Author: Matt Taylor
Date: 08:00:35 12/17/02
Go up one level in this thread
On December 17, 2002 at 09:30:07, Bob Durrett wrote: >On December 17, 2002 at 04:44:55, Matt Taylor wrote: > >>On December 17, 2002 at 01:56:26, Jesper Antonsson wrote: >> >>>On December 16, 2002 at 19:55:47, Matt Taylor wrote: >>> >>>>On December 16, 2002 at 18:35:21, Jesper Antonsson wrote: >>>> >>>>>On December 16, 2002 at 18:18:19, Alessandro Damiani wrote: >>>>> >>>>>>On December 16, 2002 at 17:42:08, Jesper Antonsson wrote: >>>>>> >>>>>>>On December 16, 2002 at 07:48:59, Sally Weltrop wrote: >>>>>>>>On December 16, 2002 at 04:52:40, Jorge Pichard wrote: >>>>>>>>>Intel could produce a microprocessor capable of 10 Ghz very soon, but they >>>>>>>>>simply won't because they have to profit gradually from 3Ghz to 4 Ghz to 5Ghz >>>>>>>>>etc... Simply the Murphy LAW is a profitable marketing strategy that has worked >>>>>>>>>gradually, if they make the mistake of producing a microprocessor capable of 10 >>>>>>>>>Ghz in the next 6 months they will lose a lot of money, by NOT squeezing our >>>>>>>>>pockets every six months as they have done for the last 20 years. >>>>>>>> >>>>>>>>excellent point. >>>>>>> >>>>>>>Sorry, no, it's far from excellent, actually the reasoning is invalid. Intel has >>>>>>>competition, and that means that if they could churn out faster processors (at >>>>>>>reasonable costs) but doesn't, AMD or someone else would do it and take over >>>>>>>Intels market shares and profits. >>>>>>> >>>>>>>/Jesper >>>>>> >>>>>>Right, if Intel and AMD don't make an agreement to go for the smaller steps to >>>>>>make more profit together. This would be against law. Great law! ;) >>>>>> >>>>>>Alessandro >>>>> >>>>>Yes. :-) And also, for such a conspiracy not to be obvious (and get police >>>>>attention), IBM, Compaq, Sun etc has to be in on it too, because it would look >>>>>peculiar if PowerPCs, StrongARMs, Alphas and SPARCs suddenly turned up at 10 >>>>>GHz, but the same didn't happen on the x86 market. >>>>> >>>>>/Jesper >>>> >>>>No it wouldn't. >>> >>>Yes it would. >>> >>>>MHz is only relevant within a design. There exist 10 GHz >>>>processors right now. >>> >>>Any of them one of the processors I mentioned? Otherwise your point is moot. >> >>Sparc is considerably different from Intel. I have done assembly for both >>machines. Sparc, Alpha, and PPC are RISC chips. RISC designs can reach higher >>clock frequencies than CISC chips since they don't have to do as much per cycle. >> >>Hearsay has it that Alpha is dead and Apple may as well be. I would be suprised >>if any of the processors you listed showed up at 10 GHz even when Intel and AMD >>are competing with 10 GHz processors, assuming they can scale that far. >> >>>>Also, consider Itanium which runs around 800 MHz. Seems >>>>slow? It outperforms a Pentium 4 3.06 GHz. IBM has a 125 MHz processor which >>>>will outperform ~1 GHz x86 chips. AthlonXP 2700 runs at 2.17 GHz and is roughly >>>>as fast as a Pentium 4 2.8 GHz. >>> >>>Yes. >>> >>>> >>>>Whether or not AMD and Intel have a private agreement is speculation; however, I >>>>find it more than curious that they release new chips at the same time. >>> >>>That kind of adaptions I don't find curious at all. >>> >>>>Also >>>>interesting is that the P4 3.06 GHz is the same silicon that my roommate has >>>>inside his Pentium 4 1.8 GHz chip. >>> >>>No, it's not the same. It may seem the same and have the same function, and even >>>have the same design, but most likely, the 3 GHz chip is produced with better >>>quality (purity, exactness and so on). >>> >>>>Bottom line: Intel was able run 3 GHz last year, but they've waited a year to >>>>release it to the general public. There's something to speculate about. >>> >>>They weren't able to run at 3 GHz reliably and in quantity last year. The >>>fine-tune production processes until the yield and quality goes up. Last year >>>they couldn't have churned out a lot of 3 GHz processors. >>> >>>/Jesper >> >>You sure? A friend of mine clocked a 1.5 GHz Williamette to 2.5 GHz. Some even >>clocked to 3 GHz on Williamette. I've heard of earlier Palomino-core AthlonXPs >>clocking BEYOND the rates AMD will stamp on them now. Another friend of