Folding Forum

bruce wrote:According to popular opinion, drivers 310.90 provide better FAH stability than previous versions so I have been running on them.

Not according to many of the posts on the NVidia Forums "Popular Topics" here and 38 pages of feedback on 310.70/310.90 drivers here.
From page 38:

Deadlockk wrote:I updated from the consumer preview to regular to Windows 8 saturday.
Running on a Geforce GTX 470.
After installing the 310.90 driver, my PC no longer boots. The windows loading icon is shown and then the screen flickers in tints of black.
I had the same result with the 310.70.

And there are 37 more pages with comments just like that. Some people swear by 310.90, and other people swear at it. It's just the luck of the draw.

bruce wrote:NOTE: Installing 310.90 graphics drivers includes something called NVidia Update which probably revises some of the supporting .dlls that TOOLKIT delivered.

No, it is an update service that periodically checks for updates from NVidia's website. A popup will notify you when updates are available.
NVIDIA Update

What is NVIDIA Update?
NVIDIA Update keeps your PC up-to-date with the latest NVIDIA drivers by notifying you when a new driver is available and directing you to the driver on http://www.nvidia.com.

PinHead wrote:Is the -platform OpenCL platformId independant? Can I run -deviceId 0,1 -platform OpenCL on a system with 1 AMD and 1 NVidia?

I get:
Explicit:
17.6456 ns/day

Implicit not supported on multiple devices.

AMD/NVidia
Win 7 64
FAHBench 0_4

I also get a no platformId specified.

Oddly enough, looks like you can

Spongebob25 wrote:OpenCL works fine but now I'm back to having an error running cuda.--OpenMMCUDA.dll:126. I have the tool kit installed and VS.

Which version of CUDA and VS are you using? Must be CUDA 5.0 and VS 2008

art_l_j_PlanetAMD64 wrote:Some people swear by 310.90, and other people swear at it. It's just the luck of the draw.

OR: all the drastic changes are done for 600's / Kepler architecture. Changes, if any, related to 500 and especially 400 series Fermi GPUs probably are much more conservative.

FAHBench

Is something supposed to be written to the logFile.txt file? Mine always shows zero bytes.

proteneer wrote:

Spongebob25 wrote:OpenCL works fine but now I'm back to having an error running cuda.--OpenMMCUDA.dll:126. I have the tool kit installed and VS.

Which version of CUDA and VS are you using? Must be CUDA 5.0 and VS 2008

Yes, CUDA 5.0 and VS2008 are installed.

Spongebob25 wrote:

proteneer wrote:

Spongebob25 wrote:OpenCL works fine but now I'm back to having an error running cuda.--OpenMMCUDA.dll:126. I have the tool kit installed and VS.

Which version of CUDA and VS are you using? Must be CUDA 5.0 and VS 2008

Yes, CUDA 5.0 and VS2008 are installed.

Can you run http://www.dependencywalker.com/ on OpenMMCUDA.dll?

Just curious...
But will there be an option to benchmark different projects later on?

I wouldn't mind seeing how P6903 works on a GPU (2,533,797 atoms).

SolidSteel144 wrote:But will there be an option to benchmark different projects later on?

Probably not. The goal of FAHBench is to compare hardware using some kind of a standard. Tom's Hardware (or your favorite hardware review site) can't get consistent numbers for FAH because they have not had access to a predefined test. Stanford can also use it to compare the results of FahCore code optimization.

WUs from various projects are already benchmarked by Stanford for comparison with other projects to establish the Points. They use standard hardware.

A standard protein across variable hardware is one thing. Standard hardware across variable proteins is another. Variable proteins with variable hardware would produce a lot more data but there really is no way to use all that information.

With the new core you can pretty much rip out the .xml files and just replace files used in the benchmark out.

So yes, you can benchmark new projects (though there is some minor work on your end). Although as a FAH User, you probably won't notice too much difference, but everything underneath is being completely restructured.

So, as it turns out, I had not in fact downloaded the beta Intel drivers when I was testing the HD4000. They're not the easiest thing to find, and later-versioned releases do not appear to include certain things, like OpenCL 1.2 for the GPU. I managed to find the beta drivers, which are version 9.17.10.2885, and install them. display-devices now outputs:
Basically, the CPU is no longer visible, and the GPU has gone from OpenCL 1.1 to 1.2 compliant. But FAHBench still fails. It gets farther than it did before, where it would crash after a couple of seconds; now, it gets as far as "Checking for accuracy...done" on the explicit test before crashing without any text output.

I have updated viewtopic.php?f=38&t=23440 with hopefully all of the new scores.

k1wi

rjbelans wrote:

proteneer wrote:no actually i do mean 2 cards working on the same WU - the problem is that kinda don't scale very well when testing internally (and they only work on explicit). I'll probably need to add another flag specifying explicit/implicit - so sorry if I end up breaking all your batch files =P

That may be a bit of an understatement. I did a small check using CUDA and my 590s. Adding one gpu increased the performance by about 33%, adding a third gpu increased the performance about another 25% of the original score, and then a fourth showed no improvement over three. This was done on the same 590s I posted about in the other thread (linked above), but I had a couple of additional things running that may have reduced scores from yesterday. I will have to run this again when I have nothing else to do on this computer to see if the reduction seen on a single gpu from yesterday to today was because of other processes or the change from v0.3 to v0.4.

It'd be quite interesting to see whether your 2 GPU result is helped/hindered by having the SLI bridge on the board. That is, what difference you would have running deviceId 0,2.

Can you run http://www.dependencywalker.com/ on OpenMMCUDA.dll?

Did as you asked. Got--->Error: At least one required implicit or forwarded dependency was not found. Warning: A least one delay-load dependency module was not found. Warning: At least one module has an unresolved import due to missing export function in a delay-load dependent module.

CUFFT32_50_35.DLL--->Error opening file. The system cannot find the file specified.
MSVCR90.DLL--->Error opening file. The system cannot find the file specified.
IESHIMS.DLL--->Error opening file. The system cannot find the file specified.

Spongebob25 wrote:

Can you run http://www.dependencywalker.com/ on OpenMMCUDA.dll?

Did as you asked. Got--->Error: At least one required implicit or forwarded dependency was not found. Warning: A least one delay-load dependency module was not found. Warning: At least one module has an unresolved import due to missing export function in a delay-load dependent module.

CUFFT32_50_35.DLL--->Error opening file. The system cannot find the file specified.
MSVCR90.DLL--->Error opening file. The system cannot find the file specified.
IESHIMS.DLL--->Error opening file. The system cannot find the file specified.

CUFFT is the culprit here, try to find the folder in which CUFFT32_50_35.DLL is installed, and set your windows PATH to point to that as well.