AMD Radeon Vega 8 "too old" for WebRender
Categories
(Core :: Graphics: WebRender, task, P3)
Tracking
()
People
(Reporter: kitynska, Unassigned)
References
(Blocks 1 open bug)
Details
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:68.0) Gecko/20100101 Firefox/68.0
Steps to reproduce:
Run Firefox 68 under Windows 10 with AMD Radeon Vega 8.
Actual results:
WebRender disabled:
WEBRENDER_QUALIFIED
blocked-device-too-old by env: Device too old
Expected results:
WebRender should be enabled by default.
Comment 1•5 years ago
|
||
What's the device id for this GPU? You can find it in about:support
Updated•5 years ago
|
(In reply to Jeff Muizelaar [:jrmuizel] from comment #1)
What's the device id for this GPU? You can find it in about:support
0x15dd
Updated•5 years ago
|
Comment 3•5 years ago
|
||
https://pci-ids.ucw.cz/read/PC/1002/15dd
It's a mobile GPU and also sold as APU.
The latter might be the cause why no battery was detected:
https://searchfox.org/mozilla-release/rev/7eb62057a7812cc152d4d22aa87566a4d51808f9/gfx/thebes/gfxPlatform.cpp#2581-2611
Comment 4•5 years ago
|
||
Same problem: AMD Ryzen 2400G, Radeon Vega 10, Ubuntu 19.10
GPU #1,
Active Yes,
Description AMD RAVEN (DRM 3.33.0, 5.3.0-19-generic, LLVM 9.0.0),
Vendor ID 0x1002,
Device ID 0x15dd,
Driver Vendor mesa/radeonsi,
Driver Version 19.2.1.0,
RAM 256MB
Comment 5•5 years ago
|
||
(In reply to Александр Плосков from comment #4)
Same problem: AMD Ryzen 2400G, Radeon Vega 10, Ubuntu 19.10
GPU #1,
Active Yes,
Description AMD RAVEN (DRM 3.33.0, 5.3.0-19-generic, LLVM 9.0.0),
Vendor ID 0x1002,
Device ID 0x15dd,
Driver Vendor mesa/radeonsi,
Driver Version 19.2.1.0,
RAM 256MB
My mistake: Vega 11 of course. It's a desktop.
The issue also affect (In reply to Jan Andre Ikenmeyer [:darkspirit] from comment #3)
https://pci-ids.ucw.cz/read/PC/1002/15dd
It's a mobile GPU and also sold as APU.
The latter might be the cause why no battery was detected:
https://searchfox.org/mozilla-release/rev/7eb62057a7812cc152d4d22aa87566a4d51808f9/gfx/thebes/gfxPlatform.cpp#2581-2611
15dd is also known as AMD Ryzen 5 2500u for mobile as equipped on some laptops (HP Envy x360 convertible in my case) first launched before the desktop version according to Wiki Chips (https://en.wikichip.org/wiki/amd/cores/raven_ridge).
Description AMD RAVEN (DRM 3.33.0, 5.3.8-300.fc31.x86_64, LLVM 10.0.0)
Vendor ID 0x1002
Device ID 0x15dd
Driver Vendor mesa/radeonsi
Driver Version 20.0.0.0
RAM 1024MB
The issue also impact the mobile version. Should the fix apply, the name is "AMD Raven"
Follow-up. Any update to enable support for AMD Raven Ridge with the provided data above?
Comment 9•5 years ago
|
||
We haven't released WebRender on mobile hardware yet. It's coming but in the mean time you can enable it with gfx.webrender.all=true
Comment 10•5 years ago
|
||
(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
We haven't released WebRender on mobile hardware yet. It's coming but in the mean time you can enable it with gfx.webrender.all=true
We can help with this (QA/dev) process? Maybe you can attach links to similar bug/hg branch fixed for another platform/hardware?
Comment 11•5 years ago
|
||
We haven't released WebRender on mobile hardware yet.
I used to enable WebRender on Windows 10 x64, but stopped after noticing poor battery life and high GPU usage by Firefox.
