Author: Robert Hyatt
Date: 12:12:36 12/02/02
Go up one level in this thread
On December 02, 2002 at 14:48:33, Aaron Gordon wrote: >On December 02, 2002 at 10:49:47, Robert Hyatt wrote: > >>On December 01, 2002 at 23:13:39, Aaron Gordon wrote: >> >>>On December 01, 2002 at 22:19:16, Robert Hyatt wrote: >>> >>>>On December 01, 2002 at 12:57:51, Aaron Gordon wrote: >>>> >>>>>On December 01, 2002 at 12:28:38, Robert Hyatt wrote: >>>>> >>>>>>On December 01, 2002 at 05:03:00, Jorge Pichard wrote: >>>>>> >>>>>>>Also a comparison of the new AMD XP 2600+ Vs Intel P4 2.53 Ghz ! >>>>>>> >>>>>>>http://www17.tomshardware.com/cpu/02q3/020821/athlonxp-04.html >>>>>> >>>>>> >>>>>>That just highlights what others have said before. AMD has a memory >>>>>>bottleneck that is hurting them significantly in the war with Intel... >>>>> >>>>>Hyatt, it's not nearly as bad as you think. It happens I have a KT333 chipset on >>>>>an Epox 8K5A2. I'm running "Turbo" settings in the bios (but running CL2.5 >>>>>instead of CL2(cl2=faster)) and I'm getting 2.5gb/s. >>>> >>>>So? That is about 1/4 what the PIV xeon can do. Go to tom's hardware page >>>>and you'll find some interesting comparisons... >>> >>>I have been going to Tomshardware for years now, watching the stuff he's been >>>saying, stuff he's posting and etc. I haven't seen anything interesting in years >>>as well. You say the P4 Xeon can do 10gb/s? (2.5x4). I think not. >> >>The number I have seen for the 2.8 xeon on the E7501 chipset was about >>6.5gb/second. I either mis-typed my response or misread your number as >>1.5, I don't remember which. But the AMDs are not approaching 6+ gb/sec >>yet in any test I have seen. And on a few of the really memory intensive >>_real_ applications such as linpack, the fastest intel box has been posting >>number 3-4x that of AMD... And I'm not talking about overclocked numbers as >>I ignore those. >> >> >>>The efficiency >>>on P4's is less than desirable. The 133fsb(533QDR) PC1066 RDRAM P4's can >>>supposedly do something like 8.5gb/s. Actual bandwidth = 3.2gb/s. Not too >>>impressive. Around 37% efficiency. As as said in the previous message an Nforce2 >>>can manage over 3gb/s at over 95% efficiency & with much lower latencies >> >>I have looked around and don't see any references to support the above. The >>2.8xeons are at the top of the bandwidth heap on every chart I have seen. >>Notice that I am specifically addressing "2.8ghz xeon using the E7500 chipset" >>as that is the machine I have ordered and that is the machine that I carefully >>studied on the intel side when making the decision... I had some crafty >>benchmark data from that machine as well as the recent 2800+ AMD and I made >>the decision based on performance and nothing else... And before you ask, >>(a) I didn't do the benchmark tests and (b) I can't reveal the people that >>did them, but you can probably at least figure out who they work for... Same >>for my Itanium2 numbers that I have mentioned in the past... But that doesn't >>mean the numbers are "phoney". Whether the athlon 2800+ dual is faster >>for another program doesn't matter to me since I work on Crafty. I'm not >>even sure when the 2800+ will be actually available, but it doesn't matter >>in the context of what I am doing anyway... >> >> >> >>> >>>If you bothered to do any testing yourself you'll find out the P4 isn't what >>>it's cracked up to be. It's odd to me that you have such a blind confidence in >>>the P4. You haven't bothered doing any extensive testing on one and you haven't >>>messed with an Athlon yet you 'feel' the P4 is faster. Why is that? >>>I can try to believe a Pinto is faster than a Ferrari but it's just not going >>>happen. :) >>> >> >>I haven't tested on AMD, you are correct. But I _have_ tested on a dual PIV >>as I said before. And I will have a dual 2.8 here pretty soon (I hope, if >>the holidays don't cause delays). >> >>I don't "feel" anything is faster than anything else. I rely on numbers posted >>by people here that I trust. Eugene is one, as I doubt anyone knows more about >>the internals of both processors (from an assembly/optimizing/etc point of view) >>than he does. So when he posts numbers I listen. And his dual PIV numbers >>were better than the dual AMD numbers. Others also have the PIV on top of the >>heap, including Tom's page of course... >> >>However, if someone has a dual AMD 2600+ (or whatever they are going to call >>the chip after admitting their "+" numbers were too optimistic) running linux, >>we can compare speeds once my machine arrives. That's the way I tend to settle >>speed disputes. Run the test... If I thought AMD was faster, I'd have bought >>a dual AMD. If I could have afforded a dual/quad itanium-2 I would have bought >>that as nothing else is close. If Cray built something in this price range I >>would have gone that way as well. I try to "follow the performance" and don't >>have any vendor-specific preferences at all, although I had a couple of very bad >>AMD experiences in the K5 days, and that did leave a sour taste... But if AMD >>can