FAHBench results (Most Powerful GPU)

A forum for discussing FAH-related hardware choices and info on actual products (not speculation).

Moderator: Site Moderators

Forum rules
Please read the forum rules before posting.

FAHBench results (Most Powerful GPU)

Postby k1wi » Mon Jan 14, 2013 9:28 pm

Most Powerful GPUs using FAHBench

disclaimer: These are unofficial values and, amongst other factors, are dependent on the version of FAHBench used and the integrity of the folder. Furthermore, completing FAHBench at a given clockspeed does not ensure F@H stability. The intention of these tables are to concisely record the values associated with each GPU and highlight top performers.

OpenCL SP Explicit:
25.3514 ns/day k1wi v0.3 Nvidia GTX570 @ 1020MHz (2056 memory) http://i49.tinypic.com/35b5xef.png
23.4681 ns/day SolidSteel144 v0.3 AMD 7970 1050/1500 MHz Catalyst 13.1 w/ latest OpenCL runtime.
21.1817 ns/day rjbelans Nvidia GTX 590 @ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.3
20.2295 ns/day NookieBandit Nvidia GTX 570 OC @ 830 MHz (750 MHz default): Intel Core 2 Quad Q9450 @ 2.66GHz Dell XPS 420; 4x2GB @1333 MHz
19.0149 ns/day Rattledagger AMD 7950 at 900 MHz
18.1 ns/day [OFFICIAL] Nvidia Tesla K20
17.0609 ns/day Spongebob25 Nvidia 670 GTX SC @1188MHz
16.1 ns/day [OFFICIAL] Nvidia GTX 660Ti
13.9299 ns/day Jesse_V v0.2 Nvidia GTX 560TI. GPU clock 823 MHz, Memory 1002 MHz, Shader 1645 MHz
13.6545 ns/day Zagen30 v0.4 2xNvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
13.5521 ns/day PinHead Nvidia Tesla C2050
13.2591 ns/day P5-133XL Nvidia GTX550@1000 3.2GHz P4, 2GB RAM, X32 Win7 SP1, Nvidia 306.94
11.6728 ns/day Zagen30 v0.2 Nvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
10.5417 ns/day DarkFoss AMD 6950 oc [extra shaders unlocked with the bios switch] 1536 shaders unified GPU clock 840MHz memory 1325MHz Driver 13.1(atiumdag 9.12.0.0) AMD app-sk 2.7 Win 7 64x
10.1111 ns/day PantherX v0.5 Nvidia GTX 675M @ 620 MHz
9.38559 ns/day artoar_11 v0.1 Nvidia GTX 650Ti OC @ 1097 MHz
8.56641 ns/day PinHead Nvidia GTX 550Ti
8.55352 ns/day Rattledagger AMD 5850 @ stock
8.42751 ns/day SolidSteel144 v0.5 AMD 6870 (950/1050 MHz) Catalyst 13.2 w/ latest OpenCL APP SDK.
6.35404 ns/day somata v0.5 AMD 6850 (stock, 775/1000) - Cat 13.1
4.94474 ns/day Zagen30 v0.5 Intel HD 4000 @ stock 3770K @4.3 GHz, 8 GB DDR3-2133
3.28365 ns/day Napoleon Nvidia GT430
2.93 ns/day Flaschie AMD 7660D @ stock
2.66677 ns/day SolidSteel144 v0.3 AMD 6570 675/900 MHz Catalyst 13.1 w/ latest OpenCL runtime.
1.69979 ns/day somata v0.5 AMD 6630M (stock, 485/600) - Cat 13.1
0.314806 ns/day Napoleon Nvidia ION iGPU


