as indicator)

I don't recall whether this was happening prior to the Omega driver release or not, sorry.

AMD FX-6300 6-core 3.5 GHz
Radeon 6850, latest catalyst
Win 7 x64

Graphics
--------

Adapter Description: AMD Radeon HD 6800 Series
Adapter Drivers: aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64
Adapter RAM: 1024
ClearType Parameters: Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 300
Device ID: 0x6739
Direct2D Enabled: true
DirectWrite Enabled: true (6.2.9200.16571)
Driver Date: 11-20-2014
Driver Version: 14.501.1003.0
GPU #2 Active: false
GPU Accelerated Windows: 2/2 Direct3D 11 (OMTC)
Subsys ID: 21f81458
Vendor ID: 0x1002
WebGL Renderer: Google Inc. -- ANGLE (AMD Radeon HD 6800 Series Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d 1.1
AzureContentBackend: direct2d 1.1
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0
This comment relates to this bug as well as a font rendering/anti-aliasing problem (bug 1096934).  Reminder, my video card is an ATI (now AMD) Radeon HD5670.

Finally got tired of the font rendering/anti-aliasing problem, and decided to just try the AMD website driver upgrade for the HD5670, or just buy a new video card if that failed.  Didn't really care anymore whether Dell had some "special sauce" in their "official" driver for my XPS 7100, which they last updated in 2009.  

I'm happy to report that downloading and installing the latest (ver 14.501.1003) AMD driver (and Catalyst utility ver 14.12) allowed me to re-enable hardware acceleration in Firefox (as well as restore the various about:config workarounds back to their defaults).  This fixed the UI slowness that appeared with FF v33.0, as well as the font rendering/anti-aliasing problem.  Everything appears to be working great now (current FF v38.0.6).  Problem solved for me!
Flags: needinfo?(jmuizelaar)
Flags: needinfo?(bas)
You need to log in before you can comment on or make changes to this bug.