[QUOTE=S34960zz;255187]

Had same problem myself this morning. Solution:

Go here ([URL]http://developer.nvidia.com/object/cuda_3_2_downloads.html[/URL]) and get the developer drivers (notebook drivers 260.99) for your operating system, turn off your antivirus, install driver, turn antivirus back on.[/QUOTE]

I got the 260.99 driver as you suggested, installed it, and got the message that there is no compatible hardware installed. On Nvidia's site, it actually says that it is required that all updates be obtained from lenovo, so I'm not sure what I can do at this point. My laptop is at the most recent from lenovo, but it's not current enough to run this. I'm running win7-64 on a ThinkPad T410 with a Quadro NVS 3100M. Any other ideas?

Thanks,
Doug

Hello,

Problem with mfaktc 0.16 identified! Status: not critical! :smile:

I can reproduce the issue on my GTX 470 (cc 2.0) using CUDA toolkit 3.0/3.1, too.

You can run mfaktc 0.16 compiled with CUDA toolkit 3.2 without any issues. Once you compile mfaktc 0.16 with an older version of the toolkit it will allways fail the builtin selftest so it can't harm the primenet progress (unless you hack the code to ignore the selftest or so...).

It seems to be a compiler bug.
[CODE]
unsigned int i;
float f;

i = 0xFFFFFFFF;
f = (float)i;
[/CODE]
For constants/constant expresions the compiler tries to replace them during compile time. This seems to fail for unsigned int to float conversion if the MSB of the integer is set...

Fix seems to be easy. :smile:
I'll provide an updated mfaktc 0.16p1 in the next few days.

Oliver

[QUOTE=TheJudger;255227]Hello,

Problem with mfaktc 0.16 identified! Status: not critical! :smile:

I can reproduce the issue on my GTX 470 (cc 2.0) using CUDA toolkit 3.0/3.1, too.

You can run mfaktc 0.16 compiled with CUDA toolkit 3.2 without any issues. Once you compile mfaktc 0.16 with an older version of the toolkit it will allways fail the builtin selftest so it can't harm the primenet progress (unless you hack the code to ignore the selftest or so...).

It seems to be a compiler bug.
[CODE]
unsigned int i;
float f;

i = 0xFFFFFFFF;
f = (float)i;
[/CODE]
For constants/constant expresions the compiler tries to replace them during compile time. This seems to fail for unsigned int to float conversion if the MSB of the integer is set...

Fix seems to be easy. :smile:
I'll provide an updated mfaktc 0.16p1 in the next few days.

Oliver[/QUOTE]

Thank you Oliver :smile:

Luigi

Hi Brain,

[QUOTE=Brain;255177]mfaktc 0.16: Overall Throughput > 200M/s (see attachment)

Intel Core i5-750 @ 3600 MHz (4 physical/logical cores)
Nvidia Geforce 560 Ti @ 823 MHz (384 CUDA cores)

Danke Oliver[/QUOTE]

any comparison how hard mfaktc drives your GPU compared to e.g. modern games or furmark? GPU temperatures and fan speed?

I recommend that you lower the number of instances and run on one or two cores LL tests.

Oliver

[QUOTE=TheJudger;255235]
any comparison how hard mfaktc drives your GPU compared to e.g. modern games or furmark? GPU temperatures and fan speed?
[/QUOTE]
Will further evaluate FurMark 2011 (about 4250 pts). CUDALucas' GPU load is 93% (compared to 99% for mfaktc).

[QUOTE=TheJudger;255235]
I recommend that you lower the number of instances and run on one or two cores LL tests.
[/QUOTE]
I knew you would say that... ;-) GPU's default job is CUDALucas. More needed worktype and keeps my CPU free. Screen was only benchmark setting.

[QUOTE=drh;255200]I got the 260.99 driver as you suggested, installed it, and got the message that there is no compatible hardware installed. On Nvidia's site, it actually says that it is required that all updates be obtained from lenovo, so I'm not sure what I can do at this point. My laptop is at the most recent from lenovo, but it's not current enough to run this. I'm running win7-64 on a ThinkPad T410 with a Quadro NVS 3100M. Any other ideas?

Thanks,
Doug[/QUOTE]

I am using a Dell Precision M6500 / Quadro FX 2800M, so we had the same error message about mismatched CUDA driver / runtime versions, but slightly different starting points. The "official approved" drivers from Dell (also as encouraged by Nvidia) were only at A04 / 8.17.12.5729 / Forceware 257.29 for the FX 2800M and FX3800M. Thus my move to the developer drivers for the CUDA Toolkit 3.2.

