ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) (02/15/90)
I don't believe this question has been addressed yet. In a recent PC Magazine,
they did a survey of about 30 SX machines. Just for the record, the machines
that won awards were the AST, the Gateway, and the Zeos. The AST has
16k of cache, the Gateway has 70 nanosecond rams, and the Zeos 60 nanoseconds.
No doubt, these played a major role in system throughput.
But one machine was different in that it forces the SX to 20 Mhz. It was
called a Citrus something or other. This was it's main distinguishing feature.
It implies that one can tweak other boards to get this kind of performance.
1.) Does anyone know how to do this?
2.) Is it worth the risk with overheating the components?
3.) BTW, WHY does the componentry heat up?
Thank you.
--
"Nobody had the guts...until now..."
|-$anjay "lock [+] on" $ingh ssingh@watserv1.waterloo.edu N.A.R.C. ]I[-|
"A modern-day warrior, mean mean stride, today's Tom Sawyer, mean mean pride."
!being!mind!self!cogsci!AI!think!nerve!parallel!cybernetix!chaos!fractal!info!ralf@b.gp.cs.cmu.edu (Ralf Brown) (02/15/90)
In article <1109@watserv1.waterloo.edu> ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) writes: }3.) BTW, WHY does the componentry heat up? The following is true for CMOS, and partially true for TTL devices: A CMOS switching element has two transistors, one between the supply voltage and output, and one between the output and ground, like so: power supply | | _____|/ | |\ | | input ----+ +------- output | | |_____|/ |\ | | ground The two transistors of the switching element have opposite polarities, so that one is off when the other is on. Due to this architecture, the current drain at rest is very low (only enough to drive the gates of the switching elements connected to the output). However, when the input switches from one state to the other, that switch is not instantaneous, so for a brief period of time, both transistors are partially on. We now have what amounts to a momentary short between power and ground. The duration of this short depends on how cleanly the input switches states, which is relatively independent of clock speed. However, the clock speed does influence how often the short occurs, so a higher clock speed results in more momentary shorts (of the same duration), causing a higher current draw, which heats up the chip. -- {backbone}!cs.cmu.edu!ralf ARPA: RALF@CS.CMU.EDU FIDO: Ralf Brown 1:129/46 BITnet: RALF%CS.CMU.EDU@CMUCCVMA AT&Tnet: (412)268-3053 (school) FAX: ask DISCLAIMER? | _How_to_Prove_It_ by Dana Angluin 24. by appeal to intuition: What's that?| Cloud-shaped drawings frequently help here.
cms2839@ultb.isc.rit.edu (C.M. Stuntz) (02/16/90)
In article <1109@watserv1.waterloo.edu> ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) writes: >I don't believe this question has been addressed yet. In a recent PC Magazine, >they did a survey of about 30 SX machines. Just for the record, the machines >that won awards were the AST, the Gateway, and the Zeos. The AST has that's funny . i've got the issue right here , and it says that the Editor's Choices were the AST , the _Austin_ , and the Zeos ... -a.stranger -- @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ @ "Imagination keeps the shadows away - Xymox @ @~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~@ @ a.stranger - CMS2839%RITVAX.BITNET@cunyvm.cuny.edu @
mlord@bnr-rsc.UUCP (Mark Lord) (02/16/90)
In article <1109@watserv1.waterloo.edu> ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) writes: >But one machine was different in that it forces the SX to 20 Mhz. It was >.. >1.) Does anyone know how to do this? >2.) Is it worth the risk with overheating the components? >3.) BTW, WHY does the componentry heat up? Many places are doing this. My own SX runs at 18Mhz (failed the 20Mhz test). To do this, simply change the 32Mhz master crystal to something bigger, like 36Mhz for 18Mhz operation, or 40Mhz for 20Mhz operation. If you get lucky, the board will happen to work at the new speed. No all boards do. Like I said, mine failed the 20Mhz attempt, but works more or less fine at 18Mhz (no noticeable problems due to clock speed). The higher frequency will tend to make the components heat up more than they otherwise would, since the gates will be switching states more frequently. Each state change draws power.. more state changes, more power.. more heat. IMHO, the heat should not be a big problem. Sure, running hotter will make things wear out sooner, but odds are good that it'll still take many years. The big trick is simply whether or not it works at all at the new speed. -- ______Mark S. Lord______________________ ______________________________ | ..uunet!bnrgate!carrsc!mlord | These are only MY opinions. | | or: bnr-rsc!mlord@bnrgate | I charge for official views. | |________________________________________|______________________________|
poffen@sj.ate.slb.com (Russ Poffenberger) (02/22/90)
In article <7982@pt.cs.cmu.edu> ralf@b.gp.cs.cmu.edu (Ralf Brown) writes: >In article <1109@watserv1.waterloo.edu> ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) writes: >}3.) BTW, WHY does the componentry heat up? > >The following is true for CMOS, and partially true for TTL devices: > > A CMOS switching element has two transistors, one between the > supply voltage and output, and one between the output and ground, > like so: > > power supply > | > | > _____|/ > | |\ > | | > input ----+ +------- output > | | > |_____|/ > |\ > | > | > ground > >The two transistors of the switching element have opposite polarities, so that >one is off when the other is on. Due to this architecture, the current drain >at rest is very low (only enough to drive the gates of the switching elements >connected to the output). However, when the input switches from one state to >the other, that switch is not instantaneous, so for a brief period of time, >both transistors are partially on. We now have what amounts to a momentary >short between power and ground. The duration of this short depends on how >cleanly the input switches states, which is relatively independent of clock >speed. However, the clock speed does influence how often the short occurs, >so a higher clock speed results in more momentary shorts (of the same >duration), causing a higher current draw, which heats up the chip. This is incorrect. Actually the reason why more current is drawn is because the output stage is driving a capacitive load at the input of the following stage. This causes the capacitance to constantly charge to an opposite state when it switches, meaning the output stage must drive current to discharge the capacitance, then more to charge it to the opposite polarity. The faster (more often) this is done, results in more current draw from the supply and resulting heat generation. Russ Poffenberger DOMAIN: poffen@sj.ate.slb.com Schlumberger Technologies UUCP: {uunet,decwrl,amdahl}!sjsca4!poffen 1601 Technology Drive CIS: 72401,276 San Jose, Ca. 95110 (408)437-5254
cs4g6ag@maccs.dcss.mcmaster.ca (Stephen M. Dunn) (02/27/90)
In article <1109@watserv1.waterloo.edu> ssingh@watserv1.waterloo.edu ($anjay "lock-on" $ingh - Indy Studies) writes:
$But one machine was different in that it forces the SX to 20 Mhz. It was
$called a Citrus something or other. This was it's main distinguishing feature.
