Message boards :
Number crunching :
Intel vs. AMD
Message board moderation
Author | Message |
---|---|
![]() ![]() Send message Joined: 15 Jul 05 Posts: 157 Credit: 34,715 RAC: 0 ![]() |
Hey guys, which do you think is better at crunching numbers in BOINC? Assume both chips have identical speeds. Absence of evidence is not evidence of absence. <img src="http://www.boincstats.com/signature/user_220108.gif"> +250 Classic Seti |
KB7RZF ![]() Send message Joined: 15 Aug 99 Posts: 9555 Credit: 3,308,926 RAC: 2 ![]() |
Well, I dunno. On one computer, I have an Intel Celeron 2.6ghz, and on the other is an Intel Pentium 4 at 1.6ghz, and the 2nd one crunches WU's faster than the faster machine. So I dunno. I like my Intels. I've had my Dell machine now for just about a year, and have had no problems at all with it (knock on wood). Of course I don;t do much installing of needless programs. Just emailing, surfing, and crunching, and being an email server. So thats my thoughts. I'd love to read others when they post. Jeremy ![]() |
[BOINCstats] Willy Send message Joined: 4 Mar 01 Posts: 202 Credit: 152,243 RAC: 0 ![]() |
Oh, man, you're unleashing a flame war again. One will say AMD, the other Intel. One thing is for sure though: assuming identical clock-speed: AMD wins. Without doubt. But, as it is now, the clock speed of AMD chips is lower then that of Intel chips. That doesn’t impact performance because AMD chips can perform more operations in one tick then Intel chips. If you want a single core CPU, your best buy will be AMD: cheaper and much faster in almost all disciplines. If you go for dual core it’s more difficult. Although the AMD X2 performs much better then Intel Pentium D, it’s also much more expensive. I would shell out the money for the X2, but that’s just me. Ow, forgot: If you take Intel, you will save money on house-warming. Your Intel will help in keeping your house nicely warm :) ![]() ![]() Join team BOINCstats |
Hammer Send message Joined: 13 Dec 02 Posts: 74 Credit: 1,773,558 RAC: 0 ![]() |
If you are going with single core: For Seti Intel with Hyperthreading is the fastest. My Prescott2M @ 4Ghz with optimized client can do around 30 Work units per day. Way more than any single core OCed Athlon64 Venice or SanDiego. But if you are crunching Einsten things are completely different, there Athlon64 wins the battle. If you are going dual core than AMD X2 is much better choice, maybe in seti it is tied or Smithfield wins by a small amount but X2 will destroy smithfield in Einstein or Predictor. And X2 is overall better choice if you can afford it. |
W-K 666 ![]() Send message Joined: 18 May 99 Posts: 19741 Credit: 40,757,560 RAC: 67 ![]() ![]() |
For crunching Seti, clock speed for clock speed the one with the bigger L2 cache wins. A Pent M with L2 of 2MByte will beat most 3GHz desktop CPU's, but being single core it will lose out on throughput. But with the right mobo chipset will do it all on about 60W of total power. Andy |
W-K 666 ![]() Send message Joined: 18 May 99 Posts: 19741 Credit: 40,757,560 RAC: 67 ![]() ![]() |
[Mistake!!! |
Simplex0 Send message Joined: 28 May 99 Posts: 124 Credit: 205,874 RAC: 0 |
My Prescott2M @ 4Ghz with optimized client can do around 30 Work units per day. Is that 30 units / 24 hours? |
Astro ![]() Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
my non overclocked laptop with a mobile AMD64 3700+ does about 22 a day. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 ![]() |
If it helps, I have a low cost Dell 2.8GHz P4 w/o HT that is returning 21 per day. May this Farce be with You |
![]() ![]() Send message Joined: 15 Jul 05 Posts: 157 Credit: 34,715 RAC: 0 ![]() |
Well, I dunno. On one computer, I have an Intel Celeron 2.6ghz, and on the other is an Intel Pentium 4 at 1.6ghz, and the 2nd one crunches WU's faster than the faster machine. So I dunno. I like my Intels. I've had my Dell machine now for just about a year, and have had no problems at all with it (knock on wood). Of course I don;t do much installing of needless programs. Just emailing, surfing, and crunching, and being an email server. One of my hosts is a dell 8200. When I bought it, I had the top of the line 2.0 Ghz (remember when that was fast?). Since then, I've fried that CPU and now it's got a 1.7 p4HT in it. It can hold it's own, just doesn't seem to be completly reliable when doing CPU intensive tasks over a long time. At least not for me. Absence of evidence is not evidence of absence. <img src="http://www.boincstats.com/signature/user_220108.gif"> +250 Classic Seti |
![]() ![]() Send message Joined: 15 Jul 05 Posts: 157 Credit: 34,715 RAC: 0 ![]() |
