Advanced search

Message boards : Graphics cards (GPUs) : GPU List

Author Message
UnknownPL1337
Send message
Joined: 22 Mar 15
Posts: 12
Credit: 530,700
RAC: 0
Level
Gly
Scientific publications
watwat
Message 40596 - Posted: 23 Mar 2015 | 23:44:29 UTC

Hello everyone,
i hope this GPU List can help anyone here.

http://gflops.info

[CSF] Thomas H.V. DUPONT
Send message
Joined: 20 Jul 14
Posts: 732
Credit: 126,845,366
RAC: 190,805
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 40597 - Posted: 24 Mar 2015 | 7:23:59 UTC - in response to Message 40596.

Hello everyone,
i hope this GPU List can help anyone here.

http://gflops.info

Thanks UnknownPL1337 :) Interesting list.
Do you plan to update it regularly and to complete it with missing GPU's?
____________
[CSF] Thomas H.V. Dupont
Founder of the team CRUNCHERS SANS FRONTIERES 2.0
www.crunchersansfrontieres

UnknownPL1337
Send message
Joined: 22 Mar 15
Posts: 12
Credit: 530,700
RAC: 0
Level
Gly
Scientific publications
watwat
Message 40600 - Posted: 24 Mar 2015 | 9:33:57 UTC - in response to Message 40597.
Last modified: 24 Mar 2015 | 9:34:51 UTC

Yes i do, there will be over 250 GPUs. Tesla, Quadro more and more.
I want also to thank you guys that i can help you to make the world better :)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 40621 - Posted: 25 Mar 2015 | 22:44:18 UTC
Last modified: 25 Mar 2015 | 22:50:58 UTC

only the Memory interface matters 384/256/192 not the memory clock speed.

It should be:"only the memory bandwidth matters, which is interface width (e.g. 384, 256 or 192 bit) times clock speed".

Together with a disclaimer that if the app is not limited by memory bandwidth on a certain GPU, memory clock speed increases only increase power draw but not performance. In such cases lower memory clocks can lower the power draw. This is generally not the case at GPU-Grid. The app is not particularly bandwidth hungry, but lower memory clocks don't improve efficiency either, as power and performance drop.

Edit: you're talking specifically about Folding@Home. I haven't followed them lately, but I'm sure the GPU architecture also plays a role there. We solved this by including an empirical "flops conversion factor". For example in past times if the theoretical flops from a CC 1.1 GPU (like G94) were x, then a CC 1.3 GPU with the same maximum theoretical flops would crunch as much as if it had 1.4x the raw performance.

That's because different architectures generally differ in the efficiency with which they can use their maximum theoretical flops in real programs. AMDs VLIW5 was notorious for providing amazing theoretical flops with limited real world applicability (outside older games).

MrS
____________
Scanning for our furry friends since Jan 2002

[CSF] Thomas H.V. DUPONT
Send message
Joined: 20 Jul 14
Posts: 732
Credit: 126,845,366
RAC: 190,805
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 40625 - Posted: 26 Mar 2015 | 7:39:48 UTC - in response to Message 40600.

Yes i do, there will be over 250 GPUs. Tesla, Quadro more and more.
I want also to thank you guys that i can help you to make the world better :)

Thanks for your reply :)
____________
[CSF] Thomas H.V. Dupont
Founder of the team CRUNCHERS SANS FRONTIERES 2.0
www.crunchersansfrontieres

UnknownPL1337
Send message
Joined: 22 Mar 15
Posts: 12
Credit: 530,700
RAC: 0
Level
Gly
Scientific publications
watwat
Message 40627 - Posted: 26 Mar 2015 | 18:52:20 UTC - in response to Message 40621.
Last modified: 26 Mar 2015 | 18:54:27 UTC

only the Memory interface matters 384/256/192 not the memory clock speed.

It should be:"only the memory bandwidth matters, which is interface width (e.g. 384, 256 or 192 bit) times clock speed".

Together with a disclaimer that if the app is not limited by memory bandwidth on a certain GPU, memory clock speed increases only increase power draw but not performance. In such cases lower memory clocks can lower the power draw. This is generally not the case at GPU-Grid. The app is not particularly bandwidth hungry, but lower memory clocks don't improve efficiency either, as power and performance drop.

Edit: you're talking specifically about Folding@Home. I haven't followed them lately, but I'm sure the GPU architecture also plays a role there. We solved this by including an empirical "flops conversion factor". For example in past times if the theoretical flops from a CC 1.1 GPU (like G94) were x, then a CC 1.3 GPU with the same maximum theoretical flops would crunch as much as if it had 1.4x the raw performance.

That's because different architectures generally differ in the efficiency with which they can use their maximum theoretical flops in real programs. AMDs VLIW5 was notorious for providing amazing theoretical flops with limited real world applicability (outside older games).

MrS


Could you quote what you would write instead that what i wrote?

example ?
"if the app is not limited by memory bandwidth on a certain GPU, memory clock speed increases only increase power draw but not performance. In such cases lower memory clocks can lower the power draw. This is generally not the case at GPU-Grid. The app is not particularly bandwidth hungry, but lower memory clocks don't improve efficiency either, as power and performance drop."

I bought a GT630 GK208, i will see how much GFlops i can make with it by Modding it :) Wish me luck hehe ^^ I got the Zotac 1GB Version with GK208. I will run this GPU on my PC because my GTX580 is TOOO HUNGRY :( 250Watt+ for 1750Gflops, about 6Gflops/watt, the GT630 G208 has 27Gflops/watt ;)

UnknownPL1337
Send message
Joined: 22 Mar 15
Posts: 12
Credit: 530,700
RAC: 0
Level
Gly
Scientific publications
watwat
Message 40663 - Posted: 28 Mar 2015 | 16:03:32 UTC

I go beast mode :p will buy a 10000+Gflops pc :p Gpugrid is my new hobby haha :D

Adding Radeon R200 series on the list... ;) i should be done today with r200, 7xxx 6xxx :)

I saw the 8xxx series and it looks like it's only for Laptops.

Post to thread

Message boards : Graphics cards (GPUs) : GPU List

//