Folding@Home Benchmark 1.2.0 (GUI)

[proteneer]

FAHBench is the official Folding@Home GPU benchmark that measures the compute performance of GPUs for Folding@Home:

Download Link -> http://fahbench.com

1.2.0 Updates:
-Updated to use OpenMM 5.1 with significantly faster speed
-Accuracy checking disabled by default on GUI (enabled by default on Command Line mode)
-Stress Test option to do a very large number of steps

1.1.2 Updates:
-Made the errors due to CUDA JIT compilation more descriptive

1.1.1 Updates:
-Added a stress test option (1 million steps)
-Now checks for NaNs at the end and only once every 50000 steps for stress test
-Added performance estimate to progress counter

1.0 Updates:
-Added FAHBench GUI

Mod Edit: Updated Post - PantherX

[k1wi]

Are you getting numerous downloads of this swamping your connection? Download speed on the 6MB file is only 15-25KB/s on my 12mbit connection...

The openmm link: http://proteneer.com/blog/openmm.org ends in an 404 error.

[proteneer]

It was pretty fast for me. I updated the openmm link, thanks

[P5-133XL]

For me, it just give an error on an x64 Win7 system with 2x GTX460's that fold correctly: "Error loading library c:\temp\FAHBench_0_1\OpenMMOpenCL.dll: 126"

[artoar_11]

For me the same error:

C:\FAH> FAHBench.exe -deviceId 0 -platform CUDA -precision single

Explicit:
Error loading library C:\FAH/OpenMMOpenCL.dll: 126
C:\FAH>

Win7/64 Pro; CPU - i5 2500K; GPU - GTX 650 Ti; NV Driver 306.97

[Jesse_V]

The download was also slow for me, and I also ran into the same error:

Code: Select all

C:\Users\Admin\Downloads\bench>FAHBench.exe -deviceId 0 -platform CUDA -precisio
n single

                                          O              O
   P R O T E N E E R     C--N              \              \               N
                         |                  C              C=O           / \-C
                         C                 /               |          N-C     \
  .C-C                 C/                  C               C           |      C
 /    \          O     |                   |               /           N      |
C     C          |     |           O       C              C                 /-C
 \_N_/ \   N    _C_    C           |      /         O    /                 C
        C-/ \_C/   \N-/ \    N   /-C-\   C          |    |           O    /
        |     |           C-/ \C/     N-/ \_   N\  /C\  -C      N    |    |
        O     |           |    |            \C/  C/   N/  \_C__/ \   C-\  C
              C           O    |             |   |          |     C-/   N/ \-C
               \_C             C             O   |          O     |          |
                  \             \-O              C                C          O
                  |                               \                \
                  C    N         Folding@Home      C--N             C
                   \   |      Benchmark  (Beta)    |                |
                    N--C                           O                |
                        \        Yutong Zhao                       C=O
                         N    [email protected]                 /
                                                                 O

               for official stats, please visit www.fahbench.com

Explicit:
Error loading library C:\Users\Admin\Downloads\bench/OpenMMOpenCL.dll: 126
C:\Users\Admin\Downloads\bench>

I notice the forward slash instead of a backslash. Perhaps that's the problem?

System:

Code: Select all

22:46:28:******************************* System ********************************
22:46:28:          CPU: Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
22:46:28:       CPU ID: GenuineIntel Family 6 Model 42 Stepping 7
22:46:28:         CPUs: 8
22:46:28:       Memory: 7.98GiB
22:46:28:  Free Memory: 6.74GiB
22:46:28:      Threads: WINDOWS_THREADS
22:46:28:   On Battery: false
22:46:28:   UTC offset: -7
22:46:28:          PID: 3156
22:46:28:          CWD: C:/Users/Admin/AppData/Roaming/FAHClient
22:46:28:           OS: Windows 7 Professional N
22:46:28:      OS Arch: AMD64
22:46:28:         GPUs: 1
22:46:28:        GPU 0: NVIDIA:2 GF114 [GeForce GTX 560 Ti]
22:46:28:         CUDA: 2.1
22:46:28:  CUDA Driver: 5000
22:46:28:Win32 Service: false
22:46:28:***********************************************************************

Also, your "Folding@Home" link to http://proteneer.com/blog/folding.stanford.edu is a 404.

