- Package:
- xserver-xorg-video-radeon
- Source:
- xserver-xorg-video-ati
- Description:
- X.Org X server -- AMD/ATI Radeon display driver
- Submitter:
- "Alexey Feldgendler"
- Date:
- 2010-04-13 15:33:06 UTC
- Severity:
- normal
With KMS and DRI2, glxgears is dramatically slower (about 140 fps) than without KMS and DRI2 and with only DRI (about 300 fps). Also, the motion is not uniformly smooth and seems to depend a lot on the background system activity. I have a shell script that updates the clock in my dzen2 status bar once per second, and the gears seem to stumble for a split second once per second as it happens. I tried stopping that script, and have confirmed that it indeed is causing the jitter, even though the load it puts on the system is tiny. This didn't happen without KMS and DRI2; not only did the animation run faster, but it was also smooth (without noticeable jitter). The kernel is linux-image-2.6.33-2-686 from experimental, and libgl1-mesa-dri is 7.7-4 from unstable.
Alexey Feldgendler wrote: glxgears is dramatically not a benchmark. Its fps output measures nothing interesting in the real life. You should use a real 3D game or whatever, but not glxgears. Brice
Alright, I admit this bug is invalid. I've run gtkperf, the best I could find in Debian, and observed no statistically significant slowdown. Tried to run these demos/benchmarks <http://unigine.com/download/> but couldn't because they require the GL_ARB_map_buffer_range extension — don't know what it is, but it sure isn't in my glxinfo output.
Alexey Feldgendler wrote:
reopen 572911 thanks This issue happens to me, too. It is most definitely not solved. Any OpenGL application I try has very poor performance when KMS is enabled. Have tried glxgears (FPS drops from ~1500 to ~300), AisleRiot (from gnome-games), and ZSNES. Using a Radeon X1550 (RV516). System information attached.
Alex Hvostov wrote: There is nothing to solve in this report, glxgears is not a benchmark, so having a "slow glxgears" means nothing. Moreover, you're not even using KMS/DRI2, so your "problem" has nothing to do here. cheers, Brice
Please reread my message. I am experiencing poor performance in AisleRiot and ZSNES as well. This is when KMS is *on*; I experience no such performance issue when KMS is off. Honestly. This is insulting.
Alex Hvostov wrote: This bug report was about glxgears, not about these. Why did you send the log with KMS off then... Brice
It seems to me that it's about poor OpenGL performance, with glxgears being used as an example. That's certainly not the only application suffering from poor performance. My mistake. Sorry about that. Attached is the log with KMS on; I hope it helps.
glxgears is showing the same problem (very poor performance) as every other application I've tested. How is that meaningless? AisleRiot gets ~15-30 FPS, which is terrible considering it's just a solitaire game. armagetronad locks up the entire X display for seconds at a time, leaving only a split second between lock-ups where the mouse cursor can be moved. AisleRiot uses libclutter (and, therefore, OpenGL) for drawing. ZSNES uses OpenGL as well. Both show very poor performance.
This issue, I believe, is well known: http://www.phoronix.com/scan.php?page=article&item=ati_ubuntu_kmsums&num=1 In my opinion this bug should be renamed to something like: degraded performance of KMS vs UMS And perhaps it should be put in wishlist and sent upstream. Andres