$It implies that one can tweak other boards to get this kind of performance.
Not advised.
$1.) Does anyone know how to do this?
Well, you don't just go in and plug in a faster crystal. The various
glue chips have to be able to handle whatever speed you're running at, and
also the timing for the expansion bus may be derived from the main
crystal. You want to keep the expansion bus running at 8 MHz since many
cards won't run any faster than this, and if it's run off the same clock
then you're out of luck.
I've heard that microprocessors are actually supposed to be able to run
at up to 1.5x the rated clock speed. This doesn't make much sense to me,
but that's what I've heard ... it may be wrong. But pushing a 16 MHz
part to 20 MHz shouldn't be a problem, given that the rest of the design
was based on the higher speed.
$2.) Is it worth the risk with overheating the components?
If you provide the box with adequate cooling, the parts don't get
hot enough to worry about ... this is what was done with the first few
33 MHz 386 machines, back when 25 MHz was the fastest chip available.
$3.) BTW, WHY does the componentry heat up?
I think the heat is generated when a gate switches states, so by
increasing the clock rate 25%, you also increase the rate at which the
gates switch by 25% and increase the heat production similarly. Of course,
this applies to the same chip run at two different speeds. If you're
comparing two chips fabbed with different scales (e.g. a 1.0 micron
chip against a 1.5), even at the same speed they'll dissipate different
amounts of heat.
--
Stephen M. Dunn cs4g6ag@maccs.dcss.mcmaster.ca
<std_disclaimer.h> = "\nI'm only an undergraduate!!!\n";
****************************************************************************
I Think I'm Going Bald - Caress of Steel, Rushphil@pepsi.amd.com (Phil Ngai) (02/28/90)
In article <25E9799E.5768@maccs.dcss.mcmaster.ca> cs4g6ag@maccs.dcss.mcmaster.ca (Stephen M. Dunn) writes: | I've heard that microprocessors are actually supposed to be able to run |at up to 1.5x the rated clock speed. This doesn't make much sense to me, |but that's what I've heard ... it may be wrong. But pushing a 16 MHz Let me assure you that this is indeed nonsense. It is true that a reputable semiconductor manufacturer will include some margin to allow for the accuracy of the IC testers used, but the amount of the margin depends on many factors and for a microprocessor is very unlikely to be 1.5. You have to agree that if there was margin, they would sell as much of it as they could. Why can't people accept that if a device is rated at a particular speed, then they should not exceed that speed? -- Phil Ngai, phil@amd.com {uunet,decwrl,ucbvax}!amdcad!phil A PC without DESQview is like Unix without ^Z.
alex@bilver.UUCP (Alex Matulich) (03/01/90)
In article <29306@amdcad.AMD.COM> phil@pepsi.AMD.COM (Phil Ngai) writes: >| I've heard that microprocessors are actually supposed to be able to run >|at up to 1.5x the rated clock speed. This doesn't make much sense to me, >Let me assure you that this is indeed nonsense. It is true that a >reputable semiconductor manufacturer will include some margin to allow >...etc... Just for general information -- someone from Motorola once told me that the margin for a 68010 is about 20%. He said that if you plan to push it that far, it's best to use the kind that comes in a ceramic case instead of bakelite. The reason is that the problems associated with pushing the clock speed arise from overheating, and the ceramic chips dissipate heat a little better than the bakelite ones. I know a few Amiga owners who successfully pushed 12 MHz 68010's to 14 MHz, so I guess it can really be done. -- /// Alex Matulich /// Unicorn Research Corp, 4621 N Landmark Dr, Orlando, FL 32817 \\\/// alex@bilver.UUCP ...uunet!tarpit!bilver!alex \XX/ From BitNet use: bilver!alex@uunet.uu.net