However, if you keep a backup of the original bios and flash it back before RMA? Nividia supplies the flash utility and I think some companies provide bios updates for their cards.

[QUOTE=owftheevil;341046]However, if you keep a backup of the original bios and flash it back before RMA? Nividia supplies the flash utility and I think some companies provide bios updates for their cards.[/QUOTE]

Good point.

However, EVGA do not appear to provide BIOS updates for any of their graphics cards. Also, if someone "bricks" their card they won't be able to re-install the backup BIOS.

Although, truth be told, the most likely cause of hardware failure is because some gamer wanted to over-clock / over-voltage their card by way of a BIOS hack -- not someone wanting to down-clock their card for computational stability.

As some of the thinking here is somewhat foreign to me, I feel a need to clarify my position.

When I mess with the bios as I do, I do so with the knowledge that what I am doing might destroy my card. I also do so without any intention of RMAing any card I have fiddled with in such a way. Even if I were to RMA such a card, I most definitely would not try to hide the fact that I had fiddled with the bios.

So again, if you follow these procedures, you do so at your own risk. Please take full responsibility for what you do.

That said, I don't believe there is much risk in this procedure. For me, the slight risk is worth the gain of a fully functional card.

[QUOTE=owftheevil;341053]So again, if you follow these procedures, you do so at your own risk. Please take full responsibility for what you do.[/QUOTE]

Absolutely.

Anyone who does anything beyond buying a card and plugging it in and returning it to the retailer or manufacturer if it doesn't work take their own risks.

Carl talked about how he found a solution to his particular problem. He was very clear that this solution may not work for everyone (or, even, anyone else).

[QUOTE=owftheevil;341053]That said, I don't believe there is much risk in this procedure. For me, the slight risk is worth the gain of a fully functional card.[/QUOTE]

It's called "covering your ass with paper".

Thanks for sharing your experience! :smile:

Edit1: Just to be clear, you, Carl, did exactly the correct thing.

Edit2: The hardware and firmware and software providers, on the other hand, might not of. They are the ones covering their asses with legal "paper". Try to enforce any of it, though.

Can I request that everyone please run a quick benchmark for me, I'd like to validate (and update) the lookup table I use to generate my [url=http://www.mersenne.ca/cudalucas.php]CUDALucas throughput page[/url]. Please run this simple benchmark on a wide variety of GPUs you have available and email the results to [email]james@mersenne.ca[/email] (or PM me here if you prefer).[code]CUDALucas -info -cufftbench 1048576 8388608 1048576[/code]

Also, please run the benchmark with v2.04 if possible.

[QUOTE=owftheevil;341053]That said, I don't believe there is much risk in this procedure. For me, the slight risk is worth the gain of a fully functional card.[/QUOTE]

Indeed. And I'm happy to report that this (appears to have) solved the problem on my card. I dropped the MemClock from 2004 MHz to 1960 MHz, and my GTX 560 has now passed three CUDALucas self-tests (-r), and over two hours of your memory test. I'm going to let it run for at least another four hours, but the card never survived more than an hour without at least one error at 2004 MHz.

Absolutely no issues with the nvflash -- your above procedure worked perfectly!

Thank you such much Carl!!! :bow:

[QUOTE=chalsall;341144]I'm going to let it run for at least another four hours, but the card never survived more than an hour without at least one error at 2004 MHz.[/QUOTE]

[CODE][chalsall@hobbit memtest013]$ ./memtest 74 100000 0 | tee 201305211516.txt

Initializing test using 1850MiB of memory on device 0

Beginning test.

Position 0, Iteration 10000, Errors: 0, completed 0.14%
Position 0, Iteration 20000, Errors: 0, completed 0.27%
...
Position 73, Iteration 90000, Errors: 0, completed 99.86%
Position 73, Iteration 100000, Errors: 0, completed 100.00%[/CODE]

Happiness!!! :smile:

Now to see if that translates into error free CuLu and CPm1.

[QUOTE=owftheevil;341176]Now to see if that translates into error free CuLu and CPm1.[/QUOTE]

An excellent question. Let's run that experiment. I will report back in a couple of days...

[CODE][chalsall@hobbit cl]$ ./CUDALucas -v

CUDALucas v2.051 Alpha

[chalsall@hobbit cl]$ ./CUDALucas 29801743 | tee 201305211935.txt

Starting M29801743 fft length = 1568K
Running careful round off test for 1000 iterations. If average error >= 0.25, the test will restart with a longer FFT.
Iteration 100, average error = 0.23153, max error = 0.34766
Iteration = 132 < 1000 && err = 0.35938 >= 0.35, increasing n from 1568K
Starting M29801743 fft length = 1600K
Running careful round off test for 1000 iterations. If average error >= 0.25, the test will restart with a longer FFT.
Iteration 100, average error = 0.09618, max error = 0.14844
Iteration 200, average error = 0.10772, max error = 0.14062
Iteration 300, average error = 0.11145, max error = 0.14600
Iteration 400, average error = 0.11349, max error = 0.14844
Iteration 500, average error = 0.11475, max error = 0.15625
Iteration 600, average error = 0.11599, max error = 0.15234
Iteration 700, average error = 0.11640, max error = 0.14062
Iteration 800, average error = 0.11684, max error = 0.14062
Iteration 900, average error = 0.11732, max error = 0.15283
Iteration 1000, average error = 0.11744 < 0.25 (max error = 0.15625), continuing test.
Iteration 10000 M( 29801743 )C, 0x865251249daf325a, n = 1600K, CUDALucas v2.051 Alpha err = 0.17188 (0:54 real, 5.4364 ms/iter, ETA 44:59:20)
Iteration 20000 M( 29801743 )C, 0x2e336b1f4a815d5b, n = 1600K, CUDALucas v2.051 Alpha err = 0.17188 (0:53 real, 5.2620 ms/iter, ETA 43:31:50)
Iteration 30000 M( 29801743 )C, 0x47e2fdcaa747334b, n = 1600K, CUDALucas v2.051 Alpha err = 0.17188 (0:52 real, 5.2627 ms/iter, ETA 43:31:20)
[/CODE]

Dear Manpowre,
Can you tell me the url of your latest incarnation of CudaLucas that works on the gtx titan?
Also if you can do this:
Try modifying your code to also run Lucas-Lehmer tests on the GPUS for Fermat numbers:
The proof of correctness of this a theorem in my MSc thesis at U of Toronto.
The quadratic for this is : x^2 -5x +1=f(x)
Everything remains the same for the LL for Fn=(2^2^n)+1
Start with S0=5 instead of 4 or x[0]=5 instead of 4
and test for S(p-2)==0(mod Fn) as S(p-2)==0 iff Fn prime
Note that the recursive poly for Fermat and Mersenne numbers is the same.
That is: S_k=(S_(k-1))^2 -2
and FFT must take insto account that the type of Binay Fn is different from Mp.
M7=1111111_2 F_1=101 F_2=10001
Al