Author: Vincent Diepeveen
Date: 05:14:54 04/13/03
Go up one level in this thread
On April 13, 2003 at 08:02:34, Vincent Diepeveen wrote: Hello Tom, For the endusers there is only 1 way to easily say what SMT is. That is that the P4 processor can split itself into 2 processors and that they get that second processor for free and it speeds DIEP up about 15% in nodes a second. It may be clear that this is not how i see SMT. I assume SMT is that intel has put 2 processors near each other which share resources. Using those resources probably is what takes care they cannot run both at full speed. In future hopefully less and less resources get shared, until only L2 cache gets shared where also reads can be done simultaneously, but only a write will be blocking other cpu from reading/writing to the L2. Then HT will give a major boost in speed. It is questionable whether we must call it HT then however. I guess intel will keep using that word for the coming 10 years though. >On April 12, 2003 at 23:29:18, Tom Kerrigan wrote: > >>On April 11, 2003 at 23:26:35, Robert Hyatt wrote: >> >>>First, I didn't say it did or it didn't. I said that tests suggest that there >>>can be imbalances. >>> >>>Second, you found a result for _one_ test. What about one that does a lot of >>>memory reads? Memory writes? Mixture? >>> >>>There are _lots_ of tests to do. >> >>Wow, Bob, you're getting quite a workout. First the furious handwaving about how >>the logical processors are imbalanced (Cray YMP this, Intel secrecy that) and >>now furious backpedaling. >> >>You have been criticizing people for "bad math" this entire thread. You rejected >>the notion of a 50%-50% division: >> >>"But I don't buy the 50% stuff, the cpu is not that simple internally. One >>thread will run at nearly full speed and the other gets slipped into the gaps" >> >>and came up with this gem of idiocy: >> >>"If your NPS goes up by 10%, then with a 1.7x multiplier on two real cpus, the >>program should run 1.07X faster using SMT." > >1 + 10% = 1.1 x (1.7 / 2.0 ) = 0.935 actual speedup > >>And now you're trying to maintain that you never said the logical CPUs were >>necessarily unbalanced? Hilarious. > >>What's even more hilarious is the way you argued your point--first saying that >>some guy came up with some numbers that I should look up (uh huh) and then >>saying you couldn't test this stuff yourself, when even a retarded 3rd grader >>could come up with a way to test it. >> >>Now you're saying my testing was incomplete? Yeah right. Any _moron_ can tell >>you that if you run a memory intesive program with a CPU intensive program, the >>CPU intensive program will get most of the CPU time, just like it utilizes most >>of the CPU on a system with one logical processor. These situations obviously >>don't need to be tested. The question at hand was logical CPU division for chess >>programs, where both threads have exactly the same performance characteristics. >>-Tom
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.