Author: Tom Kerrigan
Date: 18:44:06 07/04/03
Go up one level in this thread
On July 04, 2003 at 21:21:47, Jeremiah Penery wrote: >On July 04, 2003 at 18:21:08, Vincent Diepeveen wrote: > >>On July 04, 2003 at 17:01:36, Jeremiah Penery wrote: >> >>>On July 04, 2003 at 14:37:37, Vincent Diepeveen wrote: >>> >>>>On July 03, 2003 at 16:48:07, Tom Kerrigan wrote: >>>> >>>>>On July 03, 2003 at 16:23:10, Russell Reagan wrote: >>>>> >>>>>>Also the length of the pipeline on the P3 is 10, which means that a mispredicted >>>>>>branch costs 10 cycles. On the P4 the length of the pipeline is 20, which means >>>>>>it costs 20 cycles for a mispredicted branch. I may be wrong about the actual >>>>>>numbers (10 and 20, but I think they are close). I'm not sure what the length is >>>>>>on the Athlon. Anyone know? >>>>> >>>>>Pentium 3: 12 cycles >>>> >>>>9+ cycles. Usually it's more like 15 though when you measure. >>>> >>>>>Pentium 4: 20 cycles >>>> >>>>20+ usually it's more like 30+ though. >>>> >>>>>Athlon: 10 cycles >>>> >>>>there is no official data on this and i won't sign a NDA ever of either intel or >>>>AMD. AMD answerred my public question a few years ago: >>>> >>>>"Mr Diepeveen, it is more than the P3, but the exact amount is secret" >>> >>>You're spewing gibberish again. The pipelines of all of these processers are >>>very well documented. I don't know what numbers you're trying to give. Maybe >>>you're confusing pipeline length with branch misprediction penalty. >> >>Of course the pipe line length is not so interesting, but the resulting branch >>misprediction penalty out of it is! > >Too bad you didn't say that in the first place, thus avoiding this entire sub >thread. Instead you claimed that the pipeline lengths themselves were >different, which was blatantly wrong. Not to mention that the discussion was >about _pipeline length_, not branch misprediction penalties. > >>>The rest of what you say is completely irrelevant to the discussion. >> >>No it isn't the original poster asked why the P4/Athlon is so much faster for >>computerchess than the P4. As usual your short term memory i bet. >> >>Seems only slightly longer now than Bob's. > >The post you replied directly to was talking about pipeline lengths only. That >was also the only section you quoted. If you wanted to reply to the rest of the >stuff, then quote the text with the relevant reply directly underneath. >Otherwise, it is irrelevant to the discussion, as I originally said. Partially my mistake, I gave the pipeline length in "cycles" which doesn't make any sense (instead of stages). But branch mispredict penalties are pretty close to pipeline lengths... of course, other factors are involved, dependencies, bubbles, forwarding, queues, memory latencies... a branch mispredict penalty could be hundreds of cycles... but in general I think it's safe to say that mispredicts hurt the P4 much more than the Athlon. -Tom
This page took 0 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.