OpenCL SP Implicit:
125.036 ns/day SolidSteel144 v0.5 AMD 7970 1050/1500 MHz Catalyst 13.2 w/ latest OpenCL runtime.
114.165 ns/day k1wi v0.3 Nvidia GTX570 @ 1020MHz (2056 memory) http://i49.tinypic.com/35b5xef.png
97.407 ns/day rjbelans Nvidia GTX 590 @ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.3
95.0495 ns/day NookieBandit Nvidia GTX 570 OC @ 830 MHz (750 MHz default): Intel Core 2 Quad Q9450 @ 2.66GHz Dell XPS 420; 4x2GB @1333 MHz
84.5 ns/day [OFFICIAL] Nvidia Tesla K20
84.1613 ns/day Rattledagger 7950 at 900 MHz
83.6318 ns/day Spongebob25 Nvidia 670 GTX SC @1188MHz
77.0 ns/day [OFFICIAL] Nvidia GTX 660Ti
75.2482 ns/day DarkFoss AMD 6950 oc [extra shaders unlocked with the bios switch] 1536 shaders unified GPU clock 840MHz memory 1325MHz Driver 13.1(atiumdag 9.12.0.0) AMD app-sk 2.7 Win 7 64x
69.0537 ns/day Jesse_V v0.2 Nvidia GTX 560TI. GPU clock 823 MHz, Memory 1002 MHz, Shader 1645 MHz
61.4903 ns/day PinHead Nvidia Tesla C2050
61.0539 ns/day SolidSteel144 v0.5 AMD 6870 (950/1050 MHz) Catalyst 13.2 w/ latest OpenCL APP SDK.
56.0748 ns/day Zagen30 v0.2 Nvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
52.2023 ns/day Rattledagger AMD 5850 @ stock
51.3674 ns/day PantherX v0.5 Nvidia GTX 675M @ 620 MHz
47.9787 ns/day artoar_11 v0.1 Nvidia GTX 650Ti OC @ 1097 MHz
48.56641 ns/day PinHead Nvidia GTX 550Ti
45.9501 ns/day P5-133XL Nvidia GTX550@1000 3.2GHz P4, 2GB RAM, X32 Win7 SP1, Nvidia 306.94
40.7671 ns/day somata v0.5 AMD 6850 (stock, 775/1000) - Cat 13.1
26.6441 ns/day Zagen30 v0.5Intel HD 4000 @ stock 3770K @4.3 GHz, 8 GB DDR3-2133
20.26 ns/day Flaschie AMD 7660D @ stock
18.7606 ns/day SolidSteel144 v0.3 AMD 6570 675/900 MHz Catalyst 13.1 w/ latest OpenCL runtime.
18.4848 ns/day Napoleon Nvidia GT430
13.2178 ns/day somata v0.5 AMD 6630M (stock, 485/600) - Cat 13.1
3.6432 ns/day Napoleon Nvidia ION iGPU


CUDA SP Explicit:
35.5702 ns/day rjbelans 3xNvidia GTX 590@ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.4
35.4081 ns/day rjbelans 4xNvidia GTX 590@ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.4
30.7309 ns/day rjbelans 2xNvidia GTX 590@ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.4
29.4911 ns/day k1wi v0.2 Nvidia GTX570 @ 1020MHz (2056 memory) http://i47.tinypic.com/2cgjvjl.png
25.6 ns/day [OFFICIAL] Nvidia Tesla K20
24.0401 ns/day rjbelans Nvidia GTX 590 @ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.3
23.8437 ns/day Zagen30 v0.4 2xNvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
22.8 ns/day [OFFICIAL] Nvidia GTX 660Ti
22.6641 ns/day Spongebob25 Nvidia 670 GTX SC @1188MHz
18.0406 ns/day Jesse_V v0.2 Nvidia GTX 560TI. GPU clock 823 MHz, Memory 1002 MHz, Shader 1645 MHz
15.8791 ns/day PinHead Nvidia Tesla C2050
15.0846 ns/day Zagen30 v0.2 Nvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
14.3662 ns/day P5-133XL Nvidia GTX 460@825, Q6600@3.0 4GB 2xGTX 460@825, X64 Win7 SP1, Driver 306.94
14.3141 ns/day artoar_11 v0.1 Nvidia GTX 650Ti OC @ 1097 MHz
13.2684 ns/day P5-133XL Nvidia GTX550 @1000 3.2GHz P4, 2GB RAM, X32 Win7 SP1, Nvidia 306.94
11.3592 ns/day PinHead Nvidia GTX 550Ti

