ASUS R9 280x Crashing on Perseus Arm Survey
1 Sep 2014 12:26:19 UTC
Topic 197706
(moderation:
Since the Gravitational Wave S6 Directed Search (CasA) ended I am having severe issues with Binary Radio Pulsar Search (Perseus Arm Survey) work units crashing my GPU. The GPU crashes within 5 minutes of each work unit. you can suspend the work unit then resume and it will inch forward for 5 minutes or so then crash again. This is aggravating....
To the point of...
Reinstalling Drivers, New and Old including Beta
Replacing the PSU with a NEW PSU
Replacing the HD with a NEW HD
Installing win7
Reinstalling win8.1
Replacing the GPU with a new GPU
Changing the GPU load to 1 unit..
then back to 2...
So, with each change nothing different.
Here is the machine: NOT OVERCLOCKED
AMD FX(tm)-9370 Eight-Core Processor [Family 21 Model 2 Stepping 0]
Number of processors 8
Coprocessors CAL AMD Radeon HD 7870/7950/7970/R9 280X series (Tahiti) (3072MB) driver: 1.4.1848
Operating System Microsoft Windows 8.1
So AFTER Scratching my head until there was a bare spot, I decided Maybe it wasn't my GPU and that it was the work units causing this issue.
So I installed SETI@Home and the GPU CRUSHES those work units in less than 8 minutes each and has done so without error or crashing for at least 9 hours now.
I have installed other intensive GPU software and it runs fine. So, I have concluded that its not my machine, but the work units causing the issue... I had no issues crushing 2 Gravitational Wave S6 simultaneously with this gpu in 15 minutes or so.
Can anyone help... Im not a SETI@home fan..
ASUS R9 280x Crashing on Perseus Arm Survey
)
I have no problem with my Radeon HD 8670D, AMD A10-6700, Windows 8.1.
Tullio
Did you install Asus GPU
)
Did you install Asus GPU tweak?
If not, do so.
Then underclock the GPU memory speed slightly (or significantly) and run some tests with the Perseus Arm units and then adjust speeds as necessary.
I have seen this sort of problem more than once with some my hosts, nothing wrong with the GPU etc. but in certain motherboard/GPU/memory configurations I can't run my 7970/280X cards at full memory frequency with the BRP5 Perseus Arm App.
Reducing the GPU memory frequency by 25MHz (100MHz effective) has immediately solved the problem for me and running at the reduced speed doesn't appear to have a detrimental effect on runtime when comparing to my other similar machines which run at the intended factory clocks...
Gav.
When I look at your computer
)
When I look at your computer list I see 3 computers with a "7870/7950/7970/R9 280X" GPU. A quick glance at the Perseus Arm error list for the computers say all the tasks are "Aborted by user"
I assume it's not you that are aborting the tasks? In that case maybe something on your computers is interfering with the E@H app?
Logforme, RE: When I
)
Logforme,
Yes, I aborted all the because they were crashing my GPU consistently.
The machine is question is dedicated ONLY to Einstein. There is no other software on it to interfere. Only insttaled software other than GPU tweak is except BOINC.
Gavin,
I do have the GPU tweak installed to monitor temps. I never thought to reduce the GPU memory speed.. Ill try that and see if it helps. While writing this and after a reduction of 100mhz, It crashed one instance of the GPU already, Ill keep fiddling with the GPU MEM speed until it stabilizes...
since I wrote the original post, It has crushed SEVERAL Seti@Home units with no issues... So its got to be something simple with a setting of some sort... probably memory speed is the key... Ill update when i figure it out.
Thanks Gavin for leading me to the memory speeds.
Hello ! I had similar
)
Hello !
I had similar problem with mx Matrix HD7970. GPU memory is causing this problem. Set GPU memory@6000Mhz and set GPU memory .
If you are using ASUS GPU Tweak you will have to set GPU memory clock every time when you turn on or restart your Computer. I do not Know why but, after restart, GPU Tweak is set @6000MHZ but GPU memory is still working @ default, so you will have to move GPU memory clock slightly and get it back @ 6000MHz than hit Apply. After these settings you can run E@H.
http://einsteinathome.org/node/197483
So after a few days of
)
So after a few days of running different tasks from other applications, I have found that its Einstein's problem that keeps crashing my GPU. I can run SETI without issue, I have moved over to Milkyway and crush there WU's in about 10 seconds. I since installed another r9 280x and can run 6 WU in about 64 seconds. been doing this for 4 days strait now... NO CRASHING.
So, when you guys figure out whats going on with Einstein's issues with R9 280X GPU, let me know. Until then I'll be over at milkyway letting my monster machine munch on there WU for a while.