I am running 4 computers, 2 Intel and 2 AMD, and wondered if anybody can tell me why the Intel CPUs get more credit for the work they do. I know this has a lot to do with other things but there seems to be an Intel advantage and I wondered why.

I also wonder about Linux versus Windows. I read somewhere where Windows XP generated more results. I have run both (Windows 2K) and Windows seems to generate more credits. I know this would have a lot to do with system configuration, but I ran web servers, database servers and GUIs on both. Typically on Windows I am running IIS and Apache, and SQL Server and MYSQL whereas on Linux I ran just Apache and MYSQL.

GenuineIntel
Intel(R) Pentium(R) 4 CPU 2.40GHz [x86 Family 15 Model 2 Stepping 7] [fpu tsc sse sse2 mmx]

103592321 94025080 1 Sep 2007 5:59:40 UTC 1 Sep 2007 13:01:11 UTC Over Success Done 14,629.39 22.89 30.16
103564381 93998622 1 Sep 2007 2:44:48 UTC 1 Sep 2007 8:53:11 UTC Over Success Done 10,378.64 16.24 23.43
103525073 93961298 31 Aug 2007 22:32:48 UTC 1 Sep 2007 5:59:40 UTC Over Success Done 11,579.97 18.12 19.58
103492946 93931161 31 Aug 2007 19:26:32 UTC 1 Sep 2007 2:44:48 UTC Over Success Done 14,928.11 23.35 30.59


AuthenticAMD
AMD Athlon(tm) 64 Processor 3500+ [x86 Family 15 Model 7 Stepping 1] [fpu tsc sse 3dnow mmx]

103582821 94016120 1 Sep 2007 4:51:21 UTC 1 Sep 2007 9:37:24 UTC Over Success Done 8,663.44 27.38 12.80
103556123 93990820 1 Sep 2007 1:51:58 UTC 1 Sep 2007 7:14:03 UTC Over Success Done 8,685.56 27.45 12.89
103528286 93964353 31 Aug 2007 22:54:46 UTC 1 Sep 2007 4:51:21 UTC Over Success Done 10,692.66 33.79 34.26
103497763 93935562 31 Aug 2007 19:53:02 UTC 1 Sep 2007 1:51:58 UTC Over Success Done 10,554.94 33.35 15.90

Thanks,
WOR

---

Well I can tell you that the Intel/AMD difference is mainly to do with Intel CPU's having normally double the cache sizes to AMD. But that's about it. SSE3 isn't used (neither is SSE or SSE2) for that matter, so that doesn't make a difference. Clock speed doesn't make a difference, because AMD CPU's do more operations per clock cycle than an Intel, they have higher efficeincy (check this on your stats) but Intels sort of make up for it by having higher clock speeds, until C2D came out, they sorted out the efficiency problem with a new architecture I think (Conroe or something, or is that the core type?).
AMD's have a better FPU but lose out on Rosetta by having smaller caches, which is I think why they are cheaper than Intel. Because the SRAM used for CPU cache is expensive to manufacture.
Also I'd like to point out that you show here in your post an AMD 64 3500+ against a 2.40GHz P4. The "3500+" on the AMD is meant to show that it is equivilent to a 3.50GHz P4, even though it may have a much lower clock speed, due to the higher efficiency of AMD's mentioned earlier. So it's not a perfect comparison, but I can see your point. Intel's genuinely do outperform AMD's at rosetta because it's so cache intensive, one has only to look at the "top computers" list to see this. But AMD's are generally better at everything else (gaming, multitasking even with one core, overclocking etc) but I'm not biased, I just like em cos they're good :)

---

Winton, your machines are hidden so I couldn't post your machines' specs, but yes, there's more to getting work out of a CPU then the Ghz. L2 cache is often the primary difference.

---

Rosetta Moderator: Mod.Sense

i hope no too ot, but since i want to make the most efficient use of the crunching power i have, i would appreciate some recommendations.

i think we can agree that intels with large L2 caches excel at rosetta.

i will keep my new dell quadcore as a rosetta only cruncher.

now, i have two amd's, an XP-M 2500, and an A64 3800. they seem to be getting only about 33% of requested credit.

without starting a war, does anyone know which projects (excluding, as a matter of personal preference, Seti and Einstein) would be "efficient" for these two AMD's?

Folding@home seems to demand quite a lot of Cache. It seems machines with less than 2MB per core take twice as long to complete a task!
There are other projects such as Ufluids@home which require only small amounts of memory (<10MB) so I'm guessing would be fairly cache unintensive. But I think the reason that larger cache is good is that the application can neatly fit all it's instructions in it, rather than having to fetch them from slower RAM. Not to mention the fact that is anybody bothered what happens to microfluids, makes better satellites or something. I think charity starts at home, cure disease
I have an AMD 64 3000+ which gets about 35-40 creds per work unit and has only a 512KB L2 cache! It gets about 5-15 more creds on average than a competing P4 (with or w/o HT). So personally I find the cache argument a little puzzeling.

---

i hope no too ot, but since i want to make the most efficient use of the crunching power i have, i would appreciate some recommendations.

i think we can agree that intels with large L2 caches excel at rosetta.

i will keep my new dell quadcore as a rosetta only cruncher.

now, i have two amd's, an XP-M 2500, and an A64 3800. they seem to be getting only about 33% of requested credit.

without starting a war, does anyone know which projects (excluding, as a matter of personal preference, Seti and Einstein) would be "efficient" for these two AMD's?

Well (I know) Seti with lunatics-kwsn optimised seti works well on the AthlonXP and the Athlon64... I think the 64bit version may give a slightly faster result on the Athlon64.

Pop to WGC though and see what they are saying, either that or go where the project scores on benchmark.



Unfortunately with the two comparisons about there is really not enough data to give a real comparison, since with that snapshot it could easily be target variation, plus the time length is different.. But Pentium4's used to get a lower benchmark than they would crunch, hence whay they are probably claiming more (the benchmark is just wrong). It does not mean the Athlon64 is now slower.
That and the Athlon64 is at a lower MHz clock so a clock for clock comparison is more difficult.

---

Team mauisun.org