tom@ssd.csd.harris.com (Tom Horsley) (11/27/90)
I see a lot of the current and near current top of the line RISC chips and
other microprocessors coming out with a 33MHz clock rate. In a similar way
the previously most popular number seemed to be 25MHz.
Is there some hardware related magic about certain clock rates that makes it
likely for totally different manufacturers of totally different chips to
come out with chips with the same clock rates in similar time frames?
Obviously they all need to stay competitive, but why do they lots of them
wind up at exactly 33? Why not 30 or 35?
--
======================================================================
domain: tahorsley@csd.harris.com USMail: Tom Horsley
uucp: ...!uunet!hcx1!tahorsley 511 Kingbird Circle
Delray Beach, FL 33444
+==== Censorship is the only form of Obscenity ======================+
| (Wait, I forgot government tobacco subsidies...) |
+====================================================================+
wunder@orac.HP.COM (Walter Underwood) (11/28/90)
I see a lot of the current and near current top of the line RISC chips and other microprocessors coming out with a 33MHz clock rate. In a similar way the previously most popular number seemed to be 25MHz. Get out your slide rule and find the L and D scales. As the exponent goes through 0.0, 0.1, 0.2, ..., 10**x gives this series: 1.0 1.25 1.6 2.0 2.5 3.2 4.0 5.0 6.3 (or 6.4 if you want it to be 2x 3.2) 8.0 10.0 Do these clock speeds sound familiar: 4, 6, 8, 10, 12.5, 16, 20, 25? Sounds like 33 MHz was an attempt at one-upmanship, and everyone caught on. One other common place to see this series is in ANSI film speeds. wunder
shair@ux1.cso.uiuc.edu (Bob Shair) (11/28/90)
tom@ssd.csd.harris.com (Tom Horsley) writes: >Is there some hardware related magic about certain clock rates that makes it >likely for totally different manufacturers of totally different chips to >come out with chips with the same clock rates in similar time frames? >Obviously they all need to stay competitive, but why do they lots of them >wind up at exactly 33? Why not 30 or 35? Actually the magic looks to be in the human mind. For the same reason that events are scheduled to start at 8PM rather than at 8:17, designers set out to build chips with 50ns, 40ns, 30ns cycle times. The new IBM RISC 6000 model 550, at 41.6MHz or 24ns, may be something of an exception. I'll be interested to see whether anyone else uses that rate. Come to think of it, the 540 also runs at a somewhat unusual rate of 30MHz. -- Bob Shair shair@chgvmic1.iinus1.ibm.com Scientific Computing Specialist SHAIR@UIUCVMD (bitnet) IBM Champaign
tim@proton.amd.com (Tim Olson) (11/28/90)
In article <TOM.90Nov27065615@hcx2.ssd.csd.harris.com> tom@ssd.csd.harris.com (Tom Horsley) writes: | I see a lot of the current and near current top of the line RISC chips and | other microprocessors coming out with a 33MHz clock rate. In a similar way | the previously most popular number seemed to be 25MHz. | | Is there some hardware related magic about certain clock rates that makes it | likely for totally different manufacturers of totally different chips to | come out with chips with the same clock rates in similar time frames? | Obviously they all need to stay competitive, but why do they lots of them | wind up at exactly 33? Why not 30 or 35? It's mainly that designers think in terms of clock period (ns) rather than frequency in MHz, and 33MHz (really 33.3333...) is a 30ns cycle, just as 25MHz was a 40ns cycle. 40ns -> 30ns is a 25% reduction in cycle time. -- -- Tim Olson Advanced Micro Devices (tim@amd.com)
gillies@m.cs.uiuc.edu (11/28/90)
Possibly the harmonic numbers multiplied by 100: CPU Speed Change Speed of Instruction Fetching 1/6 = 16Mhz 1/5 = 20Mhz 20% speed increase over 16Mhz 50ns/ins 1/4 = 25Mhz 25% speed increase over 20Mhz 40ns/ins 1/3 = 33Mhz 33% speed increase over 25Mhz 30ns/ins 1/2 = 50Mhz 50% speed increase over 33Mhz 20ns/ins Don't ask me where 40Mhz came from. this pattern suggests that the next generation of chips beyond 50Mhz will run at 100Mhz! 8-) Maybe this spectrum is something that marketing figured out. Or maybe it is the optimal space-time set of speed test-points for a CPU chip tester. Or maybe it has something to do with the availability of high-speed DRAM and SRAM.