Oh, man, you're unleashing a flame war again. One will say AMD, the other Intel. Let the flamers come. I'll just ignore them. It's the people that have valid points that I'm interested in. I'd like to make a point. I forgot to include the new dual proccesor G5 from Apple in the list. My "Power" computer, the one that I built most recently, is based on an MSI board with an Athalon 64 3200+ with a gig of ram. At least as far as Boinc is concerned, I've seen much better results on the AMD. Of course, I do have dynamic oveclocking enabled. And to be fair, the Dell (my p4 machine) is running on an intel board from 4 years ago. Nevertheless, I've actually noticed that in this case, my AMD is more stable and run a hell of a lot cooler than either my Dell, or my Compaq Notebook (2.4Ghz P4). As for the dual core thing. How many people have actually purchased these things. I won't go anywhere near them until all kinks are worked out. I usually don't by the 1st gen of a new processor architecture. Just asking for trouble is what that is. But, that's only my opinion. Absence of evidence is not evidence of absence. <img src="http://www.boincstats.com/signature/user_220108.gif"> +250 Classic Seti |
![]() ![]() Send message Joined: 15 Jul 05 Posts: 157 Credit: 34,715 RAC: 0 ![]() |
If you are going with single core: There's a 4Ghz P4?! When did that happen. I knew they were close, but didn't know they broke the barrier yet. BTW, you forgot to mention the NewCastle, which is what I have. Now that I'm running optimized clients on all my computers I've seen great improvements. But the AMD had the best improvement. Around 24 - 26 work units a day vs. 12 or so before. That 30WU that you are getting is quite impressive though! Absence of evidence is not evidence of absence. <img src="http://www.boincstats.com/signature/user_220108.gif"> +250 Classic Seti |
![]() ![]() Send message Joined: 16 Jun 99 Posts: 211 Credit: 259,752 RAC: 0 ![]() |
Well, as Hammer indicated a few messages back. It really depends on the project. For SETI, here are some quick time samples. Keep in mind that the Intel's are running two work units at once, so a single work unit time is not indicative of actual speed, and must be divided by two for the overall average. 12 work unit sample All machines running Tetsuji's optimized apps. Your mileage may vary. Pentium D 840 (dual core) 3.2 ghz no HT -- 3524/2 = 1762 Pentium 4 3.4 ghz w/HT -- 4982/2 = 2491 Pentium 4 3.0 ghz w/HT -- 5467/2 = 2733 AMD 64-3200 -- 4284 Athlon XP 2800 -- 7224 (may be inaccurate due to only a 5 unit sample) It would be interesting to see how a dual core AMD 64 with 1 meg L2 cache would compare. Also, since the AMD's listed don't have SSE3 they can't run the fastest optimized clients. Later AMD's might do better. For other comparision, my iMac G5 1.6ghz running an optimized SETI app averages 6512 per unit. Typically, SETI has benefitted from large caches. Einstein benefits from Short Pipelines, and Predictor is somewhere in between. For SETI, Intel, the bigger the cache the better. For Einstein, Athlon or Power PC, for Predictor, Atlon or Power PC. Don't know about Climate or the other projects since I haven't ran them recently. Of course, more speed is always better, regardless of platform. ![]() Team MacNN - The best Macintosh team ever. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 ![]() |
In your experience are the einstein and predictor apps optimized for your platforms? If not, then comparison is meaningless it seems. May this Farce be with You |
![]() ![]() Send message Joined: 16 Jun 99 Posts: 211 Credit: 259,752 RAC: 0 ![]() |
In your experience are the einstein and predictor apps optimized for your platforms? If not, then comparison is meaningless it seems. Well, the only comparision was against various machines running SETI. The different projects utilize different types of calculations which benefit from different architectures. Regardless of whether the app is optimized or not, SETI benefits from large cache, Einstein benefits from short pipelines. (My Athlon 2800 runs neck and neck with the AMD 65 3200 due to shorter pipelines on Einstein) Predictor is somewhere in between. For what it is worth though, Predictor doesn't offer any optimized apps. Einstein already had a well optimized Windows app, and now offers optimized versions for Linux and Mac OS-X. The OS-X app is altivec optimized which currently gives it a significant advantage per gighertz. In essence, there is no one platform that is best for all of BOINC. Each platform has its strengths and weaknesses. What is best for one project, may be the worst for another. The best for SETI is currently Intel, but any of the platforms can do a good job. ![]() Team MacNN - The best Macintosh team ever. |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