CUDA SP Implicit:
114.2 ns/day [OFFICIAL] Nvidia Tesla K20
110.698 ns/day k1wi v0.2 Nvidia GTX570 @ 1020MHz (2056 memory) http://i47.tinypic.com/2cgjvjl.jpg
104.969 Spongebob25 Nvidia 670 GTX SC @1188MHz
103.7 ns/day [OFFICIAL] Nvidia GTX 660Ti
85.6293 ns/day rjbelans Nvidia GTX 590 @ 720/1440/1728 Win7 Ult 64 - CPU 980X @ 4.0 GHz - 12GB RAM @ 1600 MHz - FAHBench 0.3
67.6586 ns/day artoar_11 v0.1 Nvidia GTX 650Ti OC @ 1097 MHz
67.295 ns/day Jesse_V v0.2 Nvidia GTX 560TI. GPU clock 823 MHz, Memory 1002 MHz, Shader 1645 MHz
62.7223 ns/day Zagen30 v0.2 Nvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz
55.0354 ns/day PinHead Nvidia Tesla C2050
52.912 ns/day P5-133XL Nvidia GTX 460 @825, Q6600@3.0 4GB, X64 Win7 SP1, Driver 306.94
46.0997 ns/day P5-133XL Nvidia GTX550 @1000 3.2GHz P4, 2GB RAM, X32 Win7 SP1, Nvidia 306.94
20.203 ns/day PinHead Nvidia GTX 550Ti

Notes:
This is a place where people can start posting their benchmark scores. That way the debug thread is kept clean and there is an incentive for people to use the benchmark tool :)

I am currently moving *slowly* towards allowing two results per GPU type: 1 at stock settings as per wikipedia and 1 representing the maximum value overclocked. Top 5 results may be exempt once extra results are in.

For example, the GTX570 will be permitted two results: 1 at stock and 1 maximum when overclocked. The GTX560Ti will also be allowed two results.

As of now a link to a screenshot should be a requirement, as is notation of the version of FAHBench you are using. Information on the GPU, including clockspeed is advisable.

With numerous people submitting multiple results and scores subject to variation between tests, if you submit multiple times for a specific piece of hardware I reserve the right to keep your existing score unless there is a margin of improvement of >5%.

Please use the following format:

[System details including FAHBench version]
[Platform]
Explicit:
xx.xxxx ns/day
Implicit:
xx.xxxx ns/day

I will update this post as people send in results. As mentioned here: viewtopic.php?f=16&t=23434&p=233764#p233764, it is probably likely that OpenCL will become the primary table.

k1wi
Last edited by k1wi on Sun Mar 10, 2013 8:53 pm, edited 16 times in total.
k1wi
 
Posts: 1318
Joined: Tue Sep 22, 2009 10:48 pm

Re: Most Powerful GPU (FAHBench)

Postby Jesse_V » Tue Jan 15, 2013 1:59 am

FAHBench v0.2
Nvidia GTX 560TI. GPU clock 823 MHz, Memory 1002 MHz, Shader 1645 MHz

CUDA: http://www.jessevictors.com/files/perma ... SINGLE.png
Explicit:
18.0406 ns/day
Implicit:
67.295 ns/day

OpenCL: http://www.jessevictors.com/files/perma ... SINGLE.png
Explicit:
13.9299 ns/day
Implicit:
69.0537 ns/day
User avatar
Jesse_V
 
Posts: 2762
Joined: Mon Jul 18, 2011 4:44 am
Location: Logan, Utah, USA

Re: Most Powerful GPU (FAHBench)

Postby Zagen30 » Tue Jan 15, 2013 2:55 am

FAHBench v0.2
2xNvidia GTX 460 GPU clock 823 MHz, Memory 1804 MHz, Shader 1645 MHz

CUDA SP: http://imageshack.us/photo/my-images/171/fahbenchcudasp.png/
Explicit:
14.9828 ns/day
Implicit:
60.5381 ns/day