take the performance lead, I'd buy. But so far, they've been unable to >>even think about quad machines... while Intel has had 'em for at least 5+ >>years. That means they will _always_ be behind by a factor of nearly two or >>so... >> >> >> >> >>> >>> >>>>>This just proves once again Tom is a bad reviewer. There's already proof out he >>>>>faked a Pentium4 review and I have proof he hinders Athlons (by setting their >>>>>bios settings to the absolute lowest setting). The KT333 set to "turbo" has an >>>>>efficiency of 94%, this means out of 2.7GB/s max theoretical bandwidth at >>>>>166fsb(333DDR) it does 2538MB/s. Since mine is CL2.5 I'm only running 90% >>>>>efficiency (2430mb/s). With Nforce2 boards and running 200+ fsb you can get over >>>>>3gb/s on an Athlon easily. Same with the Epox 8K9A2. >>>> >>>> >>>>Again, that is _still_ slow compared to the PIV xeon numbers with (say) the >>>>E7501 chipset... >>>> >>>> >>>> >>>>> >>>>>Tom however is managing a horrible 74%. You can do better than this even setting >>>>>the settings to 'normal'. He has disabled other options to get the benchmarks >>>>>THAT low. 26% below regular 166fsb(333DDR) scores is pretty bad. >>>>>Even your average KT266a/333/400 board at 133fsb gets about 2gb/s. >>>>> >>>>>About the faked review he did.. it's the "Hot contraband P4-3.6GHz". He used a >>>>>Pentium4 2GHz CPU and cut&pasted some of the 3's to make it look like a 3.3GHz >>>>>P4. The picture is STILL on his page. Just go look at that article and save the >>>>>large P4 picture. Go into an image editing program and go near the bottom where >>>>>it says "80531PC3.3G0K". Drag a selection box around the second three in that >>>>>line and copy it, invert the colors and then drag that over the other three. You >>>>>will see a black box where the colors match EXACTLY. This was a cut & paste job. >>>>> >>>>>Tom isn't the only corrupt reviewer, either. >>>> >>>> >>>>I'm not going to debate that. Tom's hardware isn't the only place that is >>>>doing these measurements. There are some well-known academics that are >>>>interested in various benchmarks like streams, linpack, etc... > >You say, "I had some crafty benchmark data from that machine as well as the >recent 2800+ AMD and I made the decision based on performance and nothing >else...". The only way the 2800+ is going to be any slower is if the binary it >ran had the 1.4x speedup problem. Email your source and have them test with one >of my binaries. You could also take the dual 2800+ results and multiply them by >1.214 (21.4% diff between 1.7 and 1.4). > >Also you say the + numbers are optimistic.. How so? My 1900+ at 1.8GHz (2200+) >is on par with a P4-3GHz in Crafty. A 2600+ (Stock) could beat the P4-3.25GHz in >the results list I have AND have room to spare. I don't say that. AMD said that. And they are scaling their numbers back. They were using a raw estimate that every 66mhz increase in clock speed gave them 100mhz of effective speed increase compared to the PIV. They discovered this failed at the beyond 2000 mhz clock frequencies and was "optimistic". Again, I assume they know what they are doing since they make the things... >By the way, AMD is currently working on Quad Opteron systems and there are test >boards available. I don't know if you remember what I said in a previous message >(not in this thread) but Opterons are going to have dual-channel DDR and memory >banks and that dedicated bandwidth PER CPU. >So if you have four CPU's and you're running 200MHz(400DDR) memory w/ dual >channels thats 6.4gb/s per cpu. Total you'd have 25.6gb/s bandwidth. Oh, don't >forget the CPU has it's own memory controller. Right... and how are you going to read data out of memory where each bank has 100+ns latency? I know how supercomputers do it. I don't think you are going to see 4-port memory banks in a machine at that price point... Theoretical max is a nice concept. Attainable thruput is more interesting. At present, Intel seems to be leading that list by a significant margin. Whether they will in the future or not I don't know, as I try to not evaluate "vaporware". But for the moment, they are on top in the bandwidth war... as Linpack and other normal memory-intensive applications show time after time. I hope they can deliver a quad opteron for a resonable price. They were talking about quad K7's two years ago and not a single instance has shown up yet. Intel talked about the 8-way boxes a while back and delivered a kludge there, using a "fusion" chipset to tie two 4-way clusters of processors together into a single 8-way box, but with terrible memory performance. They tried to offset that by only offering 2M L2 caches, but that drove the price up and didn't help memory-bound large applications at all... I hope the quad opterons don't end up in never-never land as the 8-way boxes did... If I recall, the 4=way dual 2.0ghz xeon is the fastest PC-class machine around right now, by a wide margin. And the heavier the load placed on it, the wider that gap becomes...
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.