F cuda.............

Message boards : Number crunching : F cuda.............
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile Loony
Avatar

Send message
Joined: 8 Dec 99
Posts: 5
Credit: 3,193,475
RAC: 78
United Kingdom
Message 848986 - Posted: 4 Jan 2009, 0:06:47 UTC

Hmm....... Tried it........!

A 3 hour average workunit dropped to 17 minutes.......

BUT unfortunately... it kept crashing my Nvidia 8400..... even WITH the latest drivers....


Rolled back to a NON cuda version......

Remove the bugs... and I'll try again..!! ;)
ID: 848986 · Report as offensive
Profile RandyC
Avatar

Send message
Joined: 20 Oct 99
Posts: 714
Credit: 1,704,345
RAC: 0
United States
Message 849005 - Posted: 4 Jan 2009, 1:00:11 UTC - in response to Message 848986.  

Hmm....... Tried it........!

A 3 hour average workunit dropped to 17 minutes.......

BUT unfortunately... it kept crashing my Nvidia 8400..... even WITH the latest drivers....


Rolled back to a NON cuda version......

Remove the bugs... and I'll try again..!! ;)


I'll second that, but with a 9400 GT.
ID: 849005 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 849389 - Posted: 4 Jan 2009, 19:52:55 UTC - in response to Message 849005.  

Try..178.24 ;)
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 849389 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51478
Credit: 1,018,363,574
RAC: 1,004
United States
Message 849476 - Posted: 5 Jan 2009, 1:01:38 UTC
Last modified: 5 Jan 2009, 1:01:51 UTC

Well.....I am done with my Cuda bashing....really.

But the Nehi just got banged by one for zero credit.....WU 390817938.....sigh.
"Time is simply the mechanism that keeps everything from happening all at once."

ID: 849476 · Report as offensive
Profile Euan Holton
Avatar

Send message
Joined: 4 Sep 99
Posts: 65
Credit: 17,441,343
RAC: 0
United Kingdom
Message 849487 - Posted: 5 Jan 2009, 1:27:47 UTC

When you look at some of the results coming from the CUDA apps, you have to wonder if they're even remotely doing the same thing as the CPU apps.

At least that WU is getting sent out again.
ID: 849487 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51478
Credit: 1,018,363,574
RAC: 1,004
United States
Message 849489 - Posted: 5 Jan 2009, 1:32:43 UTC - in response to Message 849487.  

When you look at some of the results coming from the CUDA apps, you have to wonder if they're even remotely doing the same thing as the CPU apps.

At least that WU is getting sent out again.


My results were.....

Credit multiplier is : 2.85
WU true angle range is : 0.007167
Restarted at 6.70 percent.

Flopcounter: 19390979096301.250000

Spike count: 1
Pulse count: 3
Triplet count: 3
Gaussian count: 0

The Cuda came up with.....

SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

Flopcounter: 331863280.000000

Spike count: 30
Pulse count: 0
Triplet count: 0
Gaussian count: 0

I know which one I trust.....but I guess we'll have to see what another wingman comes up with......
"Time is simply the mechanism that keeps everything from happening all at once."

ID: 849489 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 849560 - Posted: 5 Jan 2009, 5:39:37 UTC - in response to Message 849476.  

Well.....I am done with my Cuda bashing....really.

But the Nehi just got banged by one for zero credit.....WU 390817938.....sigh.

Calm the kitties, the third host doesn't have CUDA.
                                                                 Joe
ID: 849560 · Report as offensive
Profile The Gas Giant
Volunteer tester
Avatar

Send message
Joined: 22 Nov 01
Posts: 1904
Credit: 2,646,654
RAC: 0
Australia
Message 849612 - Posted: 5 Jan 2009, 9:55:20 UTC

Just imagine if an optimised app was giving these sort of results. All hell would brought down on the devs of the app.....CUDA should be pulled and sent back to BETA.
ID: 849612 · Report as offensive
MarkJ Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 08
Posts: 1139
Credit: 80,854,192
RAC: 5
Australia
Message 849614 - Posted: 5 Jan 2009, 10:19:42 UTC - in response to Message 849487.  

When you look at some of the results coming from the CUDA apps, you have to wonder if they're even remotely doing the same thing as the CPU apps.

At least that WU is getting sent out again.


You can see why I pulled the pin on Seti CUDA. I wonder how much it has compromised the science?

I know Eric doesn't like to post on the message boards if he can avoid it, but some words of wisdom (or otherwise) would be appreciated. Maybe they'll just invalidate all the CUDA results and rerun them once the CUDA app has been corrected. Here's hoping anyway.
BOINC blog
ID: 849614 · Report as offensive
Alinator
Volunteer tester

Send message
Joined: 19 Apr 05
Posts: 4178
Credit: 4,647,982
RAC: 0
United States
Message 849802 - Posted: 5 Jan 2009, 19:41:33 UTC - in response to Message 849612.  

