sunyat@pawl.rpi.edu (Eric Hsiao) (10/04/90)
Hi,
I'm in the market of a 386sx motherboard, and I need to know some info.
First of all, this guy wants to offer a 386sx 16Mhz with a fast crystal
installed to yield 20Mhz (he says Norton SI is 21). Is this a safe thing
to do? Will it burn out the board? He wants to charge $400 for it (without
any memory I think). Is that a fair price or is it too high? I think he
wants around $80 for 1 meg of 80ns Simms. Is that a good deal?
Please E-mail me.
Eric
--
| Internet: Eric_Hsiao@MTS.RPI.EDU > Read GIF News - a freely copied news < |
| sunyat@PAWL.RPI.EDU > magazine in .GIF file format. Color < |
| Bitnet: userGJ8L@rpitsmts.bitnet > articles on reviews of programs, < |
| Download GIF News (617)354-8873 > games, industry news, IBM, Amiga, etc.< | dank@eng.umd.edu (Daniel R. Kuespert) (10/04/90)
In article <Z~R%'N|@rpi.edu> sunyat@pawl.rpi.edu (Eric Hsiao) writes: > I'm in the market of a 386sx motherboard, and I need to know some info. >First of all, this guy wants to offer a 386sx 16Mhz with a fast crystal >installed to yield 20Mhz (he says Norton SI is 21). Is this a safe thing >to do? It is not safe. While you might get lucky and get a chip that's actually capable of 20MHz operation but just isn't certified for it, more likely you'll be replacing the SX within a few hours. Unless the chip is actually specified for 20MHz, it probably can't dissipate the heat generated by the faster speed very well and will consequently fail. As you've worded it, this loser is trying to load you up with a motherboard that's also running above design speed. What goes for the SX chip also goes for the other chips on the board, so you'll probably have trouble with random failures. -- Daniel R. Kuespert | Disclaimer: Chemical Process Systems Lab | University of Maryland, College Park | "I lied; it's a vice I have." dank@eng.umd.edu |
tim@ziggurat.gg.caltech.edu (Tim Kay) (10/05/90)
>> I'm in the market of a 386sx motherboard, and I need to know some info. >>First of all, this guy wants to offer a 386sx 16Mhz with a fast crystal >>installed to yield 20Mhz (he says Norton SI is 21). Is this a safe thing >>to do? >It is not safe. While you might get lucky and get a chip that's >actually capable of 20MHz operation but just isn't certified for it, >more likely you'll be replacing the SX within a few hours. Unless the >chip is actually specified for 20MHz, it probably can't dissipate the >heat generated by the faster speed very well and will consequently >fail. As you've worded it, this loser is trying to load you up with a >motherboard that's also running above design speed. What goes for the >SX chip also goes for the other chips on the board, so you'll probably >have trouble with random failures. This is incorrect. Typically, there is absolutely no difference between the fast and slow parts that a manufacturer ships, *except* that the faster parts happen to work at the faster speed. After Intel manufactures each 386 chip, their tester determines how fast it can run reliably and stamps it accordingly. The only difference is that certain parts of the chip layout turn out to be marginal, and some of the chips manage to squeak past the tester, due to many factors including process variations, while others don't. The chips will generate roughly the same amount of heat regardless of their stamped speed rating. As part of tuning the process to produce faster chips (on average), they might redesign certain marginal parts of the mask or process. This tuning of a process slowly increases the mean speed of the chips coming off the line. The question of whether a given chip will run at a faster rate than it is stamped must really be broken into two separate cases. If you are trying to push the chip faster than the fastest chip that a manufacturer "makes," you are in case I. If you are just trying to save money by buying a less expensive chip, you are in case II. Case I: What do you suppose Intel does with a chip that passes not only the 33 MHz test but also passes the 40 MHz test? (I would guess they run the 40 MHz test and the 50 MHz test so that they can measure how well the line is maturing.) But they haven't announced a 40 MHz part, so they have to sell it stamped for 33 MHz. This is why you *can* screen a bunch of 33 MHz chips and find a few that will run just fine at a higher clock rate. The only suspicious part of it is that the screeners test equipment isn't as good as Intel's. But the fact that the chip is stamped 33 MHz doesn't mean that the chip failed Intel's faster tests. Case II: The question of whether a 16 MHz part will run at 20 MHz when the manufacturer is shipping 20 MHz parts is a different question. If a chip passed the 20 MHz test, the manufacturer would probably then sell it as a 20 MHz part. Therefore, if a chip is stamped 16 MHz, it probably failed the manufacturer's 20 MHz tests. However, manufacturer's tests are harsher than most environments that the chips end up in eventually. It is just possible that the circumstance in which the 16 MHz chip failed when tested at 20 MHz will never occur in the particular application. Also, what happens when a manufacturer's line has matured to the point that most of the chips are coming out at at least 20 MHz. Do they stop selling 16 MHz parts? Eventually they do, but for marketing reasons, they might choose to stamp some 20 MHz-capable parts as 16 MHz. Otherwise, they would have to force 16 MHz customers to buy the more expensive 20 MHz parts, or they would have to lower the price of the 20 MHz parts for the entire customer base. There is no reason that they manufacturer's process