home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ogicse!qiclab!nosun!klic!keithl
- From: keithl@klic.rain.com (Keith Lofstrom)
- Newsgroups: comp.sys.ibm.pc.hardware
- Subject: Re: Run a 486/33MHz CPU at 40 or 50MHz?
- Message-ID: <1993Jan12.201913.19407@klic.rain.com>
- Date: 12 Jan 93 20:19:13 GMT
- Article-I.D.: klic.1993Jan12.201913.19407
- References: <1868@ddbeano.Dundee.NCR.COM> <1993Jan12.132709.17786@physchem.ox.ac.uk>
- Organization: Keith Lofstrom Integrated Circuits
- Lines: 50
-
- > I have a 486DX/33MHz system, and I have been told by the manufacturer
- > that the motherboard is upgradeable to 50MHz by plugging in a new
- > crystal and CPU.
- >
- > Question:
- > Is it possible to plug in (say) a 40MHz or even a 50MHz crystal
- > into the board *without* replacing the old (33MHz) CPU?
-
- Integrated circuits can WEAR OUT.
-
- There are two types of mechanisms to worry about - heat induced degradation
- and hot carrier injection. The first problem results in hard failures,
- and can be avoided by keeping the chip cool. Hot carrier injection occurs
- at all temperatures - and is enhanced a bit by cooling the chip - and acts
- to make the chip slower with time.
-
- What this means to you is that a chip that works up to, say, 55 MHz this
- year may only work up to 45 MHz a year from now. Intel may have marked
- that chip as 33 MHz, so that the part would keep working for a decade at
- its rated speed. Intel has a powerful incentive to sell those chips at
- as fast a speed rating as they can get away with. If a batch will safely
- go at 50 MHz over its lifetime, they gain no advantage by marking them as
- 33 MHz. If they mark them as too fast, then have to recall them, they
- lose a lot of money.
-
- Heat also slows chips down, by about 0.5% per degree C. If Intel rated
- the chip at 55C, and you can guarantee the chip stays below 35C at all
- times, you've bought an extra 10% speed margin. However, if you tweek
- the machine to go as fast as it can go this winter at 18C ambient, and
- this summer it gets up to 38C *outside the case*, you are going to have
- a miserable summer. Most chips run too hot anyway. Put a fan on it.
-
- Lastly, consider how you test the modified system . I don't know if
- anybody has developed a test program for PCs that checks all the
- potential data pattern sensitivities for all the different peripheral
- combinations and chipsets - it would be a difficult problem. I'm
- pretty sure Checkit and QAPlus find many potential errors, but I
- doubt they find all the speed related ones.
-
- The bottom line? Don't do it if you count on the machine for anything.
- Go ahead and do it if you enjoy testing the hell out of it and will
- report the results of your extensive tests to the rest of us.
- Please, no "I increased my clock xx% and it hasn't crashed for days"
- stories - I can make a cursor blink with unreliable hardware, too.
-
- Keith
- --
- Keith Lofstrom keithl@klic.rain.com Voice (503)-520-1993
- KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon"
- Design Contracting in Bipolar and CMOS - Analog, Digital, and Power ICs
-