Message boards :
Number crunching :
Running Stock Experiment
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
RAC is now 2378 and I have 15 Op app pnedings left on this host When my pendings are gone I will let this thing settle down and see if the rack stabelizes with in expectations. And then will go to the four cores. And give that a fair test on MB. I did want to get my GPU involved but Im going to waite on that. I will run AP stock only on 4 cores. And see what happens. Then after a fair test will go to HT AP only stock. I know that maybe this testing seems stupid. But I need to find out for myself how this works. [/quote] Old James |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
Starting to see some daylight on pendings on this host 10 pendings left, Though one is an AP that needs a manual fix. And a RAC of 2,416. One does need to be patient though. [/quote] Old James |
Gene Send message Joined: 26 Apr 99 Posts: 150 Credit: 48,393,279 RAC: 118 |
>James My test results in a similar MB vs. AP comparison: I was doing a GPU-only test, first letting a single MB task use the GPU (Linux x41g cuda) and getting enough results for a decent average; then allowing a single AP task (Linux AP 6.07) use the GPU. The "executive summary" is that the AP tasks produce 7.5 times the credit as the MB tasks. The GTX 650 GPU is what I would consider a low-end GPU and here's a bit more detail on the MB & AP tasks. MB average run time of 3610 seconds, with CPU utilization of 6.7% and GPU around 45%. AP average run time of 4730 seconds, with CPU utilization of 12.6% and GPU around 80%. (The libsleep.so function IS implemented for the AP application.) This GPU comparison won't necessarily relate directly to your CPU testing but I've posted it here just as another reference datum. Back on April 9, (in this thread) TBar expressed an interest in comparing AP and MB tasks in a "mid-range 6xx" GPU. I will PM to him my observations and set up a test configuration for further data if he has an interest. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
>James I have no idea what you just said:) But TBar will. Im just plodding along, Learning as I go forth into the crunching world. [/quote] Old James |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
>James I think your results are a little off. Most people are reporting nVidia cards gaining around 2x the credits running APs. 3610secs for a 650 seems awful slow, unless you are running 2 at a time. It would still be a little slow at 2 at a time. It would be best to duplicate the original test. |
Gene Send message Joined: 26 Apr 99 Posts: 150 Credit: 48,393,279 RAC: 118 |
>TBar, The 7.5 factor is for just a single AP vs. a single MB on the 650. The 650 is able to do 3 concurrent MB (Lunatics x41g) and then the comparison is closer to the factor of 2 you suggest. As to the speed of the 650 on APs, roughly one hour, the 650 is a relatively low performance card, with only 2 compute units. And I don't have a clear idea what values to pass on the command line for unroll, ffa_block, etc., which might improve things. I am using unroll=4, ffa_block=2048, ffa_block_fetch=1024, sbs=256, and hp. I don't want to stray too far from the thread topic here. PM reply if you think I am... /Gene/ |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
I have 7 pendings one of which is a Ap that needs a manual poke one way or the other. The other V7's have a deadline as of 23 may. Rac is 2167 So far I have lost on average 3,233 in RAC by going stock MB only. To tell you the truth, I think Im just going to set NNT and get rid of these stock V7's and Just go to AP lunatic version. and let my little GPU run free. If it still works. [/quote] Old James |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
I still have 7 pendings on this host The deadline is for them is 23 may. So I hope they all validate,( except for that lone AP.) I am getting antsy about this test. Its taking so long. I would like to run four cores with stock to see what happens but I cant figure out how to get into BIOS. When I restart , The screen flashes by so fast I cant read all the F key options. Its an HP running Vista. I think all makers should have a standard F key for going into BIOS. Edit- I have read the manual, No info on how to go into BIOS. [/quote] Old James |
W-K 666 Send message Joined: 18 May 99 Posts: 19093 Credit: 40,757,560 RAC: 67 |
I still have 7 pendings on this host I am informed that it should be F1 or F10 when the logo screen is on for Compaq and HP computers. N.B. That info comes from a psychiatrist, take that as you will ;-) |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
I still have 7 pendings on this host Thanks winterknight. That beats restarting and hitting eack F key:) [/quote] Old James |
Wiggo Send message Joined: 24 Jan 00 Posts: 34905 Credit: 261,360,520 RAC: 489 |
I've been using this old hardware setup as a trial myself James in this thread to do the same thing, though I've started this rig from scratch (with a new ID) on AP's only just over 4 months ago, it's now set to MB only so it'll be interesting to see what happens once the AP's are out of the system (with a 10,500 average under AP's it didn't quite reach 50% of the RAC that it did under just MB V6 work back then). Cheers. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
WinterKnight, F10 was the key to unlocking Bios, I am now running for core stock MB. Lets see if the through put increases RAC. Rac as of the switch was 2175 With 7 pendings One is a stuck AP. Five others timed out. So have been resent. So lets see if the old saw about running 4 real cores can out RAC 8 on HT. Edit-this host [/quote] Old James |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
RAC is now at 2017 with 4 pendings One is that stuck AP. I started 4 real cores no HT with 2175 RAC. So I will run this for a bit and see what happens. edit--this host [/quote] Old James |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
RAC is now at 2017 with 4 pendings One is that stuck AP. If I were a betting person I'd say it will do somewhere between 1400-1700 with 4 cores. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
RAC is now at 2017 with 4 pendings One is that stuck AP. I think you might be right. I will let it run. Then go back to HT and see what happens. I never did buy the 4 core can beat 8 HT anyway. [/quote] Old James |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
RAC is now at 2017 with 4 pendings One is that stuck AP. In running only CPU tasks I have not seen an instance where HT did not provide some gain on the same system. When you toss GPUs into the mix running fewer CPU tasks has been shown to be an advantage in several instances. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
RAC is now at 2017 with 4 pendings One is that stuck AP. Which is why after the 4 real core test is finished I will got back to HT. Once that is stableized I will run HT with 7 cores and my GPU. Then will try 8 cores and a GPU. This will be all stock MB tests. [/quote] Old James |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Just recently switched from running 4 with HT on to running 8 on my i7-860. The initial results looks to be 11.1% increase in power consumption and 27.7% increase in output. So for MB tasks on the CPU a gain for the hardware used. I think I want to get 2 low/mid range GPU's, like some R7 240 or R7 250X's, to stick in that system and do some more tests similar to your planned ones. I want to also do some AP task tests & it looks like we have a semi reliable source of AP so I could perform an accurate test at some point in the future. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
RAC is 1822 with 4 pendings. One is an AP stuck in never everland.this host Im waiting for the 3 to clear and RAC to even out. I thought going to just 4 cores would make the RAC plumment. I will be interested on what happens when HT is turned back on. Hal, It just takes a so long to get RAC to even out. Im trying to be calm and collected. [/quote] Old James |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
RAC is 1822 with 4 pendings. One is an AP stuck in never everland.this host Yeah I ran mine from Apr 8th to May 30th to get a steady RAC value. If you track the daily values on the stat sites. Then you can get an idea of what the stabilized RAC will probably end up being. I like to wait until the graph in BOINC shows at least a week of a flat(ish) line myself. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
©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.