Thank you amd, now I can't use Linux

[duby229]

Look, as I said earlier if it runs fine in Windows and has a totally different set of unrelated problem with the Catalyst driver, then it cannot be anything at all to do with defective hardware other than the GPU itself, and even then only if one driver uses certain parts of the hardware (which they almost definitely don't).

The symptoms damn sure look like a hardware fault to me. It's really very simple troubleshooting to determine what the faulty hardware could be.... Stress test the existing hardware and determine triggers, then replace a piece of hardware and perform the same stress test to see if the trigger still exists. Rinse and repeat until you have narrowed down to the faulty hardware. A guy who's been doing it long enough will have experience and intuition and be able to narrow it down very quickly.

[duby229]

The symptoms damn sure look like a hardware fault to me. It's really very simple troubleshooting to determine what the faulty hardware could be.... Stress test the existing hardware and determine triggers, then replace a piece of hardware and perform the same stress test to see if the trigger still exists. Rinse and repeat until you have narrowed down to the faulty hardware. A guy who's been doing it long enough will have experience and intuition and be able to narrow it down very quickly.

Just thinking about it, I once had an old style PS2 keyboard that would prevent the computer from going to sleep. I never would have thought the keyboard would be the problem until I tried a different keyboard. Another time I come across a HDD that would prevent any MB it was plugged into from posting. Another time I had a PSU that could boot the computer and idle it, but if I tried to do anything that increased CPU load it would abruptly turn off....

Weird things happen and you gotta troubleshoot. Part of that is stress testing.

[gradinaruvasile]

Guys, we are using experimental drivers here. I too have a A8-5500 and i have random freezes, mainly when im not at the computer. Other than that i can play games for hours or compile stuff and it wont happen.

Also, if i do the same with catalyst, i have no issues. Hardware fault? Look around, there are a few reports of freezes with APUs using the dpm driver.

The driver itself works wonderfully - it has better performance than catalyst in source games for example (what i play) and has about the same thermal characteristics as catalyst. Remember APUs work differently than dedicated cards.

[pandev92]

Guys, we are using experimental drivers here. I too have a A8-5500 and i have random freezes, mainly when im not at the computer. Other than that i can play games for hours or compile stuff and it wont happen.

Also, if i do the same with catalyst, i have no issues. Hardware fault? Look around, there are a few reports of freezes with APUs using the dpm driver.

The driver itself works wonderfully - it has better performance than catalyst in source games for example (what i play) and has about the same thermal characteristics as catalyst. Remember APUs work differently than dedicated cards.

Exactly, I see that if I use kernel 3.10.7 without dpm, all run correctly and no reboots, but if I use kernel 3.11 rc6 for example, every 4or 5 hours or when I do an intensive work with the computer, it suddenly restart.

[brosis]

Look, as I said earlier if it runs fine in Windows and has a totally different set of unrelated problem with the Catalyst driver, then it cannot be anything at all to do with defective hardware other than the GPU itself, and even then only if one driver uses certain parts of the hardware (which they almost definitely don't).

It can. About 10 years ago, I had had CTD in CnC generals when doing 8er battles on huge maps.
Apparently, the game simply CRCed whole data, so any faults would surface. Its pretty similar to Prime95/Mprime. Turned out the CPU was undervolted too much, it did not surface in any desktop use.
So, 24hr mprime followed by 24hr memtest are obligatory.

Guys, we are using experimental drivers here. I too have a A8-5500 and i have random freezes, mainly when im not at the computer. Other than that i can play games for hours or compile stuff and it wont happen.

Also, if i do the same with catalyst, i have no issues. Hardware fault? Look around, there are a few reports of freezes with APUs using the dpm driver.

The driver itself works wonderfully - it has better performance than catalyst in source games for example (what i play) and has about the same thermal characteristics as catalyst. Remember APUs work differently than dedicated cards.

Maybe that DPM in open driver crosses TDP headroom. Longer measures, a bit too high timings, I can only guess..

[gradinaruvasile]

Maybe that DPM in open driver crosses TDP headroom. Longer measures, a bit too high timings, I can only guess..

I dont know if temps are the problems. I monitor its temperatures and it never crosses 60 C with the open driver (it stays ~52-55C) when under full load for hours (compiling with 4 threads) even with its tiny default cooler. The temperature behavior is just about the same with fglrx, at least from what sensors tells me.
But i dont exclude the possibility of having certain areas of the APU (GPU part) overheated or something.

There is something weird sometimes - it throttles down to ~1.9-2 GHz (max is 3.2GHz) effectively (cpufreq-aperf), despite /proc/cpuinfo showing 3.2 GHZ (the highest p-state) and temperatures being reported as ~48 C. This didnt happen with fglrx.

