ASUS Matrix HD7970 problem - Help!

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213
Topic 197483

I have problem with ASUS Matrix HD 7970. My WU time was fastest I ever seen. At first sight I thought it is great to run WU-s from 3,600 to 4,000 but almost all WU-s are Validate error. This GPU shows no errors at any other BOINC projects. Does anybody know what is going on? Help please!!!

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 76

ASUS Matrix HD7970 problem - Help!

Quote:
but almost all WU-s are Validate error.


We only see one result being a validate error, while several others show as aborted by user.

In any case, to see what you can do against a validate error, start by reading Gary's post here.

As for other projects, the only project you use the GPU actively at is DistrRTgen, all other projects that could use your GPU do not. You don't say if you run only one task at a time on the GPU, or several.

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

Oh,I am sorry! It is my

Oh,I am sorry! It is my fault. I am running just one WU every time at every project. I was disconnected from Einstein@home project since I discovered this problem. Reinstalling OS did not help and I tried Windows 7 and Windows 8.1.
I am going to switch myself to einsten@home now and you can see what is the problem.

Logforme
Logforme
Joined: 13 Aug 10
Posts: 332
Credit: 1714373961
RAC: 0

Given that the card is

Given that the card is advertised as highly overclockable I assume you have it overclocked?
If the problems persists I suggest running the card at stock speed and see if that helps.

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

No, I did not overclock. It

No, I did not overclock. It runs at stock speed @1100MHz.
After last WU with error I have made small change in UEFI/BIOS (Sabertooth Z77). PCI-E 2.0/3.0 settings were at Auto and I set it manual as PCI-E 2.0. At this time WU is running 1 hour and 5 minutes with 67.8% complete. This could be promise....

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

Not good again! Validation

Not good again! Validation inconclusive.

Logforme
Logforme
Joined: 13 Aug 10
Posts: 332
Credit: 1714373961
RAC: 0

RE: Not good again!

Quote:
Not good again! Validation inconclusive.


This is not a problem (yet). It just means you have to wait for another computer (wingman) to complete the same task so they can be compared. After that you know if your card produced the correct result.

Quote:
No, I did not overclock. It runs at stock speed @1100MHz.


I'm sorry, my post was not clear about what I meant. When I wrote stock speed I meant the specification speed for the 7970.

The original specification is 925 MHz.
Later the GHz edition version was released. This was factory overclocked by AMD to 1000 MHz.
Your card has been further overclocked by ASUS to 1100 MHz.

My idea is that you can try to restore (underclock) your card to 925 MHz to see if that solves the issue.

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

No change at all. There is

No change at all. There is difference between PCIE 1.1 and 2.0. My last WU (Arecibo) was made underclocked @900 MHz with memory clock @5800 and it has been finished faster than WU@1100 MHz!!! Now I have removed AMD driver with AMD cleanup tool and I installed ASUS R9 280X driver. Lets see is there some difference....

mikey
mikey
Joined: 22 Jan 05
Posts: 12684
Credit: 1839092349
RAC: 3820

RE: No change at all. There

Quote:
No change at all. There is difference between PCIE 1.1 and 2.0. My last WU (Arecibo) was made underclocked @900 MHz with memory clock @5800 and it has been finished faster than WU@1100 MHz!!! Now I have removed AMD driver with AMD cleanup tool and I installed ASUS R9 280X driver. Lets see is there some difference....

According to this page http://einsteinathome.org/host/10254947/tasks&offset=0&show_names=1&state=3&appid=0 you are finishing units just fine and getting good credits for them. Your invalid units, you only have 3, one of which another user had problems with too. You do have some 'error' units but it says you aborted them on your own. This might or might not be totally true, sometimes Boinc uses terms that are generic, like user aborted, when in fact the Server itself aborted the units. The key is the units were aborted by someone or something, and can be ignored in your stats.

It looks like underclocking your card may have made your crunching faster and it was working. It can take a few days for more definitive results to come in, so this time I would suggest you wait it out to see if it does or doesn't work. Not EVERY project, or subproject, benefits from overclocking.

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

I was very surprised when I

I was very surprised when I saw valid Arecibo WU ( I think this is my first valid in 5 months, since I bought this GPU). It looks like underclocking is working but I have to be certain what underclocking gives good results:GPU or memory or both. I was running GPU@900 MHz and memory@5800 MHz. That is the lowest possible settings fom ASUS GPU Tweak software. Five months ago I was running E@H several days and results were desperate. Onla couple of WU-s were validated.
I will put focus on underclocking, it seems there is a solution of this problem. I am very surprised because native Tahiti XT2 GPU clock is 1000 MHz. I will report results of testing with underclocking here.

Thank you friends!

Sasa Jovicic
Sasa Jovicic
Joined: 17 Feb 09
Posts: 75
Credit: 88879732
RAC: 119213

Underclocking does not help.

Underclocking does not help. I have 8 valid and 15 invalid WU. Something is still wrong. I really want to crunch E@H but this is frustration.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.