mahesh@caradhras.cc.nd.edu (Mahesh Subramanya) (11/28/90)
In article <1990Nov27.192504.11191@ux1.cso.uiuc.edu>, shair@ux1.cso.uiuc.edu (Bob Shair) writes: >The new IBM RISC 6000 model 550, at 41.6MHz or 24ns, may be >something of an exception. I'll be interested to see whether >anyone else uses that rate. From what I've heard, the reason behind that was that it was originally spec'd at 40 Mhz, but they ran into some race conditions. So they basically tweaked the clock around till the race cond. disappeared. The number *happened* to be 41.6. Apart from the fact that 41.6 is my second favourite number (8-)), there is nothing else really exceptional about it in terms of clock speeds. -- ************************************************************************ Mahesh Subramanya INTERNET: mahesh@caradhras.cc.nd.edu Senior Analyst Office of University Computing NeXT: mahesh@numenor.next.nd.edu University of Notre Dame Voice: (219) 239-5600 x6421 Notre Dame, IN 46556 ************************************************************************
weaver (11/28/90)
In article <TOM.90Nov27065615@hcx2.ssd.csd.harris.com> tom@ssd.csd.harris.com (Tom Horsley) writes: >I see a lot of the current and near current top of the line RISC chips and >other microprocessors coming out with a 33MHz clock rate. In a similar way >the previously most popular number seemed to be 25MHz. > >Is there some hardware related magic about certain clock rates that makes it >likely for totally different manufacturers of totally different chips to >come out with chips with the same clock rates in similar time frames? >Obviously they all need to stay competitive, but why do they lots of them There is no hardware magic about 33 vs 32, and so forth. This is competitive marketing. When you manufacture chips, the max speed for the individual chips makes a bell curve, with small numbers of very slow and very fast chips, and most in the middle. If the middle of the curve is (say) 30 Mhz, you make your chip 33 Mhz by testing for speed and throwing out any that are slower than 33 (or more likely retesting them for 25 Mhz). Of course if the middle of the curve is too low, you won't get any at the desired speed. The magic number is actually the approximate 20% improvement in speed, which probably has to do with how much speed improvement is necessary to make a difference to buyers of the chip.
salim@mozart.amd.com (Salim Shah) (11/28/90)
These numbers like 25MHz and 33Mhz when transformed to time-period round off to nice whole numbers. Nothing magical about it. Salim.
koopman@a.gp.cs.cmu.edu (Philip Koopman) (11/28/90)
Many times, the frequencies are picked based on standard crystal clock oscillator frequencies. It's real handy for system designers to be able to get an oscillator off-the-shelf instead of special-ordering them (especially for prototypes). These oscillators tend to have even-MHz values (there are one or two notable exceptions). This is the difference between system-level thinking and component- level thinking... Phil Koopman koopman@greyhound.ece.cmu.edu Arpanet 2525A Wexford Run Rd. Wexford, PA 15090 *** this space for rent ***
merriman@ccavax.camb.com (11/29/90)
