If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
- Patience is the level of endurance one can have before negativity.
- Patience is stronger than force. - Patience is the weapon that forces deception to reveal itself.
- Patience is when you're waiting for something better to happen.
- Patience is the mother of all virtues and the godmother of madness.
- Patience stagnates if isn't founded on hope.
- Patience is moral elasticity.
The interesting difference for amdgpu on my RX 480 is the cpu usage
4.8:
4.9:
I'm still not sure whether it is a regression.
Overall CPU usage is a noticeable bit lower while everything else stays the same, but I'm worried that the CPU usage on that thread is a possible bottleneck on slower CPU or with rendering while CPU load is high.
I'll test it tonight. I noticed a significant decrease in NVMe performance between 4.8.4 and later 4.8 kernels. Let's see if this regression has been fixed in 4.9.
The interesting difference for amdgpu on my RX 480 is the cpu usage
4.8:
4.9:
I'm still not sure whether it is a regression.
Overall CPU usage is a noticeable bit lower while everything else stays the same, but I'm worried that the CPU usage on that thread is a possible bottleneck on slower CPU or with rendering while CPU load is high.
Looks like its using a single core first before spawning another worker thread. Saves a lot of context switching and locking to use 2 cores fully instead of 4 partially. Just a guess. Should test if the latencies are better.
My hacky and incomplete solution for "late frames" shows less of those, while the FPS remain the same, so this indicates an improvement there.
In unigine heaven the buffer wait times remain the same, but I tried native csgo and previously when there was some action on the screen the FPS would be somewhere between 50 and 100. Now they very rarely dip below 100 FPS and the buffer wait time has been massively reduced. On the other hand there was also a small csgo update, so perhaps it got improved from that side too.
My hacky and incomplete solution for "late frames" shows less of those, while the FPS remain the same, so this indicates an improvement there.
In unigine heaven the buffer wait times remain the same, but I tried native csgo and previously when there was some action on the screen the FPS would be somewhere between 50 and 100. Now they very rarely dip below 100 FPS and the buffer wait time has been massively reduced. On the other hand there was also a small csgo update, so perhaps it got improved from that side too.
Comment