According to "NVIDIA_CUDA_C_ProgrammingGuide_3.1.pdf" Appendix A, page 98, your Quadro NVS 3100M is compute capability 1.2, 2 multiprocessors, 16 CUDA cores.

So, it would seem that your hardware has the capability, but you're still in need of correct drivers. What provides the "no compatible hardware installed" message? Could you post the exact message and which program? (At that point, one of the folks who has been doing this for a while may be able to help. I'm about out of my amateur depth on this topic.)

[QUOTE=S34960zz;255247]According to "NVIDIA_CUDA_C_ProgrammingGuide_3.1.pdf" Appendix A, page 98, your Quadro NVS 3100M is compute capability 1.2, 2 multiprocessors, 16 CUDA cores.

So, it would seem that your hardware has the capability, but you're still in need of correct drivers. What provides the "no compatible hardware installed" message? Could you post the exact message and which program? (At that point, one of the folks who has been doing this for a while may be able to help. I'm about out of my amateur depth on this topic.)[/QUOTE]

I get the message after the install completes, (double-clicking on the install icon), and the Nvidia graphics Driver version 260.99 window opens and starts with the system check. Specifically it says "Install cannot continue" "This graphics driver could not find compatible graphics hardware".

Thanks for any ideas anyone may have,
Doug

[QUOTE=drh;255299]I get the message after the install completes, (double-clicking on the install icon), and the Nvidia graphics Driver version 260.99 window opens and starts with the system check. Specifically it says "Install cannot continue" "This graphics driver could not find compatible graphics hardware".

Thanks for any ideas anyone may have,
Doug[/QUOTE]

Okay, I finally got it working. I found the site at [URL="http://www.laptopvideo2go.com"]www.laptopvideo2go.com[/URL] and downloaded the driver from there, along with a "modified" .inf file, which turns out to be the critical piece. mfaktc now runs all of the self-tests successfully and I'm ready to begin using it.

Doug

[QUOTE=James Heinrich;255195]Actually, look at [url=http://www.mersenneforum.org/showpost.php?p=255068&postcount=632]my more recent post[/url] instead, where I actually attached the file; no need to PM/email me if you need the DLL.[/QUOTE]And if you happen to need the 32-bit [i]cudart32_32_16.dll[/i], here it is.

[QUOTE=S34960zz;255187]mfaktc 0.16: Overall Throughput ~ 26 M/s (see attachment)

Brain's i5 / Geforce 560 TI is much faster, runs hotter, too, though!

I ran the same exponent as Brain's second instance, obtaining half the throughput for that instance (only took one mfaktc instance to fully load my GPU).

Intel Core i7-840QM @ 1.87 GHz (nominal) (4 cores, 8 logical CPUs)
Nvidia Quadro FX 2800M @ 600 MHz GPU core clock (96 CUDA cores), 999 MHz GPU memory clock, 1500 MHz GPU shader clock

+++++

Thank you to James Heinrich for providing me a copy of "cudart64_32_16.dll" (per post 588, [URL]http://www.mersenneforum.org/showpost.php?p=253371&postcount=588[/URL]).

+++++

drh ([URL="http://www.mersenneforum.org/showpost.php?p=255082&postcount=636%29:"]http://www.mersenneforum.org/showpost.php?p=255082&postcount=636):[/URL]



Had same problem myself this morning. Solution:

Go here ([URL]http://developer.nvidia.com/object/cuda_3_2_downloads.html[/URL]) and get the developer drivers (notebook drivers 260.99) for your operating system, turn off your antivirus, install driver, turn antivirus back on.[/QUOTE]
Looks to me like you are badly cpu bound. The only thing I can suggest is running 4 instances.

[QUOTE=henryzz;255349]Looks to me like you are badly cpu bound. The only thing I can suggest is running 4 instances.[/QUOTE]

Was this tongue-in-cheek, or am I mis-understanding something? With NumStreams=3 and CPUStreams=3 (matching Brain's settings), the GPU was running at 98+ percent. What would the expected outcome be for multiple mfaktc instances in this case?

Since yesterday, as an experiment, I've been running one core LL-D, two cores LL, and one core mfaktc feeding the GPU. Seems to be working okay.