In article <1990Nov28.010513.17760@news.nd.edu>, mahesh@caradhras.cc.nd.edu (Mahesh Subramanya) writes: > In article <1990Nov27.192504.11191@ux1.cso.uiuc.edu>, > shair@ux1.cso.uiuc.edu (Bob Shair) writes: >>The new IBM RISC 6000 model 550, at 41.6MHz or 24ns, may be >>something of an exception. I'll be interested to see whether >>anyone else uses that rate. > > From what I've heard, the reason behind that was that it was originally spec'd > at 40 Mhz, but they ran into some race conditions. So they basically tweaked > the clock around till the race cond. disappeared. The number *happened* to > be 41.6. Apart from the fact that 41.6 is my second favourite number (8-)), there > is nothing else really exceptional about it in terms of clock speeds. > -- I once heard the story that a particular, rather odd slow-speed asynch circuit bit rate (56.86 bps comes to mind) came to be because the designers were aiming for a somewhat higher speed, but found the mechanics in the printers could not keep up, so they selected the next set of gears from the Boston Gear catalog (or whatever) that would fit without too much rework, and what happened happened. > ************************************************************************ > Mahesh Subramanya INTERNET: mahesh@caradhras.cc.nd.edu > Senior Analyst > Office of University Computing NeXT: mahesh@numenor.next.nd.edu > University of Notre Dame Voice: (219) 239-5600 x6421 > Notre Dame, IN 46556 > ************************************************************************ George Merriman, Cambridge Computer Associates - New York
caveh@csl.sri.com (Caveh Jalali) (11/29/90)
33MHz is usually 33.33333 MHz. if you take the inverse to find the cycle time, we can see that it does follow a pattern: MHz cycle time (ns) 10 100 12.5 80 16.67 60 20 50 25 40 33.3 30 40 25 50 20 66.67 15 the pattern sort of makes sense in the ns column. the point of all this is a that the speed rating is pretty much arbitrary from the point of the view of manufacturers such as motorola and intel. they have no idea what your requirements are, thus they can't fine tune their speed ratings to allow you the best performance. they basically have to pick arbitrary norms at which they test their components for pass/fail. if a chip passes, it means they will require certain timing constraints that must be met in order for that chip to operate correctly, and conversely, the chip will satisfy certain timing specifications. often, a 25MHz processor is a 33MHz processor which failed the 33MHz test, so they really do have the freedom to pick arbitrary speed ratings. if you're a sufficiently large enough customer, you will suddenly see things like 15MHz 68020's altho i don't think moto ever sold those over the counter. basically, they don't want to test these chips too many times, so they might test at 33MHz, if it fails test at 25MHz, if it fails test for 20MHz, if it fails ther, it's junk. they could test for 24, 23, 22.. MHz, but i guess that's not feasible. if you develop your own processor, you have the freedom to choose your test clock frequency. that's why MIPS and IBM get to use oddball frequencies. -- 00c -- caveh@csl.sri.com "X is not a letter, it's a sentence."
mahar@jetsun.weitek.COM (Mike Mahar) (11/30/90)
In article <TOM.90Nov27065615@hcx2.ssd.csd.harris.com> tom@ssd.csd.harris.com (Tom Horsley) writes: >I see a lot of the current and near current top of the line RISC chips and >other microprocessors coming out with a 33MHz clock rate. In a similar way >the previously most popular number seemed to be 25MHz. 10Mhz = 100ns 20Mhz = 50ns 25Mhz = 40ns 33Mhz = 30ns 40Mhz = 25ns 50Mhz = 20ns 67Mhz = 15ns Most Chip designers pick their next clock based on nano-seconds per cycle rather than Mhz. -- "The bug is in the package somewhere". | Mike Mahar - Anyone who has used Ada | UUCP: {turtlevax, cae780}!weitek!mahar
jrs@hpfcso.HP.COM (John Spencer) (11/30/90)
The maximum clock frequency for the Intel EISA chips is 33MHz.
