Message boards :
Number crunching :
How do I get this type of point production?
Message board moderation
Author | Message |
---|---|
Send message Joined: 20 Nov 07 Posts: 54 Credit: 2,663,789 RAC: 0 |
Under 3 seconds per unit? http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=51931 |
Send message Joined: 13 Feb 08 Posts: 1124 Credit: 46,740 RAC: 0 |
Under 3 seconds per unit? With an app running on a big GPU card. |
Send message Joined: 24 Dec 07 Posts: 1947 Credit: 240,884,648 RAC: 0 |
But that's the cpu time. Wall clock time was over 200 seconds, but that was with 3 tasks running on the gpu. |
Send message Joined: 20 Nov 07 Posts: 54 Credit: 2,663,789 RAC: 0 |
Wow, after hearing this it kinda takes the wind out of the sail to continue using cpu's. |
Send message Joined: 20 Sep 08 Posts: 1391 Credit: 203,563,566 RAC: 0 |
It would not surprise me if this project goes out of business very shortly due to the use of GPU crunching. |
Send message Joined: 3 Jan 09 Posts: 139 Credit: 50,066,562 RAC: 0 |
I dont know about going out of business,the milkyway is a big place after all. Seems to me theres plenty of crunching can go on by both CPU and GPU for a while at least. I will upgrade my old P4 soon,looking at quad core and a GPU or will buy new computer.which ever is cheaper. |
Send message Joined: 20 Nov 07 Posts: 54 Credit: 2,663,789 RAC: 0 |
Great, just the excuse I've been waiting for to retire most boxen, save on utility costs, and upgrade a few vid cards. |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Wow, after hearing this it kinda takes the wind out of the sail to continue using cpu's. Not all projects can or will use the GPU. Some problems / algorithms are not amenable to being run on a GPU. For those applications where the GPU can be used, the speed-up is significant. The good news is that by moving some tasks onto the GPU we start using a resource that had been idle in the past but we also release the CPU to be used on other tasks and projects. Yes I can get the same number of Cobblestones in a shorter time by running tasks on the one ATI GPU I have as compared to all my other systems combined running tasks on their CPUs. In my case, however, I did not stop running MW on these other systems because the point is to do the work ... fast or slow, the point is to do the work. |
Send message Joined: 1 Dec 08 Posts: 139 Credit: 8,721,208 RAC: 0 |
Chris S wrote: It would not surprise me if this project goes out of business very shortly due to the use of GPU crunching. Why would having a GPU app cause the project to "go out of business"? |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Chris S wrote:It would not surprise me if this project goes out of business very shortly due to the use of GPU crunching. Completion of the research ... That assumes of course that there are not more things to look into and that there is a finite number of slices to be made and tested. In that Travis has indicated that there are different methods of madness to the searches it seems to me that there is much yet to be done ... On the other hand, this is the point of doing the work ... to complete the project ... to think of an idea, write the application, run the tasks and complete the research ... I grant you that I would like to keep good project around forever, the point is to complete the research ... :) |
Send message Joined: 1 Dec 08 Posts: 139 Credit: 8,721,208 RAC: 0 |
Paul D. Buck wrote: Completion of the research ... D'oh! |
Send message Joined: 26 Jan 09 Posts: 589 Credit: 497,834,261 RAC: 0 |
|
Send message Joined: 18 Nov 07 Posts: 280 Credit: 2,442,757 RAC: 0 |
Isn't the eventual goal to get 'fitness' to 0.0? ;) With the strategies currently being employed it won't matter how much processing power you throw at the problem, we're still 'only' getting to 2.85-3.00. Remember, just adding another parameter to the mix to improve the physical accuracy of the results adds a whole new dimension to the search space - I don't see us running out of slices anytime soon, but even if we did the work we can do is only bounded by the complexity of the programs. |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Isn't the eventual goal to get 'fitness' to 0.0? ;) With the strategies currently being employed it won't matter how much processing power you throw at the problem, we're still 'only' getting to 2.85-3.00. Remember, just adding another parameter to the mix to improve the physical accuracy of the results adds a whole new dimension to the search space - I don't see us running out of slices anytime soon, but even if we did the work we can do is only bounded by the complexity of the programs. I imagine over time the searches will be duplicated with more and more accurate results. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 1 Dec 08 Posts: 139 Credit: 8,721,208 RAC: 0 |
Yeah. Just like the climate researchers have decided they don't need any more CPDN WUs crunched... Yeah, and CPDN WU's are so small to start with... |
©2024 Astroinformatics Group