I forgot to include the new dual proccesor G5 from Apple in the list. [font='courier,courier new']From what I've observed, the G5 kicks x86 right in the bus when it comes to Predictor, but is roughly on-par elsewhere. I'd wait out for the PS3, hack a distro onto it, and make it a dedicated box - That'll leave everyone biting the dust.[/font] |
![]() ![]() Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 ![]() ![]() |
Well, as Hammer indicated a few messages back. It really depends on the project. For SETI, here are some quick time samples. Keep in mind that the Intel's are running two work units at once, so a single work unit time is not indicative of actual speed, and must be divided by two for the overall average. Well i have a A64 X2 4400+ but it is overclocked though to 2.5 Ghz.. The cooling solution is all air and the heatsink is barely warm , core running at 54 degrees celsius and this machine hasn't had a single glitch at all. I probably can push it further up but i'll wait for the autumn season before i'll do that. This is a cut out from "Pending Credits for Seti@Home" at ztb.dyndns.org pending credit / granted credit / total credit* / # wus* / Ø seconds_wu / Ø time_wu / Ø cc_wu* / Ø gc_wu* / last wu received from server 6,505.45 / 40,900.18 / 47,405.63 / 494 / 3,104.94 / 0h 51m 44.940s / 13.17 / 19.92 / 21 Aug 2005 21:08:37 UTC According to average times it returns two results in 3104 secs = 1552 secs for one wu.. For the Intel/Amd record.. I've worked with computers for a long time now and for me what AMD has done by squezing two cores onto one die with that low amount of heat it produces with full load is simply nothing short of a engineering miracle.. They have done such a tremendous job with this CPU so if anyone can afford it by now then buy the damn sucker. It simply destroys all Intel P4 based cpus with it's low heat output, superior performance when gaming and up to par with intel in videoapps, and finally it doesn't need a nuclear powerplant outside your window to be fed with.. But.... Intel has done a really tremendous work with their P3 based Pentium M processor and when it drops into dual core solution market i think it will dominate most intelsold systems in about 1,5 to 2 years from now.. But i don't really know but it feels like it though... Kind regards .. Vyper ![]() _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 ![]() |
Regardless of whether the app is optimized or not, SETI benefits from large cache, Einstein benefits from short pipelines. (My Athlon 2800 runs neck and neck with the AMD 65 3200 due to shorter pipelines on Einstein) Predictor is somewhere in between. I don't think this is right. The type of optimization and the use of optimized math libraries may factor things out. It seems that Einstein, like old seti and non-optimized seti, has a one size fits all approach, which means they could do better. (I just joined E@H just to see what happens and there doesn't appear to be any handshaking regarding cpu type when the app was installed.) May this Farce be with You |
Don Erway Send message Joined: 18 May 99 Posts: 305 Credit: 471,946 RAC: 0 ![]() |
I've always run 2 PCs, one Intel, and one AMD, for the last 8 or 10 years. Currently, AMD-64's are a huge win, in another way - low power use. My host mycroft can do 28 WUs a day, and do it at 100 watts less than the pentiums P4 HTs can. It is the fastest yet recorded time for the reference WU at: http://www.marisan.nl/seti/reference.htm And it is with the stock AMD heatsink and fan, and not noisy. Now the "slower but double processing", of hyperthreading P4s, which can still beat the amd's, for overall throughput, if you can get rid of enough heat. In this day and age, we really have to think environmentally, when making these choices too. The AMD Athlon 64's are overclockable as heck too, so you can save $500 on processor, and still kick it up. Don |
Ned Slider Send message Joined: 12 Oct 01 Posts: 668 Credit: 4,375,315 RAC: 0 ![]() |
No doubt about it, the AMD64 X2's are the current cream of the crop - they are significantly the fastest processors and operate at far lower power producing far less heat. If you can afford one this is definately the way to go :) Ned *** My Guide to Compiling Optimised BOINC and SETI Clients *** *** Download Optimised BOINC and SETI Clients for Linux Here *** |
©2025 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.