ndoduc@framentec.fr (Nhuan Doduc) (11/30/90)
ok, nothing magic about 33 (or really, 33.33) which is 1/30 or 25 = 1/40 or 20 = 1/50 ... but it seems that there is a reason for 4.77MHz on the good old PC : the "usual" frequency for a certain communication port is 15.31MHz, thus some other magic values: 7.16 and 9.54MHz that you may found on 8088 machines (but, *et nous sommes encore dans les valeurs magiques*, most marketing sheets and even many tech spec doc still refer to 8MHz (instead of 7.16) and 10 (instead of 9.54) --nh Nhuan DODUC, Framentec-Cognitech, Paris, France, ndoduc@framentec.fr or ndoduc@cognitech.fr, Association Francaise des Utilisateurs d'Unix, France, doduc@afuu.fr
clc5q@shamash.cs.Virginia.EDU (Clark L. Coleman) (12/01/90)
In article <1554@ftc.framentec.fr> ndoduc@framentec.fr (Nhuan Doduc) writes: >ok, nothing magic about 33 (or really, 33.33) which is 1/30 or 25 = 1/40 or >20 = 1/50 ... but it seems that there is a reason for 4.77MHz on the good old >PC : the "usual" frequency for a certain communication port is 15.31MHz, thus >some other magic values: 7.16 and 9.54MHz that you may [find] on 8088 machines I guess I can't figure it out myself, so please tell me the relationships among the following numbers : 4.77 MHz 7.16 MHz 9.54 MHz 15.31 MHz The first three have the relationship as 2,3 and 4 times 2.385 MHz, so that is no problem. Deriving them all from 15.31 MHz is tough. So I assume you meant 14.31 MHz. So, what was the famous communications port that worked at 14.31 MHz, just for my own curiosity ? ----------------------------------------------------------------------------- "We cannot talk of freedom unless we have private property." -- Gavriil Popov, Mayor of Moscow, September 11, 1990. ||| clc5q@virginia.edu
caveh@csl.sri.com (Caveh Jalali) (12/02/90)
In article <1990Dec1.152532.16888@murdoch.acc.Virginia.EDU> clc5q@shamash.cs.Virginia.EDU (Clark L. Coleman) writes: >In article <1554@ftc.framentec.fr> ndoduc@framentec.fr (Nhuan Doduc) writes: >... >The first three have the relationship as 2,3 and 4 times 2.385 MHz, so that >is no problem. Deriving them all from 15.31 MHz is tough. So I assume you meant >14.31 MHz. > >So, what was the famous communications port that worked at 14.31 MHz, just for >my own curiosity ? 14.31 MHz probably has nothing to do with communication ports (necessarity). that's 4x the colorburst frequency. aparently, the colorburst frequency was a very popular choice in the early computer days because it was easy (cheap) to get that crystal, not to mention the fact that it could be used for other purposes simultaneously. i think it was used in color TVs or something to separate the color signal from the rest of the signal. so, if your computer generated a color video signal, you only needed one crystal to generate the signal as well as clock the CPU. -- 00c -- caveh@csl.sri.com "X is not a letter, it's a sentence."
ge@wn3.sci.kun.nl (Ge' Weijers) (12/03/90)
wunder@orac.HP.COM (Walter Underwood) writes: > I see a lot of the current and near current top of the line RISC chips and > other microprocessors coming out with a 33MHz clock rate. In a similar way > the previously most popular number seemed to be 25MHz. >Get out your slide rule and find the L and D scales. As the exponent >goes through 0.0, 0.1, 0.2, ..., 10**x gives this series: >One other common place to see this series is in ANSI film speeds. That's a bit different: Common film speeds differ by factors of 2^(1/3). This used to be known as 1 DIN. The ISO scales are: (ISO = ASA + "/" + DIN) ASA 100 = ISO 100/21 (append a degree symbol) ASA 125 = ISO 125/22 (100 * 2^(1/3) = 125.9921....) ASA 160 = ISO 160/23 (100 * 2^(2/3) = 158.74.....) ASA 64 = ISO 64/19 Film speeds come in third octaves. (My Nikon actually gives 32 seconds exposure when set to 30 seconds. All exposure times are rounded from their actual values) Ge' Weijers -- Ge' Weijers Internet/UUCP: ge@cs.kun.nl Faculty of Mathematics and Computer Science, (uunet.uu.net!cs.kun.nl!ge) University of Nijmegen, Toernooiveld 1 6525 ED Nijmegen, the Netherlands tel. +3180652483 (UTC-2)