Just imagine if an optimised app was giving these sort of results. All hell would brought down on the devs of the app.....CUDA should be pulled and sent back to BETA.


LOL...

Agreed!

I've taken to using the direct disapproval approach. I scan all my tasks and abort the ones where I draw CUDA wingman (even if I have runtime already on the task).

When an app can't handle certain AR's or Dash 9's when it shouldn't and then still manages to get validated somehow, things are just not ready for prime time. This is not even taking into account the recommended CC for all users is just barely able to deal with coprocessing capability (and even then requires babysitting many people aren't willing to deal with) and the project default is to enable it if available.

From where I'm sitting it looks like everything they could do wrong, they did do wrong. Let's not forget (as some folks like to point out all the time), the vast majority of BOINC users may never even look at the front page for news, let alone read the boards. So there is just no way to tell how much worse things could get going forward. ;-)

If I wasn't running a long term study of something else of which SAH was a part of, I would just NNT the project and be done with it for now. ;-)

Alinator
ID: 849802 · Report as offensive
Iona
Avatar

Send message
Joined: 12 Jul 07
Posts: 790
Credit: 22,438,118
RAC: 0
United Kingdom
Message 849815 - Posted: 5 Jan 2009, 20:42:49 UTC - in response to Message 849489.  

When you look at some of the results coming from the CUDA apps, you have to wonder if they're even remotely doing the same thing as the CPU apps.

At least that WU is getting sent out again.


My results were.....

Credit multiplier is : 2.85
WU true angle range is : 0.007167
Restarted at 6.70 percent.

Flopcounter: 19390979096301.250000

Spike count: 1
Pulse count: 3
Triplet count: 3
Gaussian count: 0

The Cuda came up with.....

SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

Flopcounter: 331863280.000000

Spike count: 30
Pulse count: 0
Triplet count: 0
Gaussian count: 0

I know which one I trust.....but I guess we'll have to see what another wingman comes up with......




I've seen similar results myself, using a 'new' PC (albeit one thats nowhere near the 'bleeding edge') and when one of my old P4 Prescotts gave a result like that, I tracked it down to a bad OCZ DDR RAM module. There was an adage in engineering, that you measured something twice and cut it, once.




Don't take life too seriously, as you'll never come out of it alive!
ID: 849815 · Report as offensive
Profile Gecko
Volunteer tester
Avatar

Send message
Joined: 17 Nov 99
Posts: 454
Credit: 6,946,910
RAC: 47
United States
Message 849981 - Posted: 6 Jan 2009, 3:59:28 UTC - in response to Message 849612.  
Last modified: 6 Jan 2009, 4:02:07 UTC

Just imagine if an optimised app was giving these sort of results. All hell would brought down on the devs of the app.....CUDA should be pulled and sent back to BETA.


Well, you know that I agree w/ your Beta comment : > )
Volunteer developers have received comments via Berkeley in the past over far lesser bugs/issues than what this app is producing.....

It's a strange feeling to be so excited for the accomplishment...yet even more disappointed that something this flawed is in public domain.

However, trying to "keep it real".....just because it's generating some garbage doesn't mean they have to USE the garbage.
If one wanted to see/draw-out all the issues....a public release will accomplish this much quicker than a smaller test-group. Perhaps that's the intent....burn-up larger-scale, but shorter-term crunching time to air-out as many problems as possible. This would conform to Ned's "omelet" analogy.

I am confident that Eric & gang are keenly aware of the issues & are mitigating any risks to project.
If it was truly catastrophic & damaging beyond their controls....they'd have already pulled-it.

So...while frustrating & inconvenient for "us"...it appears to be within acceptable tolerance levels for the project ATM.

Not saying I like it....but afterall, I have made a New Years resolution to emulate more of Mr. Segur's patience in 2009 ; > )
ID: 849981 · Report as offensive
Vid Vidmar*
Volunteer tester
Avatar

Send message
Joined: 19 Aug 99
Posts: 136
Credit: 1,830,317
RAC: 0
Slovenia
Message 850072 - Posted: 6 Jan 2009, 11:08:36 UTC - in response to Message 849981.  

Yea, big F this and this!
Nuff said.

BR,
ID: 850072 · Report as offensive
Profile John R. @ SETI.USA

Send message
Joined: 3 Jul 99
Posts: 13
Credit: 82,149,277
RAC: 47
United States
Message 850127 - Posted: 6 Jan 2009, 14:32:32 UTC

I am serious considering taking my rigs elsewhere until this CUDA fiasco has been resolved.
I am not happy finding 0 credits for a W/U that took 1,000 secs.
I am REALLY not happy finding 0 credits for a W/U that took 10,000 secs.
GET IT FIXED!!!!!!!!!!!!!!!!
ID: 850127 · Report as offensive
Claes Holmerup

Send message
Joined: 25 Jul 99
Posts: 7
Credit: 9,446,728
RAC: 25
Sweden
Message 850235 - Posted: 6 Jan 2009, 23:11:34 UTC

After several errors with endless loop calculations of those #¤%&¤ CUDA-WU's (as it seems anyway, because only like 1 out of 5 CUDA-WU's even started showing a percentage after a couple of hours, while a normal WU would have been finished twice or more...), sudden extremely slow screen updates, blue screens etc, I deactivated CUDA downloads in the preferences and then deleted all CUDA-WU's already downloaded.
Luckily, my PC had 3 AP-WU's running at the same time as the 1 CUDA WU, so the time hasn't been a complete waste anyway - but CUDA is a big NO-NO if you ask me!

