Message boards :
Number crunching :
Which CPU is faster Pentium II or the K6-2/3?
Message board moderation
Author | Message |
---|---|
Iopodx Send message Joined: 3 Apr 99 Posts: 13 Credit: 208,352 RAC: 0 |
Hey guys, I've got a smart question for you... I'm going to have a small farm with about 20 computers... Well, I've not much money, but I'd like to have most of the power I can buy for the money I have. So I have the choice between Intel Pentium II CPUs and AMD K6-2 or K6-3. I think the Pentium CPUs are a little bit faster than the AMD CPUs... In fact, a Pentium III would be better, because of the SSE support, but Pentium III is too expensive for a student like me. I would be glad about an answer... Thanks. Best regards, Iopodx PS: Sorry for my really bad English... |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
Well you can look at mine if you want. The K6 3D's are K6-2/500's and the PIII is a Katmai 550. IIRC, PII's are generally more or less equivalent to the K6-2/500's probably a little better if you get late model ones clocking in the 400 MHz region. Personally I think you'd be better off going with fewer PIII's on Socket 370 (Coppermine and up) from a total cost viewpoint. One advantage to K6's over PII's is there seems to be a lot more Socket 7 and Super 7 boards out there than Slot 1's. In any event, running any "vintage" farm really isn't cost effective, since almost any modern "econobox" will blow away a small army of old timers, unless you you were planning to be experimenting with clustering and would be running them 24/7 anyway. HTH, Alinator |
Iopodx Send message Joined: 3 Apr 99 Posts: 13 Credit: 208,352 RAC: 0 |
Uhhm... The Pentium II compared to the K6-3 is a little bit faster as the benchmark says. But this is in a range in which this difference is equal. Thanks for your information, I think I'll buy the cheapest ones... Cya Iopodx |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
From my experience, AMD K6-2s and K6-3s were not as fast as the PIIs. Here are some actual BOINC benchmark numbers from my K6-2 450/500 & my Pentium II 333: AMD 500: FPU=346 million ops/sec, ALU=578 million ops/sec. AMD 450: FPU=325 million ops/sec, ALU=502 million ops/sec. PII 333: FPU=269 million ops/sec, ALU=501 million ops/sec. PIII 533B: FPU=439 million ops/sec, ALU=759 million ops/sec. Notice the PII 333 clocks the same as the K6-2 450 with a 66MHz bus compared to the K6-2's 100MHz FSB! A PII 450 would surely beat the AMD 500 with no problems. I threw in the PIII 533B (Katmai core, 133MHz FSB) just for comparison. My advice: if the PIIs are about the same price or only a little more than the K6's, the PIIs are a much wiser investment from a SETI@Home standpoint. If they are quite a bit more expensive, then you might be able to afford more K6's and thus produce more results through brute force (having more of them) than going with the PIIs. |
Online Game Commands Send message Joined: 5 Sep 04 Posts: 15 Credit: 25,868,225 RAC: 0 |
I have a few spare PII 350, PII 400Mhz and maybe couple PIII 500 (CPUs only) becoming spare. If they are for SETI and you pay the postage from the UK they are yours for the taking. But they must be for SETI, i hate giving old stuff away then to see people selling it on Ebay. Online Game Commands |
ADLKIRK Send message Joined: 3 Sep 04 Posts: 90 Credit: 48,158 RAC: 0 |
I have a PIII hanging from my car's mirror and am using a PII with a clip magnet glued on it for messages..... I am currently running a P4D (820) on an ASUS P5LD2..and will be shortly upograding to the 950EE.. Benchmarks are, frankly, simple guidlines. What you actually end up with is dependent upon what you combine them with: RAM, mobo, vid card, etc... For an example...running in RAID0 configuration, 1.024 gigs DDR2 RAM, 256 Meg ATI vid card, I generally blow away the comparable AMD in a rather run-of-the-mill setup. |
Iopodx Send message Joined: 3 Apr 99 Posts: 13 Credit: 208,352 RAC: 0 |
I have a PIII hanging from my car's mirror and am using a PII with a clip magnet glued on it for messages..... Well I think my 3200+, 2x 7900GTX (Club3D), 2GB A-DATA Vitesta 500 is much faster then your stupid Intel ;) MfG Iopodx |
Online Game Commands Send message Joined: 5 Sep 04 Posts: 15 Credit: 25,868,225 RAC: 0 |
|
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
I don't know how we got off topic and started bragging about fast PCs when that was clearly not the topic, but yes, it's true, benchmarks are quite synthetic and aren't always a good idea to go by. However, in this case, these BOINC benchmarks will help the OP with their decision on which CPU to buy for a new farm. The CPUs in question were about old K6's and PIIs, not P4Ds or Athlon 64s. Generally speaking, this synthetic BOINC benchmark is very relative to the question asked since SETI is more of a synthetic science application instead of a multi-jump gaming application. Can we please stay on topic? No need to start a flame war about Intel vs. AMD or who has RAID 0 (which helps very little with SETI since most of the work is limited to RAM). |
Macroman1 Send message Joined: 30 May 99 Posts: 67 Credit: 12,532,684 RAC: 0 |
Hey guys, The edge goes to the PII. The FPU in the K6-2/3 is (for want of a better term) nasty slow. |
Toby Send message Joined: 26 Oct 00 Posts: 1005 Credit: 6,366,949 RAC: 0 |
Are you getting some parts for free here? If not why not just make 1 or 2 dual core systems for the same price? If you look at my cross-project stats page you will see that my P3 (500 MHz) gets about 40 cobblestones/day (RAC). My dual cores are getting over 700. So a single one of my dual cores will probably get more credit/day than 20 PIIs. Now I spent over a $1,000 on mine but AMD will be dropping their price on dual cores soon and an Intel dual core can be had for a pretty cheap price already. I also got a fancy case and a good video card. If you replace those with cheaper parts I think you could probably build a dual core for $600 or less and it would crunch better than 20 PIIs. Also consider power. 20 PIIs will be sucking more electricity than 1 or 2 newer systems. A few things to think about :) A member of The Knights Who Say NI! For rankings, history graphs and more, check out: My BOINC stats site |