will mature at exactly the same rate as the demand for faster chips. -------- I have very successfully "pushed" chips. I ran a one wait state AT clone at zero wait states for about two years. I then replaced the motherboard with a 20 MHz 386 motherboard, but the 386 was rated at only 16 MHz. It has been running fine for the last two years. I bought the 386 motherboard when 20 MHz 386 chips were relatively new, so this 386 might qualify as a case I chip. However, let me say that I am not suggesting that you go out and save money in this way. You don't want to mess with these things unless you are capable and willing to deal with potential problems that might occur. Many people might attribute certain hardware problems to software or visa verse. The only way to point the finger in the correct direction is to run the software (*all* the software) on a different machine. If you aren't willing to chase down these loose ends, go with a more conservative system. Tim
james@bigtex.cactus.org (James Van Artsdalen) (10/05/90)
In <Z~R%'N|@rpi.edu>, sunyat@pawl.rpi.edu (Eric Hsiao) wrote: > [...] this guy wants to offer a 386sx 16Mhz with a fast crystal > installed to yield 20Mhz (he says Norton SI is 21). Is this a safe > thing to do? Not really. Intel started shipping 20MHz 386SXs by sorting the original 16MHz line, but went to a different line for 20MHz in volume (presumably different process - 1 micron?). The shortages in 20MHz 386SXs suggests to me that (1) the speed-sorting of 16MHz parts wasn't yielding well and (2) the new line didn't yield well either (or perhaps the new line didn't come up at all). > Will it burn out the board? Nope. But it might be real flaky. After the machine heats up, watch out. -- James R. Van Artsdalen james@bigtex.cactus.org "Live Free or Die" Dell Computer Co 9505 Arboretum Blvd Austin TX 78759 512-338-8789
james@bigtex.cactus.org (James Van Artsdalen) (10/05/90)
In <tim.655078682@ziggurat>, tim@ziggurat.gg.caltech.edu (Tim Kay) wrote: ] this guy wants to offer a 386sx 16Mhz with a fast crystal installed ] to yield 20Mhz (he says Norton SI is 21). Is this a safe thing to do? | It is not safe. While you might get lucky [...] > This is incorrect. Typically, there is absolutely no difference between > the fast and slow parts that a manufacturer ships, *except* that the faster > parts happen to work at the faster speed. No, it was correct. The word "typically" was a good choice. > After Intel manufactures each 386 chip, their tester determines how > fast it can run reliably and stamps it accordingly. [...] There are two different processes for the 386, the original 1.5 micron and a newer 1 micron. I think the 1 micron has been shipping since last summer? In any case, all 33MHz parts are the new process, and all 20MHz parts are the old process. 25MHz parts can come from either process depending on what the motherboard design requires (the parts aren't quite the same). A 1micron 25MHz part might run at 33MHz, but a 1.5micron is a lot less likely I think. As for the SX, it is quite possible that the 20MHz parts are a different process entirely than the 16MHz. Certainly the shortages of 20MHz SXs several months ago suggest Intel wasn't having much luck sorting 16MHz parts up to 20MHz. Clearly you can get lucky, but sometimes maybe not. -- James R. Van Artsdalen james@bigtex.cactus.org "Live Free or Die" Dell Computer Co 9505 Arboretum Blvd Austin TX 78759 512-338-8789
davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (10/08/90)
Okay, let me see if I can clarify this, since a lot of people are
posting parts of the answer.
When chip are available in several sppeds, there are two cases: (1)
the chips are a diferent process, or (2) the faster chips are selected
from the same line as the slower.
If the chips are case (1), using a slower part is a bad idea for two
reasons, it is very unlikely to work reliably, and it may overheat. As
the clock speed goes up the power disipation goes up (not linearly) due
to driving the internal capacitance (and some other stuff). Damage to
the chip is unlikely, but as it gets hotter it changes some internal
timings, and it may not give correct answers.
If the chips are case (2), the manufacturer will test all chips at the
higher speed and label the ones that pass FAST (whatever that means for
that chip). Then the ones which fail are tested at the slower speed, and
if they pass they are labeled SLOW, or they are not labeled at all.
However, as the type (2) process improves, the line will produce more
FAST chips than are needed. Therefore chips in excess of demand are only
tested SLOW.
When using a 16MHz chip at 20MHz, you may be using
- a chip of another process, not designed for the speed
- a chip which failed at the higher speed
- a chip not tested at the higher speed
* which is okay
* which would fail is tested FAST
I have regularly pushed 16MHz 386DX parts to 20MHz, and 286 from 6->8
MHz. In the case of the 386SX, I believe that the 20MHz part is another
process, and it is not a good idea, although Intel was getting their
20MHz parts by picking the 16 (I'm told) for a while, and while the
yield was not good, the parts didn't run overly hot.
Note that I have not told you what to do, but at least this may clear
the air. You really have to know if the situation is type (1) or (2),
then make the decision. This is why the spole saying "don't ever do it"
and "that's how Intel does it" are both partially right.
--
bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen)
sysop *IX BBS and Public Access UNIX
moderator of comp.binaries.ibm.pc and 80386 mailing list
"Stupidity, like virtue, is its own reward" -me