Now, the problem is that the freezes occur usually when doing very light stuff like browsing or just typing in terminals with nothing big going on in the background or, mostly, being away from the computer, temperatures reported (when the monitor was on) ~15C (not real, when the CPU is idle, the temps are not accurate, they get real after its loaded). But it did happen sometimes when i was compiling stuff, but usualy the temperatures reportedly were ~45-48C.

It NEVER happened so far when i was playing games (tf2, nexuiz, DoD). Interesting.

Also, sometimes right after boot the CPU is powered with the highest available voltage even when idle (1.34V) until i either blank/unblank (dpms) the screen or switch to a vt and back after which the voltage starts to change dynamically (0.9 to 1.34 V) and temps go down ~10C instantly.

I also recompiled the kernel many times with diffrent latency/scheduler options, tried adding R600_VA=0, R600_DEBUG=nodma, to /etc/environment, radeon.msi=0, disabling iommu=pt, iommu=off to the kernel command line etc. Nothing helped so far. Too bad cause the driver from a performance standpoint is excellent for the workloads i use (this bugs me the most).

It runs better than fglrx (maybe a bit lower throughput, but for my games (tf2 in particular) its way better than fglrx also it has vdpau that is available widely (although it does crash mplayer from time to time and cause massive memory leaks in xbmc after ~30 minutes).

[duby229]

I dont know if temps are the problems. I monitor its temperatures and it never crosses 60 C with the open driver (it stays ~52-55C) when under full load for hours (compiling with 4 threads) even with its tiny default cooler. The temperature behavior is just about the same with fglrx, at least from what sensors tells me.
But i dont exclude the possibility of having certain areas of the APU (GPU part) overheated or something.

There is something weird sometimes - it throttles down to ~1.9-2 GHz (max is 3.2GHz) effectively (cpufreq-aperf), despite /proc/cpuinfo showing 3.2 GHZ (the highest p-state) and temperatures being reported as ~48 C. This didnt happen with fglrx.

Now, the problem is that the freezes occur usually when doing very light stuff like browsing or just typing in terminals with nothing big going on in the background or, mostly, being away from the computer, temperatures reported (when the monitor was on) ~15C (not real, when the CPU is idle, the temps are not accurate, they get real after its loaded). But it did happen sometimes when i was compiling stuff, but usualy the temperatures reportedly were ~45-48C.

It NEVER happened so far when i was playing games (tf2, nexuiz, DoD). Interesting.

Also, sometimes right after boot the CPU is powered with the highest available voltage even when idle (1.34V) until i either blank/unblank (dpms) the screen or switch to a vt and back after which the voltage starts to change dynamically (0.9 to 1.34 V) and temps go down ~10C instantly.

I also recompiled the kernel many times with diffrent latency/scheduler options, tried adding R600_VA=0, R600_DEBUG=nodma, to /etc/environment, radeon.msi=0, disabling iommu=pt, iommu=off to the kernel command line etc. Nothing helped so far. Too bad cause the driver from a performance standpoint is excellent for the workloads i use (this bugs me the most).

It runs better than fglrx (maybe a bit lower throughput, but for my games (tf2 in particular) its way better than fglrx also it has vdpau that is available widely (although it does crash mplayer from time to time and cause massive memory leaks in xbmc after ~30 minutes).

You're a very well educated person. You clearly have the knowledge to troubleshoot this problem so now you need to find out exactly what the trigger is that causes it. It be that you can't fix it yourself, but at the very least you'll be able to post better bug reports.

It's just that with the information you've given so far it's still unclear on how exactly to duplicate this issue. Once the trigger is found it is a lot easier to understand...

My guess is that you are a programmer, If I'm right, then you already understand how code works and can debug. Apply that knowledge to this issue.

[agd5f]

Are you using a distro that uses gcc 4.8? I've noticed a lot of problems with 4.8 related to variable sized array handling which we use a lot in the radeon driver. You may be hitting some issues related to that. Can you try a distro with an older version of gcc?

[duby229]

Exactly, I see that if I use kernel 3.10.7 without dpm, all run correctly and no reboots, but if I use kernel 3.11 rc6 for example, every 4or 5 hours or when I do an intensive work with the computer, it suddenly restart.

In your particular example it seems to me like dpm is putting the hardware in a configuration that isnt working right for your hardware. It seems that it is a hardware fault somewhere.

[pandev92]

Are you using a distro that uses gcc 4.8? I've noticed a lot of problems with 4.8 related to variable sized array handling which we use a lot in the radeon driver. You may be hitting some issues related to that. Can you try a distro with an older version of gcc?

I will try with an older ubuntu version.