CUDA App einsteinbinary 3.10 for Windows available for Beta Test

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2955519863
RAC: 717692

RE: Using CUDA device #0

Message 94379 in response to message 94377

Quote:

Using CUDA device #0 "GeForce GTX 260" (806.98 GFLOPS)

WT* is that. no offense, but I'm very hesitate to believe those numbers...

holmis's reports 300 gflops as well, is the app mis-reading the gflop number?????????????????????


Reported five tasks on host 1001564 - two validated so far, three pending. Times all very close to 6 hours, no significant change from 3.07cuda.

Card is a 9800GT, stock speed, rated at 60 BOINC GFlops, but reported here as 508.03 NVidia marketing GFlops (as also shown in the GPUGrid FAQ). I've commented before that there's a factor of 8 or 9 between 'marketing' and 'crunching' GFlops - it doesn't matter which you use for card comparison purposes, so long as you always state which units you're using, and don't mix them.

TJ
TJ
Joined: 11 Feb 05
Posts: 178
Credit: 21041858
RAC: 0

RE: Ok, now I´m getting

Message 94380 in response to message 94378

Quote:

Ok,
now I´m getting also work for the two remaining cpu cores.
@tjreuter: The 9800GX2 has two chips (2x8800GTS if I´m correct). It´s SLI on one card.

I like that cuda stuff ;)

hotze

Hi Hotze,

Off course that card has 2 processors, I was thinking at SLI and then they work as one.
If you like cuda you could take a look at GPUGRID, for about 7 hours work you can get 4000-5000 credits.

Greetings from
TJ

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 984
Credit: 25171376
RAC: 43

RE: Is there "pure"

Message 94381 in response to message 94377

Quote:

Is there "pure" CUDA-application is in developer's plans?

Yes. However, some part of the app will still run on the CPU - it'll never be "pure". We aim to reduce the amount of work done on the CPU as much as we can of course.

Quote:


Using CUDA device #0 "GeForce GTX 260" (806.98 GFLOPS)

WT* is that. no offense, but I'm very hesitate to believe those numbers...
what are the specs of that card?????
MHZ?
the damn card better be; Liquid cooled, running at 800 mhz....
please post the message logs of the boinc start-up...
ok, nvm; i'll admit i'm quick to jump to conclusions....
holmis's reports 300 gflops as well, is the app mis-reading the gflop number?????????????????????

No, it's not. You need to know that this value is computed by the device properties and not determined by benchmarks. It's the card's theoretical maximum performance which can be computed like this: number of multiprocessors * number of cores/multiprocessor * clock rate * flop/core/cycle. For your GTX 260 that's 24 * 8 * 1.4 GHz * 3. By the way, is this an overclocked card? Usually they "only" run at 1.3 GHz...

No one claimed that either your card or any given application could realistically achieve that performance ;-) As always, you may relatively compare different cards using the same measure, despite the absolute values not being very informative.

Cheers,
Oliver

Einstein@Home Project

rbpeake
rbpeake
Joined: 18 Jan 05
Posts: 266
Credit: 1129357797
RAC: 725971

RE: RE: Is there "pure"

Message 94382 in response to message 94381

Quote:
Quote:

Is there "pure" CUDA-application is in developer's plans?

Yes. However, some part of the app will still run on the CPU - it'll never be "pure". We aim to reduce the amount of work done on the CPU as much as we can of course.

Cheers,
Oliver


And we imagine that the GPU will continue to improve in the speed of processing as well, which of course will yield even more efficiency! :)

zpm
zpm
Joined: 1 Apr 06
Posts: 19
Credit: 604277
RAC: 0

RE: RE: Is there "pure"

Message 94383 in response to message 94381

Quote:
Quote:

Is there "pure" CUDA-application is in developer's plans?

Yes. However, some part of the app will still run on the CPU - it'll never be "pure". We aim to reduce the amount of work done on the CPU as much as we can of course.

Quote:


Using CUDA device #0 "GeForce GTX 260" (806.98 GFLOPS)

WT* is that. no offense, but I'm very hesitate to believe those numbers...
what are the specs of that card?????
MHZ?
the damn card better be; Liquid cooled, running at 800 mhz....
please post the message logs of the boinc start-up...
ok, nvm; i'll admit i'm quick to jump to conclusions....
holmis's reports 300 gflops as well, is the app mis-reading the gflop number?????????????????????

No, it's not. You need to know that this value is computed by the device properties and not determined by benchmarks. It's the card's theoretical maximum performance which can be computed like this: number of multiprocessors * number of cores/multiprocessor * clock rate * flop/core/cycle. For your GTX 260 that's 24 * 8 * 1.4 GHz * 3. By the way, is this an overclocked card? Usually they "only" run at 1.3 GHz...

No one claimed that either your card or any given application could realistically achieve that performance ;-) As always, you may relatively compare different cards using the same measure, despite the absolute values not being very informative.

Cheers,
Oliver

thanks, sometimes i go nutz...

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6588
Credit: 315666989
RAC: 329359

RE: Card is a 9800GT, stock

Message 94384 in response to message 94379

Quote:
Card is a 9800GT, stock speed, rated at 60 BOINC GFlops, but reported here as 508.03 NVidia marketing GFlops (as also shown in the GPUGrid FAQ). I've commented before that there's a factor of 8 or 9 between 'marketing' and 'crunching' GFlops - it doesn't matter which you use for card comparison purposes, so long as you always state which units you're using, and don't mix them.


Ah, so that's why. So 1 BGFlop ~= 8 NVMGFlops

[aside]
Reminiscent of :

- the 1 MB vs 1 million bytes issue ( and GB .... )
- the Cyrix processor's 'Performance Rated' speed ( 'as good as a Pentium' vs actual )
- the quoting of CD drive speeds ( angular or linear velocity? )
[/aside]

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

zpm
zpm
Joined: 1 Apr 06
Posts: 19
Credit: 604277
RAC: 0

new driver available, 190.62,

Message 94385 in response to message 94384

new driver available, 190.62, which is working fine on my vista machine.

zpm
zpm
Joined: 1 Apr 06
Posts: 19
Credit: 604277
RAC: 0

a new wu run under 190.62.

Message 94386 in response to message 94385

a new wu run under 190.62. works fine.

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 437
Credit: 139002861
RAC: 0

RE: RE: Card is a 9800GT,

Message 94387 in response to message 94384

Quote:
Quote:
Card is a 9800GT, stock speed, rated at 60 BOINC GFlops, but reported here as 508.03 NVidia marketing GFlops (as also shown in the GPUGrid FAQ). I've commented before that there's a factor of 8 or 9 between 'marketing' and 'crunching' GFlops - it doesn't matter which you use for card comparison purposes, so long as you always state which units you're using, and don't mix them.

Ah, so that's why. So 1 BGFlop ~= 8 NVMGFlops

[aside]
Reminiscent of :

- the 1 MB vs 1 million bytes issue ( and GB .... )
- the Cyrix processor's 'Performance Rated' speed ( 'as good as a Pentium' vs actual )
- the quoting of CD drive speeds ( angular or linear velocity? )
[/aside]

Cheers, Mike.

BOINC is telling us a much lower figure. This is funny because when you put a ATI card in with it their cards are reported as being much faster (by BOINC) under 6.10.0. While there is differenced between the cards the formula needs to be standardised, at least in the BOINC world.

zpm
zpm
Joined: 1 Apr 06
Posts: 19
Credit: 604277
RAC: 0

my last cuda wu finished in

Message 94388 in response to message 94387

my last cuda wu finished in 5-6 hrs rt.

http://einsteinathome.org/task/137117612

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.