CUDA DP: http://imageshack.us/photo/my-images/685/fahbenchcudadp.png/
Explicit:
4.03733 ns/day
Implicit:
5.70847 ns/day

EDIT: Finally got OpenCL to work

OpenCL SP: http://imageshack.us/photo/my-images/809/fahbenchopenclsingle.png/
Explicit:
11.6728 ns/day
Implicit:
56.0748 ns/day

OPenCL DP: http://imageshack.us/photo/my-images/189/fahbenchopencldouble.png/
Explicit:
3.4908 ns/day
Implicit:
4.88287 ns/day
Last edited by Zagen30 on Wed Jan 16, 2013 2:26 am, edited 2 times in total.
Image
Zagen30
 
Posts: 1799
Joined: Tue Mar 25, 2008 12:45 am

Re: Most Powerful GPU (FAHBench)

Postby P5-133XL » Tue Jan 15, 2013 9:07 am

Image
P5-133XL
Site Moderator
 
Posts: 4006
Joined: Sun Dec 02, 2007 4:36 am
Location: Salem. OR USA

Re: Most Powerful GPU (FAHBench)

Postby artoar_11 » Tue Jan 15, 2013 10:46 am

Add to all my results - Memory 1350 MHz, FAHBench version - 01

artoar_11 Nvidia GTX 650Ti OC @ 1097 MHz, Memory 1350 MHz
i5 2500K @4.30GHz; P8P67 Pro; 2x4GB @1600MHz; ENGTX 460@750MHz/1GB; Win7/64b (SMP+GPU)
i5 2500K @4.20GHz; Z68 Extreme3 Gen3; 2x2GB @1600MHz; GTX 650Ti @1.12GHz; Win7/64b (SMP+GPU)
User avatar
artoar_11
 
Posts: 389
Joined: Sun Nov 22, 2009 8:42 pm
Location: Bulgaria /Team #32435/artoar_home

Re: Most Powerful GPU (FAHBench)

Postby proteneer » Tue Jan 15, 2013 4:24 pm

This is a great list - thanks for doing this.
He who has a why to live for can bear almost any how

www.proteneer.com
User avatar
proteneer
Pande Group Member
 
Posts: 110
Joined: Fri Sep 28, 2012 11:03 pm
Location: Stanford, CA

Re: Most Powerful GPU (FAHBench)

Postby Napoleon » Tue Jan 15, 2013 7:18 pm

Caught the batch file kraze too, so here's my (very clumsy) version for my ION iGPU:
Code: Select all
@echo Results will be appended to ION.txt
@echo FAHBench 0.3 >> ION.txt
@echo NVidia ION, GPU clock 500MHz, memory 1218MHz, shaders 1500MHz >> ION.txt
@echo - >> ION.txt

@echo OpenCL SP >> ION.txt
start /B /WAIT /HIGH FAHBench --disable-splash -deviceId 1 -platformId 0 -platform OpenCL -precision single >> ION.txt
@echo - >> ION.txt


FAHBench 0.3
NVidia ION, GPU clock 500MHz, memory 1218MHz, shaders 1500MHz

OpenCL SP
Explicit:
0.314806 ns/day
Implicit:
3.6432 ns/day

This is the best my ION iGPU can do, viewtopic.php?f=16&t=23434&start=75#p233820. It'd be great if k1wi takes the time to edit my ION result in the original post. This is the last time I resurrect the ION for FAH, or FAH benchmarking. I promise... :wink:

Full v0_3 GT430 results, including DP, to follow as soon as the batch file for that is done running.
Win7 64bit, FAH v7, OC'd
2C/4T Atom330 3x685MHz - GT430 2x810MHz - ION1 9400M 3x466.7MHz
4x OGR27 - Core_15 - display & Core_11

Mint 14 64bit, FAH v7 & v6, at stock
2C/2T E2220 - 9400GT
cpu:2 Core_a? - display @ Core_11 (v6 WINE)
User avatar
Napoleon
 
