Noob hardware advice, please

[Nathan_P]

I would expect about 7.5k PPD per 260 but i don't have one to confirm

P5769 - 7438PPD
P5766 - 7438PPD
P5772 - 7438PPD
P5771 - 7261PPD = 29578PPD

Kinda like that?

Nice

[Tobit]

Code: Select all

 Project ID: 5769
 Core: GROGPU2
 Credit: 353
 Frames: 100

 Name: ganymede - GTX 260
 Path: \\ganymede\fah\gpu0
 Number of Frames Observed: 300

 Min. Time / Frame : 00:00:34 - 8,970.4 PPD
 Avg. Time / Frame : 00:00:37 - 8,243.0 PPD

[kiore]

I would expect about 7.5k PPD per 260 but i don't have one to confirm

P5769 - 7438PPD
P5766 - 7438PPD
P5772 - 7438PPD
P5771 - 7261PPD = 29578PPD

Kinda like that?

That looks better, overclocked mine do 1-1.5 k ppd each more than that on those units so you could do even better than that if your cooling can handle it.
k

[Slowride]

That looks better, overclocked mine do 1-1.5 k ppd each more than that on those units so you could do even better than that if your cooling can handle it.
k

I'm not sure if my cooling can handle it or not. What are the max temps I should run at to prevent a meltdown?

Also, if you could recommend a good source of information for a novice overclocker, I would appreciate it!

[ikerekes]

Also, if you could recommend a good source of information for a novice overclocker, I would appreciate it!

http://en.fah-addict.net/articles/artic ... 00-gtx.php

[kiore]

Max temps and temps you are comfortable with are different, my GTX 260's were running at 85-95C when I had them in a case but rarely go over 70C now I run them in the open, the cards themselves can handle just over 100C but I was not comfortable at those temps and seemed to get more EUEs. It is a problem having the cards jammed close together, I have a desk fan blowing over them to help move air around.
Overclocking with things like rivatuner it is just a matter of moving a slide bar, so the technique is not so complicated, finding the right shader number will depend on your cards, some clock better than others and at the edge to performance a stable overclock on one unit will crash another type. rivatuner also allows you to increase the fan speed on the cards.
The numbers you are getting are not too bad, so really up to you if you want to squeeze them some more.
Although that article in Fahaddict says there are benefits in overclocking the RAM, I found it marginal at the least and way more unstable so I don't bother.
kiore.

[shdbcamping]

EVGA Precision is another program for OC'ing NV cards.

As to max temps for cards, it is as kiore states, up to what you feel comfortable with. I have one GTX260 that runs at an OC that rocks. Thermal throttling kicks in at just over 100C so I let it go there. I chose to OC to 95C on the 353 point wu's. That leaves the 783's bumping thermal throttle, but rocks on the rest.

The hotter you run your cards, the less the card's lifespan though. I wish the WU's ran a bit more consistently for heat, but they all do science a bit differently.

[Slowride]

The hotter you run your cards, the less the card's lifespan though

The two inner cards are running 80C ~ 85C so I think there is room for some tweaking but I have a problem.

Card #1 started giving me this:

[19:18:26] Project: 5769 (Run 7, Clone 73, Gen 403)
[19:18:26] - Read packet limit of 540015616... Set to 524286976.
[19:18:26] - Error: Could not get length of results file work/wuresults_00.dat
[19:18:26] - Error: Could not read unit 00 file. Removing from queue.
[19:18:26] EUE limit exceeded. Pausing 24 hours.
[01:10:45] + Working...

Based on what I read on here I ran memtestG80, if I am using it correctly, gave a boat load of errors so, it's time for an RMA, right? Seems fine otherwise, anything else I can check?

[bruce]

The hotter you run your cards, the less the card's lifespan though

Partly true. For high temperatures, you're correct. For moderate temperatures, the lifespan may decrease from 10^6 years down to 10^5 years but you'll never see it.

