Author: Robert Hyatt
Date: 13:23:40 12/02/02
Go up one level in this thread
On December 02, 2002 at 15:36:50, Sune Fischer wrote: >On December 02, 2002 at 13:12:39, Robert Hyatt wrote: > >>On December 02, 2002 at 12:20:46, Sune Fischer wrote: >> >>> >>>>Second, I don't see anything related in the two. A photograph has _no_ >>>>technical content. IE what can you tell by looking at the top of one of my >>>>xeon 700's? Absolutely nothing. >>> >>>He is sloppy, about pictures, data etc... >> >>Have you never seen an article about a new product, where a picture was >>unavailable? >>I have seen this _many_ times. Both with computers and with other things... > >Where, at AMD or at Intel? Neither. But there are many reasons why it could happen. I gave some. Eugene might have hit the nail on the head however, namely that it was an attempt to conceal the "ID" of the processor that someone was "leaking" past a ND agreement. >Of course not, and in such a case there would be a remark about the picture not >being authentic. Maybe or maybe not. If a "forgery" is obvious enough, I might not say anything about it knowing that everyone could see that it was "doctored"... But I'm not a mind reader and don't know what went on there and I won't try to speculate without _some_ information to go on... > >>>People have been saying for a long time they are having touble reproducing the >>>extemely low numbers he is getting for AMD systems. Either he is incompetent or >>>biased, the site is hard to take seriously in any case. >> >>The only evidence I would take for such a claim is for someone to take two exact >>machines >>such as the 2.8 xeon with RDRAM and an athlon 2800+ and compare them carefully >>and >>show numbers that are better than his. > >I recently got to test my program on a P4 2.53 GHz, it ran like a 1550 MHz >Athlon. So if you extrapolate the Intel to 3.00 and the AMD to 2.00 you will >find the AMD to be the faster by an even larger margin. Possibly, if you believe clock for clock scales linearly, which I don't believe for any processor family. But, again, I am talking about dual vs dual in my comments, and so far, the dual intel has delivered a higher NPS than a dual AMD, when running Crafty. I don't care about other applications if Crafty is what I want to run... > >I think a 4-way Opteron with hypertransport will knock the socks off anything >Intel has, but let's wait and see :) that's the point. It won't knock anything off _today_. And who knows what Intel will answer with. They have not been leading the field by lack of planning for future improvements... > >> without resorting to unsafe tricks such >>as overclocking >>or anything else. >>I have basically refused to answer problem questions about crafty when the cpu >>is overclocked >>in any way. I am not sure _any_ non-engineer understands the overclocking >>issues, and just >>because "it worked on my test" doesn't mean it works for _all_ tests. And it >>doesn't even mean >>it worked correctly for _that_ test. If a particular circuit takes 400 >>picoseconds to settle, >>and you give it 350, it might settle some of the time, none of the time, or all >>of the time, >>depending on luck. Yet people say "it works fine" which may or may not be true >>for _all_cases. >>I've seen too many cases of failure to even think about the overclocking issue. > >Of course, I'm not talking about overclocking here, just optimizing the bios for >interleaving, latency and latest drivers, basic stuff that isn't default package >settings. > >>So standard machines, compared as done on THG, could be used to discredit the >>numbers >>posted there. But not "hey, my AMD does better after I ...." Because 99.99999% >>of the >>people run 'em right out of the box... Including me.. > >I strongly suspect Tom's is bought by Intel, there is a lot of circumstancial >evidence if you know where to look. > >>>Speaking of being scientific, I don't understand how you can form an opinion >>>when you haven't even tried them or "know anyone with an AMD machine"? >> >>Simple. I know Eugene quite well. He's an extremely reliable source. Crafty >>is >>in the SPEC suite so I have talked to reps from _every_ chip manufacturer over >>the >>past couple of years and know the results. I don't have any AMD machines that I >>can run on, and I don't know anybody _here_ that has AMDs handy for me to test >>on. But I have gotten plenty of results from all over. Enough to say with >>pretty >>good confidence "AMD and Intel are close on the 2.8 xeon vs 2800+ athlon when >>running Crafty on a single CPU, but Intel leads the way when a dual-cpu machine >>is used." Remember that in the last WMCCC event Crafty participated on, the >>machine was a dual AMD, because that is what the operator had. I had plenty of >>comparison numbers back then, on a daily basis. >> > >I hope you are not implying the people here are faking the results? ;) > >-S. Nope. I suspect that some numbers are wrong. Nothing more. But in the case of Crafty it is easy to test as I'll have a dual 2.8 this month (I hope). If someone gets/has a dual athlon 2800+ then we can compare apples to apples using the same O/S and compiler... It would be interesting to do it in a carefully controlled way rather than with lots of yelling and shouting.
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.