Alex Kan Send message Joined: 4 Dec 03 Posts: 127 Credit: 29,269 RAC: 0 |
The edge goes to the PII. The FPU in the K6-2/3 is (for want of a better term) nasty slow. True, the K6-2/III's FPU is slow, but it also has 3DNow! for single-precision floating point SIMD, which the Pentium II doesn't have. Has anyone tried compiling 3DNow!-optimized crunchers? I realize that this can't possibly be on anyone's priority list, since most AMD chips do SSEn now, but older versions of FFTW (prior to 3.1) can be configured with --enable-3dnow to use the instructions supported on the K6-2 and K6-III. |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
Toby: Good points. Perhaps spending the money on a very cheap dual core would be easier in the long run (electricity) than a bunch of slower/cheaper machines. Something to think about. Alex: Good point as well. I have two K6's that I'd love to speed up but they don't have SSEx, so the only optimizations they can support is the antiquated MMX instruction set. I'm sure 3DNow! would provide a better performace boost than MMX on these chips. Unfortunately (and I'm generalizing here) most optimized clients are using Intel code which (obviously) doesn't support 3DNow! CPUs, and very few are going to take the time to code for 3DNow! themselves (nor would it come with as good of optimizations as the manufacturer direct code). But perhaps somebody could work on it? Eventually? |
KWSN - Chicken of Angnor Send message Joined: 9 Jul 99 Posts: 1199 Credit: 6,615,780 RAC: 0 |
Good points indeed :o) I'm planning to try it, actually - OzzFan1 is right, there is no way (AFAIK) to use either ICC or IPP to optimize for 3dNow!. So, I still also believe that a version compiled with the Microsoft Compiler or DevCPP and 3dNow! optimizations would best an MMX client with IPP on non-SSE AMD chips. I haven't yet had the time to put this hypothesis to the test, though. Also, I'm unsure whether anyone ever tried compiling FFTW with ICC/IPP (or whether this even makes sense) instead of standard compilers, and whether this might not also speed up things. Regards, Simon. Donate to SETI@Home via PayPal! Optimized SETI@Home apps + Information |
ADLKIRK Send message Joined: 3 Sep 04 Posts: 90 Credit: 48,158 RAC: 0 |
I have a PIII hanging from my car's mirror and am using a PII with a clip magnet glued on it for messages..... Gee..wazzat a flame??? :D You can "think" what you want. ;) Facts please and don't link to those stupid benchmark sites because as I said..... component selection is quite important. |
ADLKIRK Send message Joined: 3 Sep 04 Posts: 90 Credit: 48,158 RAC: 0 |
I don't know how we got off topic and started bragging about fast PCs when that was clearly not the topic, but yes, it's true, benchmarks are quite synthetic and aren't always a good idea to go by. I don't know how we got there. I simply answered the topics question and someone felt the need to go to the "who's is faster" b.s. And yes, SETi is RAM dependent...but that wasn't really the question. I provided that in a RAID configuration, over an IDE a pc can and will operate more quickly. |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
Facts please and don't link to those stupid benchmark sites because as I said..... component selection is quite important. I don't know if this will help, but I've randomly selected users of STANDARD boinc clients and recorded the following database. These are apples to apples so hopefully it will be of some help. tony |
hiamps Send message Joined: 23 May 99 Posts: 4292 Credit: 72,971,319 RAC: 0 |
[quote]Facts please and don't link to those stupid benchmark sites because as I said..... component selection is quite important. I don't know if this will help, but I've randomly selected users of STANDARD boinc clients and recorded the following database. These are apples to apples so hopefully it will be of some help. tony I am the only one using win 2003 Pro that isn't a server? Why doesn't mine say XP Pro 64? Is there a difference? And you missed my Overclock...LOL Official Abuser of Boinc Buttons... And no good credit hound! |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
[quote]Facts please and don't link to those stupid benchmark sites because as I said..... component selection is quite important. I can only get what is shown. When this started some users manually gave me OC facts, but I can't get them from the available info, so the OC section is exceptionally blank. LOL What I find interesting is how little the benchmark changes between 500 Meg and 1 G ram on some systems. There is even ONE system listed with 16 GB ram, but a LOW benchmark. I guess there's a limit to benchmark improvement via ram installation. I don't know if additional ram increases speed as that info isn't available via my search method. Also, Joe Segur is the only one with a processor Boinc couldn't recognize, it says "unknown" and he's successfully crunching seti with Win95. |
hiamps Send message Joined: 23 May 99 Posts: 4292 Credit: 72,971,319 RAC: 0 |
[quote]Facts please and don't link to those stupid benchmark sites because as I said..... component selection is quite important. Mine had 4 gigs but pulled 2 out as the only difference I could see was a limiting to my Overclock. More CPU and a bit less memory helped me quite a bit...I was joking with you about my overclock...You did a lot of work there and I find it very interesting. Official Abuser of Boinc Buttons... And no good credit hound! |
©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.