bit-tech.net

thehippoz · 7th Jan 2011, 01:13

I score around 18k/s in pyrit using cuda and a gtx 460.. but my friend scores 60k/s using a 5850 with the same benchmark

to put in it perspective.. 4 gtx295 score 80k in the same test- cuda is lies!!

coolamasta · 7th Jan 2011, 08:30

Eh?! Is this what you are rambling about? - http://code.google.com/p/pyrit/

How can your mate be getting 60k when a more powerful 5870 gets 33k!! Must be hell of an overclock eh!

Are you slightly Anti-Nvidia by any chance? lol But thanks for sharing anyway

thehippoz · 7th Jan 2011, 22:02

I dunno but that's what he's getting.. even that chart is probably skewed

don't you find it odd in that even in that chart, one 5870 running opencl is doing what 4 nvidia gpu's is doing.. I think ati cards are held back by the massive hype nvidia has built..

folding paid off?

n0pz · 7th Jan 2011, 22:24

Using Cal++ can get higher score. Top one was bruteforce using crunch to feed to pyrit to cowpatty. Bottom one is benchmark. Only thing overclocked is cpu, no gpu overclock; just a regular ATI HD 5850. And no, it's not photoshoped =]

saspro · 7th Jan 2011, 23:25

I'm sure it's down to pipeline efficiency & drivers for folding

I may be wrong, I often am

StreetSam · 8th Jan 2011, 03:23

Quote:

Originally Posted by thehippoz

don't you find it odd in that even in that chart, one 5870 running opencl is doing what 4 nvidia gpu's is doing..

I promised myself that I wouldn't reply to this post, but here I am, I must be drunk. Again.

Anyway the cheddar that I was using didn't seem to mark the blackboard at all, I'm sure that blackboard manufacturers are deliberately skewing their boards so that only chalk can write on them.

Quote:

I think ati cards are held back by the massive hype nvidia has built..

Yea that hype really does affect their computational power, it must be the way it affects the planets magnetic field.

Quote:

folding paid off?

A conspiracy theory, brilliant, your post has everything now.

How I wish this was true, an institute of learning doing ground breaking medical research has the forethought to extort money from a multi-national electronics company along the way, they should make a film about it.

I apologise to all those people that just wanted this thread to go away, sorry.

n0pz · 8th Jan 2011, 17:46

Which is why I stopped folding since have bad ati support. Look at pyrit, when you have an ati card doing better than 3 nvidia sli cuda cards, there something fishy when it comes to folding. Like they put loops in for the ati guy and all nvidia ppds are multipled by 4 =]

Tangster · 8th Jan 2011, 18:05

I thought it was down to ATI's double precision being badly implemented for folding. It's single precision power does blow everything else out of the water though.

Ph4ZeD · 8th Jan 2011, 18:29

I thought it was purely a question of the client.

7th Jan 2011, 01:13	#1
thehippoz Banned Join Date: Dec 2008 Location: Fresno, CA Posts: 5,780	Why isn't ati better? I score around 18k/s in pyrit using cuda and a gtx 460.. but my friend scores 60k/s using a 5850 with the same benchmark to put in it perspective.. 4 gtx295 score 80k in the same test- cuda is lies!!

7th Jan 2011, 08:30	#2
coolamasta Folding@Home CC Captain 2010/11/12 Join Date: Apr 2009 Location: Sunny Brizzol! Posts: 2,338	Eh?! Is this what you are rambling about? - http://code.google.com/p/pyrit/ How can your mate be getting 60k when a more powerful 5870 gets 33k!! Must be hell of an overclock eh! Are you slightly Anti-Nvidia by any chance? lol But thanks for sharing anyway __________________ My Main Rigs:~ H2O'd i7 3770K @ 4.5 - 16GB - H2O'd GTX 670 \| Asus Z9PE-W8, EP-Xeon 2670 @ 3.2 - 16GB \| ~~~ My Current Dual EP-Xeon 2670 Project ~~~ \| ~~~ My Mitsubishi Evo Track Car Project ~~~ \|

7th Jan 2011, 22:02	#3
thehippoz Banned Join Date: Dec 2008 Location: Fresno, CA Posts: 5,780	I dunno but that's what he's getting.. even that chart is probably skewed don't you find it odd in that even in that chart, one 5870 running opencl is doing what 4 nvidia gpu's is doing.. I think ati cards are held back by the massive hype nvidia has built.. folding paid off? Last edited by thehippoz; 7th Jan 2011 at 22:10.

7th Jan 2011, 22:24	#4
n0pz What's a Dremel? Join Date: Jan 2011 Location: California Posts: 2	The old bench probably using OpenCL Using Cal++ can get higher score. Top one was bruteforce using crunch to feed to pyrit to cowpatty. Bottom one is benchmark. Only thing overclocked is cpu, no gpu overclock; just a regular ATI HD 5850. And no, it's not photoshoped =] Last edited by n0pz; 8th Jan 2011 at 06:47.

7th Jan 2011, 23:25	#5
saspro Infrastructure Specialist Join Date: Apr 2009 Location: Kingston upon Thames Posts: 8,497	I'm sure it's down to pipeline efficiency & drivers for folding I may be wrong, I often am __________________ If what I've said helps you please remember to hit the rep button on the left Missing the old meeting place? Find us here

bit-tech.net

Site Links

Our Other Sites

8th Jan 2011, 17:46	#7
n0pz What's a Dremel? Join Date: Jan 2011 Location: California Posts: 2	Which is why I stopped folding since have bad ati support. Look at pyrit, when you have an ati card doing better than 3 nvidia sli cuda cards, there something fishy when it comes to folding. Like they put loops in for the ati guy and all nvidia ppds are multipled by 4 =]

8th Jan 2011, 18:05	#8
Tangster Butt-kicking for goodness! Join Date: May 2009 Location: York, UK Posts: 2,626	I thought it was down to ATI's double precision being badly implemented for folding. It's single precision power does blow everything else out of the water though. __________________ Uni Box:i5 2500K@5GHz\|EK Supreme LT+Coolstream XT 240\|16GB HyperX DDR3@1866MHz(in blue)\|EVGA GTX670SC 4GB+EK FC670\|MSI P67A-GD55\|\|M4 128GB\|BeQuiet Dark Power Pro 750W\|U3011\|Raven RV-02E\|KBT Pure+Sidewinder X5 + Intuos 4M Audio: PC->TosLink->TEAC UD-H01->Schiit Asgard-> Denon 5000\|PFE232\|Truth B2030A Monitors BT Feedback

8th Jan 2011, 18:29	#9
Ph4ZeD Banned Join Date: Jul 2009 Posts: 3,806	I thought it was purely a question of the client.