GPUGRID and ATI
Message boards :
Graphics cards (GPUs) :
GPUGRID and ATI
Message board moderation
Previous · 1 . . . 10 · 11 · 12 · 13 · 14 · 15 · 16 . . . 26 · Next
| Author | Message |
|---|---|
 Paul D. BuckSend message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level  Scientific publications     
|
I think that we understood in a small test code what is the problem and why we did not get the performance. If we can replicate the fix to ACEMD and it works there as well, then there could be some very good news. The best news I have had in, like, forever ... Especially when I am on the verge of having 3 each HD5870 cards ... :) If I recall correctly, GPU Grid will never be able to be run on less ... anyway ... I look forward to more ... |
|
Send message Joined: 18 Aug 08 Posts: 121 Credit: 59,836,411 RAC: 0 Level  Scientific publications  
|
Good news :) If it will work it will be the fourth project for ATI. 1. MilkyWay@home 2. Collatz Conjecture 3. DNETC@HOME 4. and GPUGRID I am keeping my fingers crossed! POLISH NATIONAL TEAM - Join! Crunch! Win! |
|
Send message Joined: 31 Mar 10 Posts: 45 Credit: 103,429,292 RAC: 0 Level  Scientific publications
|
Good news :) Seti & Primegrid not fully optimized ;) |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications
|
Seti & Primegrid not fully optimized ;) SaH is a new application in testing, the last I checked the PG is not even in a beta test, it has only been tested in stand alone modes... even the one CUDA application has been retired because the sub-project is over. As of a couple a days ago there was no GPU application on PG in use ... |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications
|
Well, Mr. Project Guy ... one more Nvidia pulled and replaced with ATI card ... puts my project standing in dire straits sadly ... but ... I have to cut my power and increase my throughput and this is the cheapest way for me to get from here to there ... The one piece of good news is that though I only have 20 hours of testing in ... the GTX295 card I pulled from GPU Grid because Vista was giving me daily "Driver stopped" messages may be working better under Win7-64 (none so far, crosses fingers, raps wood ...) if that holds I will be putting that card back to work here ... so that is one plus out of all this ... |
skgiven Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level  Scientific publications |
A big Plus, i'd say! |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications
|
A big Plus, i'd say! Been running not quite 24 hours ... going to take a couple of days to see a difference in stats I know ... and with the "pulsing" in the numbers it is hard to know where you are till you get at least a weeks data or better a month ... then you can average it out ... today's update had me at 705K for the day, not bad, but less than yesterday's 720, or the day before's 790 ... My pending on Collatz is at an all time high though ... so ... that is good ... making money on DNETC and they seem to be making small changes over the batches so maybe we can close in on the problems they are seeing there ... I have been getting errors and hard to know why, asked for more logging and hours later I have a bunch of results with nice logs ... Not sure if it is going to help with the "hang" problem though, if the task hangs and gets aborted if they just opened the log once it might be still open when the kill comes in and the log data will be lost ... if they do an open, append, close for each statement we might get some where ... though I saw a synch line in there that worries me ... it waits for the GPUs to end ... what if the GPU does not know it has ended, or sent the signal already? |
|
Send message Joined: 11 Nov 09 Posts: 23 Credit: 668,841 RAC: 0 Level  Scientific publications
|
ATi OpenCL Stream SDK 2.1 is out now. Maybe it helps you with some bugfixing! |
|
GDF Send message Joined: 14 Mar 07 Posts: 1958 Credit: 629,356 RAC: 0 Level Scientific publications |
Thanks, in this version, there could be the first working application. We are working hard on ATI now and we like the hardware. gdf |
|
Zydor Send message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level  Scientific publications
|
Which of the ATI cards are being used for development ? Regards Zy |
|
GDF Send message Joined: 14 Mar 07 Posts: 1958 Credit: 629,356 RAC: 0 Level Scientific publications |
One ATI HD5850 and a HD5670. The current speed of a HD5850 on ACEMD is twice as slow as a GTX275 (one month ago was ten times). There is still scope for improvement. The code is not stable yet, maybe with 2.1. gdf |
|
Zydor Send message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level Scientific publications
|
Okie Doke - Many Thanks Regards Zy |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications
|
One ATI HD5850 and a HD5670. Ugh, I would expect based on other projects that those numbers would be more like reversed ... :( Still, as you note, still room for improvement... Looking forward to being able to try it ... I seem to be down to only one card on GPU Grid in that I put the GTX295 card back into rotation but all the tasks seem to be failing ... :( So, down to only one GTX280 card ... which on DNETC, Collatz and MW is considerably slower than the HD5870s which are my current backbone... |
|
Send message Joined: 5 May 10 Posts: 1 Credit: 0 RAC: 0 Level  Scientific publications 
|
Hi I have a Win 7 64bit computer with Radeon 5870. I would like to help if it is needed. |
 Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level  Scientific publications |
Ugh, I would expect based on other projects that those numbers would be more like reversed ... :( Just don't forget that ATIs 5-wide execution units give you more raw compute horse power per transistor, but are also harder to use efficiently. So you can't expect ATI to be faster for every algorithm.. otherwise nVidias design would be really stupid ;) MrS Scanning for our furry friends since Jan 2002 |
|
skgiven Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications |
... :) That is funny. No, hilarious. |
|
GDF Send message Joined: 14 Mar 07 Posts: 1958 Credit: 629,356 RAC: 0 Level Scientific publications |
The SDK2.1 is better but still has several bugs in parts which are required to run the code reliably. We must wait for the next release. In the meanwhile we are working on the performance with the help of AMD. gdf |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications
|
The SDK2.1 is better but still has several bugs in parts which are required to run the code reliably. We must wait for the next release. Thats is good news once again ... and if they are helping, maybe they will see first hand the problems with the SDK and that will improve the chances that that will be fixed as well ... Thanks for the update ... |
Beyond Send message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level  Scientific publications
|
In the meanwhile we are working on the performance with the help of AMD. Very good news. Thanks! |
|
Send message Joined: 9 May 10 Posts: 6 Credit: 0 RAC: 0 Level Scientific publications
|
The SDK2.1 is better but still has several bugs in parts which are required to run the code reliably. We must wait for the next release. New here with a 5870 open for testing if required. Awesome to read of the work going on here by everyone involved, and the help from AMD! |
©2025 Universitat Pompeu Fabra