Posts: 1080
Joined: Wed May 26, 2010 2:31 pm
Location: Finland

Re: Most Powerful GPU (FAHBench)

Postby artoar_11 » Tue Jan 15, 2013 9:15 pm

FAHBench v0.3
GTX 650 Ti OC @1097 MHz, Memory 1350 MHz

CUDA -precision single
Explicit:
14.2457 ns/day
Implicit:
67.5528 ns/day
-------------------
OpenCL -precision single
Explicit:
9.38315 ns/day
Implicit:
47.867 ns/day

Screenshot: http://storage4.album.bg/19f/fahbench_0 ... 292219.png
User avatar
artoar_11
 
Posts: 389
Joined: Sun Nov 22, 2009 8:42 pm
Location: Bulgaria /Team #32435/artoar_home

Re: FAHBench results (Most Powerful GPU)

Postby Napoleon » Tue Jan 15, 2013 10:57 pm

GT430.bat
Code: Select all
@echo Results will be appended to GT430.txt
@echo FAHBench 0.3 >> GT430.txt
@echo NVidia GT430, GPU clock 810MHz, memory 810MHz, shaders 1620MHz >> GT430.txt
@echo - >> GT430.txt

@echo CUDA SP >> GT430.txt
start /B /WAIT /HIGH FAHBench --disable-splash -deviceId 0 -platform CUDA -precision single >> GT430.txt
@echo - >> GT430.txt

@echo OpenCL SP >> GT430.txt
start /B /WAIT /HIGH FAHBench --disable-splash -deviceId 0 -platformId 0 -platform OpenCL -precision single >> GT430.txt
@echo - >> GT430.txt

@echo CUDA DP >> GT430.txt
start /B /WAIT /HIGH FAHBench --disable-splash -deviceId 0 -platform CUDA -precision double >> GT430.txt
@echo - >> GT430.txt

@echo OpenCL DP >> GT430.txt
start /B /WAIT /HIGH FAHBench --disable-splash -deviceId 0 -platformId 0 -platform OpenCL -precision double >> GT430.txt
@echo - >> GT430.txt


A bit of editing, and you get the list below. OpenCL beats native CUDA at SP implicit... most interesting!

FAHBench 0.3
NVidia GT430, GPU clock 810MHz, memory 810MHz, shaders 1620MHz

CUDA SP
Explicit:
4.32625 ns/day
Implicit:
16.7943 ns/day

OpenCL SP
Explicit:
3.32794 ns/day
Implicit:
18.593 ns/day

CUDA DP
Explicit:
1.20118 ns/day
Implicit:
1.78589 ns/day

OpenCL DP
Explicit:
1.02851 ns/day
Implicit:
1.50206 ns/day
-
User avatar
Napoleon
 
Posts: 1080
Joined: Wed May 26, 2010 2:31 pm
Location: Finland

Re: FAHBench results (Most Powerful GPU)

Postby Rattledagger » Wed Jan 16, 2013 9:04 pm

Numbers from v0.3, using my own version of batchfile so a picture doesn't really tell much, but anyway.

Amd-7950, 900 MHz: picture
Single-precision tests
Explicit:
19.0149 ns/day
Implicit:
84.1613 ns/day

Double-precision tests
Explicit:
6.37455 ns/day
Implicit:
20.2941 ns/day

Code: Select all
New benchmark-run, date and time is listed below
16.01.2013
17:46
                                                                               
                                          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    proteneer@gmail.com                 /           
                                                                 O           
                                                                               
               for official stats, please visit www.fahbench.com               

[1] compatible platform(s):
  -- 0 --
  PROFILE = FULL_PROFILE
  VERSION = OpenCL 1.2 AMD-APP (1084.2)
  NAME = AMD Accelerated Parallel Processing
  VENDOR = Advanced Micro Devices, Inc.