Even though it didn't work, I do like the molecule artwork that you were able to draw in the command line.

[Napoleon]

Do I actually have to install the whole CUDA5 Toolkit to try this? I've got Visual Studio Express already, as well as NVidia 310.90 WHQL driver, which does have CUDA5 support.

[proteneer]

Are you guys using the latest drivers? Also, can you try running

http://www.dependencywalker.com/

on the OpenMMOpenCL.dll file that came with the benchmark?

[proteneer]

Do I actually have to install the whole CUDA5 Toolkit to try this? I've got Visual Studio Express already, as well as NVidia 310.90 WHQL driver, which does have CUDA5 support.

Unfortunately yes. The CUDA platform does a lot of JIT compilation and hence requires the nvcc compiler. We assume the user has the NVIDIA GPU COMPUTING TOOLKIT 5.0 installed (and hence why its able to find CUFF).

[proteneer]

Updated the download with some additional libraries - please redownload. Should fix OpenMMOpenCL.dll errors.

[k1wi]

Looks like the 126 error is fixed by installing the CUDA toolkit.

Having installed Visual Studio 2012 and then the CUDA toolkit rebooting I was stuck at the PATH error until I restarted the computer after adding the folder to the PATH list.

Having restarted with the computer I now get the following error:

Code: Select all

Error launching CUDA compiler: -1
nvcc : fatal error : nvcc cannot find a supported cl version. Only MSVC 9.0 and 
MSVC 10.0 are supported

It would look like by installing Visual Studio 2012 I have version 11 of MSVC?

Visual C++ 2012 (known also as Visual C++ 11.0) was released on April 15, 2012, and it is currently the latest stable release.

RTM version number: 17.00.50727.1

[proteneer]

Looks like the 126 error is fixed by installing the CUDA toolkit.

Having installed Visual Studio 2012 and then the CUDA toolkit rebooting I was stuck at the PATH error until I restarted the computer after adding the folder to the PATH list.

Having restarted with the computer I now get the following error:

Code: Select all

Error launching CUDA compiler: -1
nvcc : fatal error : nvcc cannot find a supported cl version. Only MSVC 9.0 and 
MSVC 10.0 are supported

It would look like by installing Visual Studio 2012 I have version 11 of MSVC?

Yes it's looking for CUFFT in the 5.0 toolkit.

We compiled our code using Visual Studio 2008 (MSVC 9.0)- you can get the free express version here:

http://www.microsoft.com/en-us/download ... px?id=6506

Please don't use 2012 (MSVC11?) it's not supported. I'm not sure about 2010 (MSVC10).

Again, make sure the PATH includes the folder in which Visual Studio 2008's cl.exe is installed.

[k1wi]

Will I have to uninstall VS2012 before installing 08?

Might be useful to put that information on http://fahbench.com/ as your link goes to the 2012 version. ah already done.

[proteneer]

Will I have to uninstall VS2012 before installing 08?

Might be useful to put that information on http://fahbench.com/ as your link goes to the 2012 version. ah already done.

No you don't need to uninstall VS2012. You just need to set your PATH so it points to the cl.exe in VS 2008

[Napoleon]

I should have the prerequisites in place now, but "FAHBench.exe -deviceID 0 -platform CUDA -precision single" only reports "Invalid Device (please set valid deviceID by using --display-devices)". Output from --display-devices is

