Message boards :
Number crunching :
Threaded number crunching vs real CPU core for each WU
Message board moderation
Author | Message |
---|---|
gs Send message Joined: 18 May 99 Posts: 45 Credit: 5,412,660 RAC: 8 |
I was wondering, if it would have a real impact on the RAC, if I would replace my Intel Core i3 540 (2 cores, 4 threads) with an Intel Core i5 760 (4 cores, 4 threads, no hyperthreading)? Would the i5 crunch the WUs faster than the i3? IMHO it would, because through i3 hyperthreading, a CPU core would deal with two threads in parallel, by that spreading computing power of one core on two WUs while with the i5, each WU would be computed by a single core with full computing power available. |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
you should gain a bit of performance. for the reasons you mentioned and likely because of the i7's architecture is vastly better than an i3 In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
He talks about I5 not I7 (sure a lot better). I have few hosts with I5 you could compare the crunching times with yours, but remember i use GPU crunching so the I5 needs to feed the GPU too. SETI works better (faster) with NO HT a I7. I imagine is the same on a I3. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
He talks about I5 not I7 (sure a lot better). Ive seen pros and cons to that statement. Im waiting for my I7s to stablelize in Rac. My first test will be to free a core for my 550 Ti's and see how that works out. Then I will disable HT and see how the rac hold up to that so called fact. Might take a while so be patient. I think my Rac on those two machines is getting close to stable. [/quote] Old James |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
SETI works better (faster) with NO HT [on] a I7. I imagine is the same on a I3. I am not so sure of that. On my I7-3820 with HT on, I see no stretchout of CPU WUs (i.e., Elapsed time >> CPU time) as I would expect if 2 HTed WUs on the same real core were contesting for the shared resources implicit in HT. That is, it is "HT" because the 2 "cores" are not completely separate, but share some of the physical logic. All those WUs (running 7 of 8 virtual cores except when doing an AP on GPU) run at 95+% CPU/Elapsed. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
As allways YMMV, each host is unique, all depends on large number of things, specialy what is running besides the crunching programs itself, NO HT is the recomended setting (you could verify that in a lot of places in this forums). But sometimes is hard to feel de diference. I inclusive run with HT On on my I7 because even if it could run a little faster with HT OFF (crunch the CPU WU a little faster, i made few long term test in the past) but with HT ON it´s video apears to run smoothly (i use 2x690 on it) just a question of feeling hard to explain. And BTW keep thinking on CPU crunching speed on a GPU based cruncher is meaningless, each one of the 690 GPUs crunch a lot times faster than any I7 AFAIK Anyway a I5 must be faster than an I3 on similar conditions, that i belive was the initial question. Who much? That´s why i say to compare the WU (similar AR of course) crunching times obtained by his I3 against the times of my I5 hosts, so he could see by himself. Talking about the GS specific host, he has a 660, so comparing with the 660 performance the change of the processor form a I3 to a I5 will make little impact in his RAC since most of it comes realy from the 660 itself. He maybe get a better increse in the RAC if he run 2 WU at a time (if he allready don´t do that) or run 1AP+1MB instead of 2MB for example. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I was wondering, if it would have a real impact on the RAC, if I would replace my Intel Core i3 540 (2 cores, 4 threads) with an Intel Core i5 760 (4 cores, 4 threads, no hyperthreading)? The 760 does have more actual cores, but a lower clock speed than the 540. The gains of the actual cores might be lost with the reduction in clock speed. Unless you were planing you lock the i5 in turbo. Then it with a doubt be faster. Previously I was running my I7-860 at home with HT on but only 4 threads. At work I was running 5 i7-860 with 8 threads. With the machines only doing MB my home machine had nearly the same RAC, 80-90%, as the machines running 8 threads. However the machines running 8 threads were always higher. I have not done this same comparison since SETI@home v7 was released. 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.