(2) device(s) found on platform 0:
  -- 0 --
  DEVICE_NAME = Tahiti
  DEVICE_VENDOR = Advanced Micro Devices, Inc.
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.2)

  -- 1 --
  DEVICE_NAME =        Intel(R) Core(TM) i7-2700K CPU @ 3.50GHz
  DEVICE_VENDOR = GenuineIntel
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.2)

Invalid Platform (please use either OpenCL or CUDA)
Single-precision tests
Explicit:
Checking for accuracy...done
19.0149 ns/day
Implicit:
Checking for accuracy...done
84.1613 ns/day
Double-precision tests
Explicit:
Checking for accuracy...done
6.37455 ns/day
Implicit:
Checking for accuracy...done
20.2941 ns/day


Ati-5850, whatever is default speed. Don't have a picture here:
Single-precision tests
Explicit:
0.677802 ns/day
Implicit:
51.5667 ns/day

Double-precision tests
Explicit:
0.309532 ns/day
Implicit:
6.70594 ns/day

Code: Select all
New benchmark-run, date and time is listed below
16.01.2013
17:50
                                                                               
                                          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    proteneer@gmail.com                 /           
                                                                 O           
                                                                               
               for official stats, please visit www.fahbench.com               

[1] compatible platform(s):
  -- 0 --
  PROFILE = FULL_PROFILE
  VERSION = OpenCL 1.2 AMD-APP (1084.2)
  NAME = AMD Accelerated Parallel Processing
  VENDOR = Advanced Micro Devices, Inc.

(2) device(s) found on platform 0:
  -- 0 --
  DEVICE_NAME = Cypress
  DEVICE_VENDOR = Advanced Micro Devices, Inc.
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.2)

  -- 1 --
  DEVICE_NAME = Intel(R) Core(TM) i7 CPU         920  @ 2.67GHz
  DEVICE_VENDOR = GenuineIntel
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.2)

Invalid Platform (please use either OpenCL or CUDA)
Single-precision tests
Explicit:
Checking for accuracy...done
0.677802 ns/day
Implicit:
Checking for accuracy...done
51.5667 ns/day
Double-precision tests
Explicit:
Checking for accuracy...done
0.309532 ns/day
Implicit:
Checking for accuracy...done
6.70594 ns/day
Rattledagger
 
Posts: 189
Joined: Thu Dec 06, 2007 9:48 pm
Location: Norway

Re: FAHBench results (Most Powerful GPU)

Postby k1wi » Wed Jan 16, 2013 9:27 pm

@Napoleon: I'm glad someone else had the same OpenCL results as me.

@Rattledagger: Wow, those 5850 explicit numbers are low, but the implicit values much more healthy! I guess it goes to show the improvements being made in GPU design and possibly the difficulties associated with benchmarking.
k1wi
 
Posts: 1318
Joined: Tue Sep 22, 2009 10:48 pm

Re: FAHBench results (Most Powerful GPU)

Postby Rattledagger » Fri Jan 18, 2013 11:36 pm

In case the new 13.1-drivers gives any differences, new benchmark-runs from my two systems:

Amd-7950, 900 MHz:
Single-precision tests
Explicit:
18.4186 ns/day
Implicit:
80.3796 ns/day

Double-precision tests
Explicit:
6.38072 ns/day
Implicit:
20.038 ns/day

Code: Select all
New benchmark-run, date and time is listed below
19.01.2013
00:03
                                                                               
                                          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    proteneer@gmail.com                 /           
                                                                 O           
                                                                               
               for official stats, please visit www.fahbench.com               

[1] compatible platform(s):
  -- 0 --
  PROFILE = FULL_PROFILE
  VERSION = OpenCL 1.2 AMD-APP (1084.4)
  NAME = AMD Accelerated Parallel Processing
  VENDOR = Advanced Micro Devices, Inc.