Comment 12•5 years ago
|
||
(In reply to Alexander Ploskov from comment #10)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
We haven't released WebRender on mobile hardware yet. It's coming but in the mean time you can enable it with gfx.webrender.all=true
We can help with this (QA/dev) process? Maybe you can attach links to similar bug/hg branch fixed for another platform/hardware?
The best thing to do is to test with gfx.webrender.all=true turned on and see if you notice any substantial regressions in CPU or GPU usage.
The main thing that we know about that's still worse is full-screen video playback. That's tracked in bug 1579235.
Other big performance issues are tracked from this meta bug 1607452.
Comment 13•5 years ago
|
||
(In reply to jimbo1qaz from comment #11)
We haven't released WebRender on mobile hardware yet.
I used to enable WebRender on Windows 10 x64, but stopped after noticing poor battery life and high GPU usage by Firefox.
When did you last try?
Comment 14•5 years ago
|
||
According to about:support, I'm using AMD Radeon(TM) RX Vega 10 Graphics.
On 12/06/2019 (probably Firefox 69 or 70), I sent the following chat messages to another person:
i disabled it after my laptop battery (i reduced the maximum charger level in bios) died in 70 minutes
and i know for sure windows taskmgr said webrender ate 10% gpu on hacker news (?)
and like 1-3% after i disabled webrender
and windows battery says Firefox ate energy
I retested on Windows 10 x64, on either Firefox 72 or 72.0.1 (64-bit), and was unable to reproduce these issues on a clean Firefox profile with WebRender enabled, or on my primary profile. I tested the following sites (with power adapter plugged in):
- http://patorjk.com/misc/scrollingtext/timewaster.php when not scrolling
- https://invidio.us/watch?v=528Jc3q86F8 when not playing the video
- https://hackaday.com/2020/01/09/beos-the-alternate-universes-mac-os-x/
- https://news.ycombinator.com/item?id=22002062
The idle GPU usage is 2-3%. I got 3-4% on Hacker News, and 2-3% on other pages. Every time I load a page, GPU usage spikes to 10-20% for around 2 seconds.
From preliminary testing on this bug report page (not the above URLs), I get around the same results with power adapter unplugged (which is when I originally experienced power drain).
Maybe WebRender currently doesn't have excessive GPU usage (at least in the steady-state). I will switch Firefox back to WebRender, use it continually, and report back if I have any other findings.
Comment 15•5 years ago
|
||
(In reply to jimbo1qaz from comment #14)
According to about:support, I'm using AMD Radeon(TM) RX Vega 10 Graphics.
On 12/06/2019 (probably Firefox 69 or 70), I sent the following chat messages to another person:
i disabled it after my laptop battery (i reduced the maximum charger level in bios) died in 70 minutes
and i know for sure windows taskmgr said webrender ate 10% gpu on hacker news (?)
and like 1-3% after i disabled webrender
and windows battery says Firefox ate energyI retested on Windows 10 x64, on either Firefox 72 or 72.0.1 (64-bit), and was unable to reproduce these issues on a clean Firefox profile with WebRender enabled, or on my primary profile. I tested the following sites (with power adapter plugged in):
- http://patorjk.com/misc/scrollingtext/timewaster.php when not scrolling
- https://invidio.us/watch?v=528Jc3q86F8 when not playing the video
- https://hackaday.com/2020/01/09/beos-the-alternate-universes-mac-os-x/
- https://news.ycombinator.com/item?id=22002062
The idle GPU usage is 2-3%. I got 3-4% on Hacker News, and 2-3% on other pages. Every time I load a page, GPU usage spikes to 10-20% for around 2 seconds.
From preliminary testing on this bug report page (not the above URLs), I get around the same results with power adapter unplugged (which is when I originally experienced power drain).
Maybe WebRender currently doesn't have excessive GPU usage (at least in the steady-state). I will switch Firefox back to WebRender, use it continually, and report back if I have any other findings.
Thanks for testing. Please do report back. We're going to be landing some changes in the next couple of days on Nightly that should improve things further.
Comment 16•5 years ago
|
||
(In reply to kitynska from comment #2)
0x15dd
Description
•