Code: Select all

=== 1 OpenCL platform(s) found: ===
  -- 0 --
  PROFILE = FULL_PROFILE
  VERSION = OpenCL 1.1 CUDA 4.2.1
  NAME = NVIDIA CUDA
  VENDOR = NVIDIA Corporation
  EXTENSIONS = cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_d3d9_sharing cl_nv_d
3d10_sharing cl_khr_d3d10_sharing cl_nv_d3d11_sharing cl_nv_compiler_options cl_nv_device_attribute_
query cl_nv_pragma_unroll
=== 1 OpenCL device(s) found on platform:
  -- 0 --
  DEVICE_NAME = GeForce GT 430
  DEVICE_VENDOR = NVIDIA Corporation
  DEVICE_VERSION = OpenCL 1.1 CUDA
  DRIVER_VERSION = 310.90
  DEVICE_MAX_COMPUTE_UNITS = 2
  DEVICE_MAX_CLOCK_FREQUENCY = 1620
  DEVICE_GLOBAL_MEM_SIZE = 1073741824
  -- 1 --
  DEVICE_NAME = ION
  DEVICE_VENDOR = NVIDIA Corporation
  DEVICE_VERSION = OpenCL 1.0 CUDA
  DRIVER_VERSION = 310.90
  DEVICE_MAX_COMPUTE_UNITS = 2
  DEVICE_MAX_CLOCK_FREQUENCY = 1500
  DEVICE_GLOBAL_MEM_SIZE = 536870912
Invalid Platform (please use either OpenCL or CUDA)

FWIW, FAH v7.2.9 reports

Code: Select all

20:08:11:******************************* System ********************************
20:08:11:          CPU: Intel(R) Atom(TM) CPU 330 @ 1.60GHz
20:08:11:       CPU ID: GenuineIntel Family 6 Model 28 Stepping 2
20:08:11:         CPUs: 4
20:08:11:       Memory: 2.75GiB
20:08:11:  Free Memory: 1.53GiB
20:08:11:      Threads: WINDOWS_THREADS
20:08:11:   On Battery: false
20:08:11:   UTC offset: 2
20:08:11:          PID: 5004
20:08:11:          CWD: C:/FAH/Data
20:08:11:           OS: Windows 7 Ultimate
20:08:11:      OS Arch: AMD64
20:08:11:         GPUs: 2
20:08:11:        GPU 0: NVIDIA:1 C?? [ION]
20:08:11:        GPU 1: NVIDIA:2 GF108 [GeForce GT 430]
20:08:11:         CUDA: 2.1
20:08:11:  CUDA Driver: 5000
20:08:11:Win32 Service: false
20:08:11:***********************************************************************

I didn't install the development driver in the CUDA toolkit, as it seemed to be pretty old (306.23, IIRC). 310.90 WHQL should support CUDA5. The nvcc.exe is the first thing in the PATH, now that the CUDA toolkit is installed. Tried both regular console and admin console, same result.

I can run for example "Stereo Disparity Computation" sample program which came with the toolkit, so I assume the toolkit installation was successful. According to GPU-Z it actually seems to do something on the GT430 because core clock briefly jumps from idle (50MHz) to 3D (810MHz).

Stereo Disparity Computation (SAD SIMD Intrinsics)

A CUDA program that demonstrates how to compute a stereo disparity map using SIMD SAD (Sum of Absolute Difference) intrinsics. Requires Compute Capability 2.0 or higher. Minspec: GeForce 400

Code: Select all

GPU Device 0: "GeForce GT 430" with compute capability 2.1

> GPU device has 2 Multi-Processors, SM 2.1 compute capabilities

Loaded <../3_Imaging/stereoDisparity/data/stereo.im0.640x533.ppm> as image 0
Loaded <../3_Imaging/stereoDisparity/data/stereo.im1.640x533.ppm> as image 1
Launching CUDA stereoDisparityKernel()
Input Size  [640x533], Kernel size [17x17], Disparities [-16:0]
GPU processing time : 12.7466 (ms)
Pixel throughput    : 26.762 Mpixels/sec
GPU Checksum = 4293895789, GPU image: <output_GPU.pgm>
Computing CPU reference...

What else could I try? Or is it just that my GT430 is too old to run the benchmark?

EDIT: if I try to run cl.exe from the console, I get "The program can't start because mspdb110.dll is missing from your computer. Try reinstalling the program to fix this problem." Well, this is a fresh install of the Visual Studio 2012 Express and all Windows updates are also installed, so I'm stumped with that one.

EDIT2: Never mind, I suppose I'll just have to try with the older version of VS Express.