Card #1 started giving me this:

[19:18:26] Project: 5769 (Run 7, Clone 73, Gen 403)
[19:18:26] - Read packet limit of 540015616... Set to 524286976.
[19:18:26] - Error: Could not get length of results file work/wuresults_00.dat
[19:18:26] - Error: Could not read unit 00 file. Removing from queue.
[19:18:26] EUE limit exceeded. Pausing 24 hours.
[01:10:45] + Working...

Based on what I read on here I ran memtestG80, if I am using it correctly, gave a boat load of errors so, it's time for an RMA, right? Seems fine otherwise, anything else I can check?

Yes, if you get MemtestG80 errors it's time for an RMA.

The messages that you quoted are not necessarily a sign of memory errors. They might be, but other folks are getting similar errors who do not have memory errors. When you get the Pausing 24 message, though, something is wrong and the Pande Group recommends that you take that client off-line and fix it.

[Slowride]

For moderate temperatures, the lifespan may decrease from 10^6 years down to 10^5 years but you'll never see it.

What temperature range is considered moderate, 80C ~ 90C?

When you get the Pausing 24 message, though, something is wrong and the Pande Group recommends that you take that client off-line and fix it.

I have taken it off line. Just out of curiosity though, I ran memtestG80 on another of my cards which is folding away like crazy, gave me a bunch of errors as well. I don't understand Maybe I don't have the test configured correctly or is there another possible cause of the EUEs on the first card?

[kiore]

You will hear all sorts of numbers as to what is 'moderate', in my experience it depends on the individual card as well as card type. Also what you measuring it with.
Using speedfan to measure temps, my GTX 260's 'prefer' <90C and rarely have issues at <80C. My 9800 GT's are in trouble if they go over 80C and I see more frequent errors. My GTX 275 never seems to go over 75C no matter what.
My preference is to run everything as cool as possible as they are all overclocked and running 24/7, to achieve this I have lots of airflow and run the GTX 260's in the open, taking them out of a box and having the mobo horizontal dropped temps by nearly 20C.

[bruce]

Just out of curiosity though, I ran memtestG80 on another of my cards which is folding away like crazy, gave me a bunch of errors as well. I don't understand Maybe I don't have the test configured correctly or is there another possible cause of the EUEs on the first card?

There are other possible causes of EUEs. The only hardware that we can test easily are (A) Memory, and (B) the card as a whole.

What were the nature of your "bunch of errors"?

[Slowride]

What were the nature of your "bunch of errors"?

Sorry about the not so scientific description. Every phase of every iteration shows huge amounts of errors. However, after looking closer I realized I was still testing GPU 0, the card that won't fold. Apparently the test wasn't acknowledging my flags so, I guess I need help running the test on another of the GPUs. I read the "readme" file, still can't get it to run on a different GPU.

[shdbcamping]

You will hear all sorts of numbers as to what is 'moderate', in my experience it depends on the individual card as well as card type. Also what you measuring it with.
Using speedfan to measure temps, my GTX 260's 'prefer' <90C and rarely have issues at <80C. My 9800 GT's are in trouble if they go over 80C and I see more frequent errors. My GTX 275 never seems to go over 75C no matter what.
My preference is to run everything as cool as possible as they are all overclocked and running 24/7, to achieve this I have lots of airflow and run the GTX 260's in the open, taking them out of a box and having the mobo horizontal dropped temps by nearly 20C.

Good advice . Airflow is everything and a drop of 20c cantake a GPU to coolsville. However 90C on any NV GPU is 15 Deg below max. I gave up on cool or even cool with F@h long ago. There is a core fix for heat on the Core11 WU's but that has been nullified with the 10101 and 10102 WU's. All my GPUs were fine at 85% throttle until these. Then came 108 and POP of the gpu. NO advance warning

Sean

[shdbcamping]

What were the nature of your "bunch of errors"?

Sorry about the not so scientific description. Every phase of every iteration shows huge amounts of errors. However, after looking closer I realized I was still testing GPU 0, the card that won't fold. Apparently the test wasn't acknowledging my flags so, I guess I need help running the test on another of the GPUs. I read the "readme" file, still can't get it to run on a different GPU.

Please, post a log