mine >>religiously babbled about clocking his Thunderbird 1 GHz chip to 1.7 GHz (read: >>AthlonXP 2000 equivalency) when 1.4 GHz Thunderbirds were the best you could >>buy. He could have debatably gone further if he had more multipliers to work >>with. >> >>Additionally, my roommate has clocked his P4 1.8 GHz to 2.4 GHz, but his pc2100 >>ram couldn't sustain 177 MHz. I also know someone who is running an AthlonXP >>around 2.6 GHz. When he bought the chip, the fastest available was AthlonXP 2400 >>(2 GHz). You still can't buy a 2.6 GHz AthlonXP. Hmm... >> >>How is it that AMD and Intel have such problems clocking their chips high when >>nobody else has problems clocking them that high? >> >>-Matt > >An electrical engineer's perspective: > >One must consider more than what's discussed above. A good design is good in >all aspects and not just a select few. In particular, the product should not be >provided to the public until adequate life testing has been performed and it is >demonstrated that the product will perform RELIABLY over the advertised life >with the worst-case conditions. Also, ample margin of safety must be verified, >preferably by test, before the product is put on the market. In the case of >microprocessors, the thermal considerations are very important. > >It would be an economic disaster to put the product on the market only to see >dozens of lawsuits and thousands of returned processors. Admittedly safety is a concern. It seems to me that they have fairly standard techniques for lowering thermal output. Whenever it would seem that they've hit a thermal brick wall, the temperature fairy comes along and saves the day. Additionally, they increase requirements for heatsinks and such. Besides, I find it difficult to believe that it took Intel 6 months to test an increase of 300 MHz on the Pentium 4 Northwood. Again, is it sheer coincidence that Intel and AMD announce newer iterations of their processors within a week or so of each other? Obviously they each know a -lot- about what the other company is doing. They used to both announce price cuts at the same times, too. Intel slackened their cuts while AMD accelerated them to compete in a branded, economically-depressed market. It also strikes me as interesting that AMD had samples of Clawhammer in 2001, yet it won't be available until almost June of 2003. They've put on shows with quad-Opterons since June or so, and since then they've decided to delay Opteron almost six months? Initially they planned to launch in 2H '02 which would have put Clawhammers in Christmas PCs. This would be very lucrative, and one would think that AMD would want that market. Why did they miss their deadline? One explanation is performance. Maybe they couldn't get the clock rates up? Nope. AMD is secretive about details, but someone managed to run the OpenSSL benchmark. Opteron was scoring near the top. Perhaps yields? Well, they had alpha silicon two years ago. One would think that by now they've got it working and in mass production, particularly after the hype machine took over. They've been talking about their silicon on insulator process for a while, too. Once you start hyping, you'd better be prepared to deliver or die. Perhaps thermal considerations? Clawhammer and Opteron include a heat spreader to make it easier to correctly install a heatsink and harder to crack. The CPU itself is really just an Athlon with some extra registers, more complex decoders, a faster multiply unit, 64-bit GPRs, and an integrated memory controller. They might also delay Clawhammer if they couldn't get its SMP-capable sibling Opteron working efficiently in SMP. An unlikely proposition since they've been displaying quad-Opteron systems for something like six months and dual-Opterons for much longer. All evidence points toward Opteron being ready to ship (except for mass produced units), yet AMD is pushing the deadline back further rather than pushing it forward as you'd expect. If there was no evidence, I wouldn't think about it. If there was a little circumstantial evidence, I would think twice. If there was a lot of circumstantial evidence, I would think more than twice.
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.