Message boards :
Number crunching :
some basics in amd gpu computing
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
I think I'll just run cpu tasks and keep it simple. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Hmmmm...more errors. BTW, if you just ran the Lunatics Installer and installed only the CPU App, it should have deleted your ap_cmdline...ati file. If you run the Lunatics installer again and add the ATI AP App it will install a fresh Blank ap_cmdline...ati copy. Then the card would run using the defaults, not at -unroll 18. Every time you run the installer it should install Blank cmdline files for the Apps you install. |
Admiral Gloval Send message Joined: 31 Mar 13 Posts: 20326 Credit: 5,308,449 RAC: 0 |
My machine sounds dirt slow. It takes 30 hours approximatly to do a AP. It is a ATI Radeon 3000 series gpu. Now how do you do a wu on a smartphoneâ‰ï¸ |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
My machine sounds dirt slow. It takes 30 hours approximatly to do a AP. It is a ATI Radeon 3000 series gpu. Now how do you do a wu on a smartphoneâ‰ï¸ https://play.google.com/store/apps/details?id=edu.berkeley.boinc&hl=en SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Admiral Gloval Send message Joined: 31 Mar 13 Posts: 20326 Credit: 5,308,449 RAC: 0 |
My machine sounds dirt slow. It takes 30 hours approximatly to do a AP. It is a ATI Radeon 3000 series gpu. Now how do you do a wu on a smartphoneâ‰ï¸ F F F FIDDLESTICKS!!! I HAVE A IOS phone. |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
Thanks TBar, I'll check it out. Hmmmm...more errors. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
My machine sounds dirt slow. It takes 30 hours approximatly to do a AP. It is a ATI Radeon 3000 series gpu. Now how do you do a wu on a smartphoneâ‰ï¸ Well if you have an Android device of any kind you could give it a ago. However I would imagine if you have an iPhone you would also have an iPad of some sort if you have a tablet. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
So my Merle-PC is no more. I put it out of it's misery and put two slugs in it and threw it in the dumpster. Bye-bye, tough guy. See you in computer heaven. |
Admiral Gloval Send message Joined: 31 Mar 13 Posts: 20326 Credit: 5,308,449 RAC: 0 |
Well if you have an Android device of any kind you could give it a ago. However I would imagine if you have an iPhone you would also have an iPad of some sort if you have a tablet. I wish...but there are two in the house. Just not mine. The App store has a SETI educational download but no crunching at all. GRRRR... My iPhone is a pad of sorts. It will do everything a iPad will do but on a small small screen. Someone needs to write a SETI crunching app for iOS. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Someone needs to write a SETI crunching app for iOS. They are still Beta testing the Apps for Mac desktop and laptops GPUs. Once they get those done, maybe but one thing at a time. ;) |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Someone needs to write a SETI crunching app for iOS. Someone would still need to create & publish a version of BOINC for iOS. Would Apple would be willing to fund that development like HTC did for the Android app? SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
TBar, I set the unroll to 8 and it took about 7 hours to complete. I also have a TdrDelay of 8 set in my registry which I was told long ago that it would help. My normal screen graphics are a little slow so I am snoozing my boinc when I do other tasks. My cpu takes about 10-11 hours doing 4 or 5 at a time. Thanks for your help. The last 3 completions on that card show an -unroll setting of 18. It should run best around unroll 6 to 8. You might want to check it out. Not long ago an out of memory warning was fatal. The task would immediately error out with 'out_of_resources'. Nice to see it now allows the task a chance to complete. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
TBar, That's about the same as my 4670 running the latest and greatest, http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=17649973 You might get better screen graphics with an unroll of 6 or even 4. I think my best times in XP were a little over 6 hours with unroll 4. I have it higher currently 'cause the other cards need it higher. The ATI HD 4600 series will work with AstroPulse v7 however, I haven't seen a CAL version of AstroPulse v7. This might be the last hurrah for the old CAL only cards, ie. HD3 & HD2. |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
My card is actually a 4670. I misspoke earlier. I'll try unroll 5 and see how it is. I guess all I have to do is restart boinc to get it to work. --edit Yeah, 5 works much better. Thanks again. |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
I guess all I have to do is restart boinc to get it to work. Just 'Snooze GPU' (after 5-10 s un-Snooze it) Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
thanks BilBg |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
Joe, But how does one know what the real speed vs theoretical peak gflops is? Related question: Sometimes for a lengthy period my r7 265 and 7770 have an average combined usage of perhaps 50-60 percent. Is that because of the AP blanking and it will be fixed by v7 at least partially, or is it something else? [somehow I missed your posting on the first go round.] lol |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Joe, What really matters is the relative speed on AP or MB tasks, so basically it's a matter of observing live run times or running comparative tests offline. For your host with an HD 7770 and an R7 265, I estimate the 7770 is taking about 2.5 times as long as the R7 for AP work. That's largely based on Task 3751359063 for the 7770 and Task 3751363773 for the R7, but times for other tasks seem to follow the same pattern of slightly over 4000 seconds for the 7770 and well under 2000 seconds for the R7 (tasks with little or no blanking). I won't guess about MB, too little data. Related question: That certainly could be heavy blanking effects, one way to get a rough idea about blanking on a running task is to look at the size of the indices.txt file in the slot for the task. A zero byte indices.txt corresponds to no blanking, anything under 3K bytes or so probably low enough blanking to not have major effects. OTOH, some users have noted low GPU usage for other reasons and in some cases I think the reason hasn't ever been found. Joe |
merle van osdol Send message Joined: 23 Oct 02 Posts: 809 Credit: 1,980,117 RAC: 0 |
Thank's Joe, I thought that the r7 was much faster compared to the 7770 than the gflops were telling me. That confirms it. Now if the 6570 were 2.5 times slower than the 7770, it would hardly be worth it to add it to the mix. But I don't think it would be 20x slower than the r7. OTOH, some users have noted low GPU usage for other reasons and in some cases I think the reason hasn't ever been found. But half the time both gpu's work on average maybe 85% or more of capacity which says to me they are healthy working together. |
©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.