(2) device(s) found on platform 0:
  -- 0 --
  DEVICE_NAME = Tahiti
  DEVICE_VENDOR = Advanced Micro Devices, Inc.
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.4)

  -- 1 --
  DEVICE_NAME =        Intel(R) Core(TM) i7-2700K CPU @ 3.50GHz
  DEVICE_VENDOR = GenuineIntel
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.4)

Invalid Platform (please use either OpenCL or CUDA)
Single-precision tests
Explicit:
Checking for accuracy...done
18.4186 ns/day
Implicit:
Checking for accuracy...done
80.3796 ns/day
Double-precision tests
Explicit:
Checking for accuracy...done
6.38072 ns/day
Implicit:
Checking for accuracy...done
20.038 ns/day


Ati-5850, default clock:
Single-precision tests
Explicit:
8.55352 ns/day
Implicit:
52.2023 ns/day

Double-precision tests
Explicit:
2.89185 ns/day
Implicit:
6.69461 ns/day

Code: Select all
New benchmark-run, date and time is listed below
18.01.2013
23:57
                                                                               
                                          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    proteneer@gmail.com                 /           
                                                                 O           
                                                                               
               for official stats, please visit www.fahbench.com               

[1] compatible platform(s):
  -- 0 --
  PROFILE = FULL_PROFILE
  VERSION = OpenCL 1.2 AMD-APP (1084.4)
  NAME = AMD Accelerated Parallel Processing
  VENDOR = Advanced Micro Devices, Inc.

(2) device(s) found on platform 0:
  -- 0 --
  DEVICE_NAME = Cypress
  DEVICE_VENDOR = Advanced Micro Devices, Inc.
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.4)

  -- 1 --
  DEVICE_NAME = Intel(R) Core(TM) i7 CPU         920  @ 2.67GHz
  DEVICE_VENDOR = GenuineIntel
  DEVICE_VERSION = OpenCL 1.2 AMD-APP (1084.4)

Invalid Platform (please use either OpenCL or CUDA)
Single-precision tests
Explicit:
Accuracy checking disabled.
8.55352 ns/day
Implicit:
Accuracy checking disabled.
52.2023 ns/day
Double-precision tests
Explicit:
Accuracy checking disabled.
2.89185 ns/day
Implicit:
Accuracy checking disabled.
6.69461 ns/day


Where's a huge difference for the implicit on the 5850, maybe this has to do was running 12.11beta4 on this system before?
Also, didn't do the accuracy-tests for this system, but if this has any effect it's a bug in the benchmark...

Also worth noting, according to CCC the GPU-usage on the 7950 was 85% for single-precision and 94% for double-precision.
According to GPU-Z, the GPU-usage on the 5850 was 97% for single-precision and 99% for double-precision.
Rattledagger
 
Posts: 189
Joined: Thu Dec 06, 2007 9:48 pm
Location: Norway

Re: FAHBench results (Most Powerful GPU)

Postby Zagen30 » Fri Jan 18, 2013 11:39 pm

I've added OpenCL results to my post above.
Zagen30
 
Posts: 1799
Joined: Tue Mar 25, 2008 12:45 am

Re: FAHBench results (Most Powerful GPU)

Postby proteneer » Sat Jan 19, 2013 2:45 am

Thanks guys - very useful.
User avatar
proteneer
Pande Group Member
 
Posts: 110
Joined: Fri Sep 28, 2012 11:03 pm
Location: Stanford, CA

Re: FAHBench results (Most Powerful GPU)

Postby SolidSteel144 » Sat Jan 19, 2013 5:39 pm

FAHBench 0.3
Radeon HD 7970 1050/1500 MHz Catalyst 13.1 w/ latest OpenCL runtime.

OpenCL SP:
Explicit - 23.4681 ns/day
Implicit - 105.033 ns/day

OpenCL DP:
Explicit - 7.94841 ns/day
Implicit - 25.748 ns/day

I hope this gives us a good idea how the new fahcores will perform.
AMD's performance (at least with GCN) has improved quite a bit.
Last edited by SolidSteel144 on Sat Jan 19, 2013 7:01 pm, edited 1 time in total.
User avatar
SolidSteel144
 
Posts: 420
Joined: Sun Apr 13, 2008 4:34 am
Location: Texas, United States

Next

Return to FAH Hardware

Who is online

Users browsing this forum: No registered users and 1 guest