Now, things are back to normal and running smoothly, as usual. :)
ID: 850235 · Report as offensive
Profile David @ TPS

Send message
Joined: 30 Sep 04
Posts: 70
Credit: 11,323,275
RAC: 0
United States
Message 850239 - Posted: 6 Jan 2009, 23:17:22 UTC - in response to Message 850127.  

I am serious considering taking my rigs elsewhere until this CUDA fiasco has been resolved.
I am not happy finding 0 credits for a W/U that took 1,000 secs.
I am REALLY not happy finding 0 credits for a W/U that took 10,000 secs.
GET IT FIXED!!!!!!!!!!!!!!!!



I would have to agree.

I was running 21K + RAC, now more like 19K. I have a bunch of 0 credit WU's but cannot go through an purge all with CUDA wingmen.
ID: 850239 · Report as offensive
Profile Fred J. Verster
Volunteer tester
Avatar

Send message
Joined: 21 Apr 04
Posts: 3252
Credit: 31,903,643
RAC: 0
Netherlands
Message 850292 - Posted: 7 Jan 2009, 0:47:07 UTC - in response to Message 850239.  

Hi , maybe it only works (with the fastest cards like the 88/89(?) and 90XX series off nVIDIA,) by NOT using them for displaying, but only for crunching and use onboard-video or an (cheaper) ATI card, for display.
Most, not too old, MoBo's have multiple PCI-Ex16/8, PCI-Ex1 and PCI, slots.
I'm gonna try it, when I can get me a nVIDIA 9800 or something as fast as that.
In the latest 2 QUAD builds, I used ATI EAH2400 cards, good and cheap ($15 a piece).

ID: 850292 · Report as offensive
cowboy

Send message
Joined: 2 Aug 08
Posts: 51
Credit: 18,580
RAC: 0
United States
Message 850294 - Posted: 7 Jan 2009, 0:48:35 UTC

If you don't want to run a CUDA app, go into your account, click on Seti@home Preferences, edit the preferences to where your computer is, weather Home/Work/School, ect, change the setting of "Use Graphics Processing Unit (GPU) if available" to no, save your changes, open BOINC Manager, click on your SETI project, click update, and you should no longer get CUDA work. Easy way to opt out of running it until its fixed.
ID: 850294 · Report as offensive
Profile Euan Holton
Avatar

Send message
Joined: 4 Sep 99
Posts: 65
Credit: 17,441,343
RAC: 0
United Kingdom
Message 850298 - Posted: 7 Jan 2009, 1:06:25 UTC - in response to Message 850294.  

If you don't want to run a CUDA app, go into your account, click on Seti@home Preferences, edit the preferences to where your computer is, weather Home/Work/School, ect, change the setting of "Use Graphics Processing Unit (GPU) if available" to no, save your changes, open BOINC Manager, click on your SETI project, click update, and you should no longer get CUDA work. Easy way to opt out of running it until its fixed.

The issue isn't with those of us who do not want to use CUDA, as I doubt that anyone in this thread who is expressing frustration with this rollout is using it.

The issue is that there is a concern that the CUDA application is poorly written and is returning bad information, crashing, and requesting ridiculously small amounts of credit, forcing workunits to be unnecessarily distributed to additional hosts for further verification. Or, to put it succinctly, CUDA is harming the project and alienating some previously strong advocates of the project.
ID: 850298 · Report as offensive
Profile dragon1

Send message
Joined: 17 Sep 05
Posts: 33
Credit: 4,438,013
RAC: 0
Canada
Message 850312 - Posted: 7 Jan 2009, 1:36:08 UTC

I've been fairly busy lately and have not checked my system for results in a couple of weeks and to my horror today I found that my system has been churning out nothing but Client ERRORS on the past 110 WU's since roughly Dec29th. What a waste of time! Now having read a few posts around the board it seems CUDA may be to blame. I wasn't aware that my preferences were going to default to a "YES" for GPU if available. (I have an NVIDIA 8300 with 512GB). Yesterday there were over 40WU's in my queue... today there are NONE and the system says I've used my quota of "2" for the day. I've gone into my preferences and turned off the GPU item and updated and run benchmark. Now what!
ID: 850312 · Report as offensive
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : F cuda.............


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.