Author |
Message |
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
I have just discovered this page thanks to a user which came with the latest server update and we did not know.
http://www.gpugrid.net/gpu_list.php
It reports the average performance of each gpu. Maybe you know it already.
gdf |
|
|
nateSend message
Joined: 6 Jun 11 Posts: 124 Credit: 2,928,865 RAC: 0 Level
Scientific publications
|
Doesn't seem to work... no way a 580 is (0.18) relative to a 590. |
|
|
|
It depends on the quality of user data the server is provided with. If few people here run GTX580 (most probably use GTX570 for cost-effectiveness or GTX590 for maximum bang), then an odd result from some user will skew the average. That's why I wouldn't trust the Linux numbers either ;)
MrS
____________
Scanning for our furry friends since Jan 2002 |
|
|
TarHealSend message
Joined: 24 Sep 11 Posts: 9 Credit: 10,103,862 RAC: 0 Level
Scientific publications
|
Could the odd performance of the 580 relative to the 590 and other cards have anything to do with the frequency with which GTX 580 users on the Grid overclock their cards, compared to users of the GTX 590? And is one more overclockable than the other? Or, could 590 users have a higher error rate percentage, because they are more prone to pushing their hardware's temps higher? Thus, maybe more GTX 590 processing time is "wasted" on WU's that ultimately error out?
Could we get another version of this chart with updated performance information and an indication of how many work units are referenced in determining each card's score? That would provide us with a confidence level for comparing the data from each card. Also, it would be good to filter out the cards that are no longer supported by ACEMD applications and add in numbers for the new Nvidia 600 series as they become available.
I realize that the GTX 580 score is probably off quite a bit, but just based on the chart as it is now, I note that you can pay $420 to get a GTX 580, which is 3.15X what you might pay for a GTX 550 Ti ($130), but you only get 2.95X the performance. Of course, that's not counting the cost of electricity required to run each board and its host PC 24/7 for a year. But it's something to think about; the "fastest" GPU might not be the best GPU for GPUGrid, and the cheapest GPU is often not the most cost effective, when you factor in energy efficiency and other variables. |
|
|
skgivenVolunteer moderator Volunteer tester
Send message
Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level
Scientific publications
|
Boinc perceives the GTX 590 as having one GPU. It has two. So the stats are per GPU. Basically double the GTX 590's performance to get a comparison.
It's worth noting that none of these stats consider Boinc or operating system settings. Some people will be on XP, some on W7, and the difference is at least 11%, but often 30%. The settings are also relevant; if you have configured Boinc to not use the GPU when the system is active, it will make a difference. If you don't use fan controlling software then some GPU's are going to downclock. How many CPU cores you free, and if you use the system for gaming or watching video is also going to influence these stats.
So, don't read too much into these stats.
Until we know exactly how well the GTX 680 performs, the top GPU's for here are,
Geforce GTX 590, 580, 570, 480, 560 Ti (448), 470, and 465.
Relative Comparison of Recommended Cards, with approximated CC Correction Factor values (in brackets):
GTX 590 GF110 40nm Compute Capable 2.0 2488 GFlops peak (3359)
GTX 580 GF110 40nm Compute Capable 2.0 1581 GFlops peak (2134)
GTX 570 GF110 40nm Compute Capable 2.0 1405 GFlops peak (1896)
GTX 560 Ti 448 GF110 40nm Compute Capable 2.0 1311 GFlops peak (1770)
GTX 480 GF100 40nm Compute Capable 2.0 1345 GFlops peak (1816)
GTX 295 GT200b 55nm Compute Capable 1.3 1192 GFlops peak (1669)
GTX 470 GF100 40nm Compute Capable 2.0 1089 GFlops peak (1470)
GTX 465 GF100 40nm Compute Capable 2.0 855 GFlops peak (1154)
GTX 560 Ti GF114 40nm Compute Capable 2.1 1263 GFlops peak (1136)
GTX 285 GT200b 55nm Compute Capable 1.3 695 GFlops peak (973)
GTX 560 GF114 40nm Compute Capable 2.1 1075 GFlops peak (967)
GTX 275 GT200b 55nm Compute Capable 1.3 674 GFlops peak (934)
GTX 260-216 GT200b 55nm Compute Capable 1.3 596 GFlops peak (834)
GTX 460 GF104 40nm Compute Capable 2.1 907 GFlops peak 768MB (816)
GTX 460 GF104 40nm Compute Capable 2.1 907 GFlops peak 1GB (816)
GTX 550 Ti GF116 40nm Compute Capable 2.1 691 GFlops peak (622)
GTS 450 GF106 40nm Compute Capable 2.1 601 GFlops peak (541)
GT 545 GF116 40nm Compute Capable 2.1 501 GFlops peak (451)
____________
FAQ's
HOW TO:
- Opt out of Beta Tests
- Ask for Help |
|
|
wiyosayaSend message
Joined: 22 Nov 09 Posts: 114 Credit: 589,114,683 RAC: 0 Level
Scientific publications
|
Thanks for this post. Since I planning a new computer build for sometime in the next three months or so, I was looking for this kind of information.
Thanks again.
____________
|
|
|
TarHealSend message
Joined: 24 Sep 11 Posts: 9 Credit: 10,103,862 RAC: 0 Level
Scientific publications
|
Thanks for the clarification re: GTX 590. And I take your point with regard to all the other variables which may cause each card to perform below its theoretical peak level.
I didn't know that choice of OS could result in a 11-30% difference in performance. That's pretty huge. I browsed through the FAQ pages and didn't see this addressed, so I'll ask: Is there a "best" operating system for running GPU Grid? If there is a consensus that choosing Windows 7 or Ubuntu 11 or whatever makes a significant impact on performance (and I definitely consider 11% to be 'significant'), then I'll gladly make the change on my machines. |
|
|
skgivenVolunteer moderator Volunteer tester
Send message
Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level
Scientific publications
|
Linux (x64 only) is generally considered the fastest, but XP (x64/x86) comes close, as do similar win servers; 2003/2008. W7/Vista are to be avoided if possible. W8 doesn't work properly as yet.
64bit OS's are generally faster for CPU tasks (other projects), but this has little importance for here. However, only Linux x64 can be used here.
____________
FAQ's
HOW TO:
- Opt out of Beta Tests
- Ask for Help |
|
|
5potSend message
Joined: 8 Mar 12 Posts: 411 Credit: 2,083,882,218 RAC: 0 Level
Scientific publications
|
Example: My W7 machine NEVER goes above 82% utilization of my 570. I haven't run Linux since I've bought my 570 on that rig, so I don't have a comparison, but if linux runs 99%+ than that's 17% loss from W7 on my rig.
On a normal task that's about a 30 min loss of productivity. |
|
|