Author: Robert Hyatt
Date: 17:42:34 02/14/00
Go up one level in this thread
On February 14, 2000 at 18:20:04, Peter Kasinski wrote: >On February 14, 2000 at 17:52:54, Robert Hyatt wrote: > >>On February 14, 2000 at 15:47:37, Dave Gomboc wrote: >> >>>On February 14, 2000 at 13:55:27, Robert Hyatt wrote: >>> >>>>On February 14, 2000 at 12:53:27, Tom Kerrigan wrote: >>>> >>>>>From what I've read, AMD is fully supporting the Kryotech machine, to the point >>>>>of providing a warranty for the processor. This tells me that they are pretty >>>>>serious about stability; so they probably wouldn't be sending Kryotech >>>>>processors that just barely work at 750MHz. I won't comment on the situation >>>>>with Digital because I don't know anything about it. >>>>> >>>>>-Tom >>>> >>>>DEC did the same thing... although I didn't ask about the financial details (I >>>>am sure Kryo pays DEC more for a chip to be overclocked than others pay for a >>>>stock chip, just to cover the warranty issues). I have no idea how they sort >>>>chips nowadays. Intel obviously sorts parts into X, Y and Z clock bins, for >>>>some devices. Maybe AMD has a 750+ bin they sort parts into for the Kryo guys? >>>>who knows. >>>> >>>>However I still think that the gains are from raising the core cpu voltage >>>>more than from trying to lower the core cpu temperature... >>> >>>If AMD already has a demonstration 1.1GHz, does this mean Kryotech could use it >>>to put together, say, a 1.4 or 1.5Ghz box? >>> >>>Dave >> >> >>I assume they will, in fact. Although I have not followed them to see if they >>did this with newer alphas, for example. I think the concept of 'overclocking' >>is a very risky one. The vendors know what voltage level and clock speed the >>things will handle reliably. Going above and beyond is not something I would >>do on a machine that I wanted to be reliable. > > >Bob, > >vendors have to cater to the lowest common denominator. I.e. also to people who >later call support lines complaining about broken coffee cup holders (one >particularly well publicized case of misuse of a CD tray). Intel makes CPU?s >that work _RELIABLY_ even if owners keep their PC?s in greenhouses in Mombassa. > >Over-clocking (like over-anything) can get out of hand pretty quickly. But well >documented experiments show that, for example, Celeron chips can run at 75MHz >bus speed (instead of 66 MHz = 13 % gain) with _NO SIDE EFFECTS_. > >Should your answer be that such use decreases chip?s life from 30 to 15 years, >then well, I think you would be guilty of over-estimating the significance of >such assertion. > >Regards, >PK You need to subscribe to a couple of the linux mailing lists. You will discover that there are _huge_ problems with overclocking _any_ chip. They work perfectly (or seem to) until you hit that one cpu instruction that has one extra gate in the path. And you get bogus answers. The linux lists are _full_ of horror stories about overclocking, to the point that the developers won't even talk to anyone running an overclocked cpu, until they return it to factory spec and still produce the reported bug. Generally going back to spec cures things instantly. Feel free to do so. But I don't buy the "it is ok" stuff. Sure Intel seems to down-clock some chips for supply/demand reasons, and yes they will usually overclock back to their original spec just fine. But we are talking about _real_ overclocking here, and that is bad, in general. IMHO of course, but also in the HO of _many_ linux kernel guys/developers as well...
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.