herrickd@iccgcc.decnet.ab.com (12/03/90)
In article <1554@ftc.framentec.fr>, ndoduc@framentec.fr (Nhuan Doduc) writes: > ok, nothing magic about 33 (or really, 33.33) which is 1/30 or 25 = 1/40 or > 20 = 1/50 ... but it seems that there is a reason for 4.77MHz on the good old > PC : the "usual" frequency for a certain communication port is 15.31MHz, thus > some other magic values: 7.16 and 9.54MHz that you may found on 8088 machines > (but, *et nous sommes encore dans les valeurs magiques*, most marketing sheets > and even many tech spec doc still refer to 8MHz (instead of 7.16) and 10 > (instead of 9.54) > > --nh > Nhuan DODUC, > Framentec-Cognitech, Paris, France, ndoduc@framentec.fr or ndoduc@cognitech.fr, > Association Francaise des Utilisateurs d'Unix, France, doduc@afuu.fr You are too generous with the 4.77 MHz. It gets used, at least on the original (1981) IBM PC Motherboard (should we say Great GrandMotherboard?) for a crucial frequency in producing the imitation NTSC signal to go to an RF Modulator and be transmitted to your home television receiver. The color was messed up on mine (I can't remember the symptom) and when I phoned and complained, they said "bring it in". While I waited, they tweaked the main processor clock oscillator to match the picture signal to my TV receiver's front end. I was flabbergasted. dan herrick herrickd@astro.pc.ab.com
weaver (12/05/90)
In article <6852@uceng.UC.EDU> dmocsny@minerva.che.uc.edu (Daniel Mocsny) writes: > >I.e., will a manufacturer ever rate parts at 25 MHz which could run >perfectly well at 33 MHz? Yes. For a high demand chip like a 486, I would not expect this to happen. But at my previous employer, we were selling low cost, 8-bit microcontrollers. In one instance, a chip was sold in two speed grades, and the yield for fast parts was significantly better than expected. The supply of fast chips greater than demand, and the supply of slow chips less than demand. The company chose to mark fast parts as slow rather than re-negotiate contracts which had been signed months in advance of production. In the longer term, price adjustments should keep manufacturers from having to do this.
daveh@cbmvax.commodore.com (Dave Haynie) (01/08/91)
In article <1990Nov28.021703.28668@weitek.COM> weaver@weitek.UUCP (Michael Weaver) writes: >In article <TOM.90Nov27065615@hcx2.ssd.csd.harris.com> tom@ssd.csd.harris.com (Tom Horsley) writes: >>I see a lot of the current and near current top of the line RISC chips and >>other microprocessors coming out with a 33MHz clock rate. In a similar way >>the previously most popular number seemed to be 25MHz. >There is no hardware magic about 33 vs 32, and so forth. Well, there is a little hardware magic. If you look at popular clock speeds, and the clock periods that go with them, all is revealed: Popular Speed Actual Speed Clock Cycle 16MHz 16.67MHz 60ns 20MHz 20.00MHz 50ns 25MHz 25.00MHz 40ns 33MHz 33.33MHz 30ns 40MHz 40.00MHz 20ns 50MHz 50.00MHz 15ns Basically, it would seem that chip makers like to give chip testers, designers, etc. (who of course work in ns, not MHz), nice, even numbers to work with, whenever possible. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "Don't worry, 'bout a thing. 'Cause every little thing, gonna be alright" -Bob Marley
john@IASTATE.EDU (Hascall John Paul) (01/09/91)
In article <17213@cbmvax.commodore.com>, daveh@cbmvax.commodore.com (Dave Haynie) writes: > Popular Speed Actual Speed Clock Cycle > 16MHz 16.67MHz 60ns > 20MHz 20.00MHz 50ns > 25MHz 25.00MHz 40ns > 33MHz 33.33MHz 30ns > 40MHz 40.00MHz 20ns \__ not in this > 50MHz 50.00MHz 15ns / reality! :-) 40 MHz == 25 ns, 50 MHz == 20 ns, 66.67 MHz == 15 ns -- John Hascall An ill-chosen word is the fool's messenger. Project Vincent Iowa State University Computation Center john@iastate.edu Ames, IA 50010 (515) 294-9551