Slashdot Mirror
Pentium III 1.13: Tops For Speed, 'F' For Price?
fjordboy writes: "CPUscorecard has some multiple reviews and benchmarks of recent CPUs. Somewhat surprisingly, the Pentium III 1.13 GHz processor tops the list, but only for speed. CPU Scorecard also gave the Pentium III an "F" when it came to pricing. Is the high price tag worth it for the top-notch speed? Click here to see how the Pentium III stacked up to the other CPU's."
A 500 MHz G4 is twelve percent slower than a 1.1 GHz Pentium III? If that's true, then that's the real story here. Isn't this what everyone was trying to vehemently deny a while back, that a G4 is equal to a Pentium of double the clock speed?
That's right, this site gives the impression
that P3 1.13 Ghz is the fastest CPU in the industry.
In fact, both Alpha and U-Sparc-III are TWICE faster.
Well, no. By far the most credible cross-platform CPU benchmark is SPEC; if you had bothered to check the SPEC CPU2000 scores you would find that a 1GHz P3 is precisely as fast as a 900MHz US-III on SPECint2000_base (438) and only 30% slower in SPECfp2000_base (327 vs. 427). Assuming linear scaling (not generally a great idea but close enough in this case), a 1.13GHz P3 will be 13% faster than the USIII-900 in SPECint, and 16% slower in SPECfp. In other words, they will be essentially equal.
Now, the Alpha actually is quite a bit faster than the P3, with SPECint/fp scores of 514/591 for an 833MHz chip. That makes it 17%/80% faster than our 1GHz P3, and an estimated 4%/60% faster than a 1.13GHz P3.
Now, on the other hand, both the Alpha and the US-III systems tested cost many times more than the i840 1GHz P3 system; not only do the chips cost a good deal more, but they get the benefit of much faster (and more expensive) buses to memory, etc. This makes quite a difference even in the SPEC CPU tests, and if the chips could somehow be placed on equivalent platforms, the P3 would easily win SPECint outright, and might be rather competitive even on SPECfp. (In case you were wondering, the reason the P3 sucks at SPECfp is because it is saddled with the register-starved x87 floating-point implementation for backwards compatability reasons; the P4 will go quite a ways towards solving this problem--at least as far as newly compiled code goes--with its SSE2 instructions.)
On the third hand, as has been pointed out, the 1.13 GHz P3 does not exist, and never did. (Intel "launched" what amounted to several engineering samples which turned out not to work properly anyways. Other than a couple dozen sent off to review sites and OEMs for validation, no 1.13 GHz P3's ever left the company.)
On the fourth hand, a chip which does currently exist in much higher quantities, the 1.5 GHz P4, looks like it will quite forcefully take the SPECint crown away from Alpha, and depending on Intel's progress in optimizing their compilers for SSE2, might even take the SPECfp crown as well. We'll get to find out when it is officially released in about a month or so.
On the fifth and final hand, though, the real advantage of the Alpha and US-III is their platforms, which give them much greater i/o throughput--often more important than CPU power for server applications anyways--and allow them to scale to configurations of 32 and 64 CPUs and beyond; Intel has a long way to go to compete on these measures.
When buying new hardware you need to look at two price points and choose in the middle. There is the point when the price takes off to catch those who will pay through the nose to get the 'best' (only to realize in two months that something better came out), and the other price point is where they catch those who try to be cheap (yea, it's the cheapest, but for $5 more you can get something twice as big/fast). Those who go the cheap way get disappointed and lose money on upgrades.
When I buy a new machine I mentally graph out cost vs. performance. Usually you will see a slow rise in price and performance takes off. Then it will level out somewhere in the middle and then about 2 generations from the top the price will skyrocket while the performance just creeps along. I buy at the point just before where the price takes off. This way I get very good hardware that only a little behind the best, and it doesn't cost much more then the junk I would get if I was cheap. I can then use the money on peripherals (a good monitor is better than a 50 MHz jump any day).
I have a second rule of computer buying: When you own a computer you need to put $XXX into it every year. The more you put into it the closer to cutting edge it will be. If you put $1000 into you machine every year you will have the best machine on the block. If you only put $50 or less into the machine you will be facing obsolesce. Custom PC's don't become obsolete, they only become obsolete through neglect. They are more like a car than a VCR.
Yes, I know my two rules sound obvious to any geek, but there is wisdom in there. This is the same spiel I give to anybody who comes to me for advice in buying a computer.
I've never known it to make sense to pay top dollar for Intel's fastest CPU. From a value perspective it just doesn't make sense, especially with SMP boards so cheap. The CPUs one, two or even three steps down from top of the line running in pairs often deliver *better* performance than the top CPU and at a lower cost(YMMV, IANAL, etc etc). Even in single-CPU installations the price differences are so great and the performance advantages so minimal between Intel's #1 CPU and the lower-clocked family members that the marginal gain doesn't exceed the marginal cost.
It would only seem to be of value in those rare situations where the marginal gain in processing is so profitable or desirable that it outweighs the marginal costs -- but places like that probably already buy bigger machines than Intel boxes.
The thing is, its always been this way. I remember buying a P150 when P200 was out and the price difference was astronomical. To this day, when I use my P150 (running @ 166) I still don't feel the performance difference is substantial over a P200.
Since the Pentium !!! 1.13GHz has been withdrawn, its price seems pretty irrelevant.
The pricing of the Pentium 4 is much more interesting...
--
This comment was brought to you by And Clover.
the speed of a chip in Mhz means nothing, when comparing two different types of chips. People are never going to realize this though, they have to start looking at mips and flops to get the low down on how the chip performs...
The program they make the speed tests with is Intel's iCOMP. Is it any wonder that Intel's CPUs come top?
Does my bum look big in this?
It's always nice to see the chip at the top of the list doesn't work.
-atrowe: Card-carrying Mensa member. I have no toleranse for stupidity.
Look at your history. Any time in the past 8 years that x86 has come close to Alpha, something happens where Alpha again leapfrogs in speed. That's not about to change.
.25um to a hybrid .18/.25um process. The only change we can expect there is a higher clock speed--around 1 GHz by January. A full .18um EV68 will ship by maybe March or April; that ought to hit 1.2-1.3 GHz and might include an on-die L2. Still, we're safe assuming that the EV68 will scale linearly with clock speed at best, with perhaps a slight bump from the on-die L2 in March or April.
.13um process, will have much lower power, voltage and cooling requirements than Willamette. This sort of thing happens with every new Intel core: the first implementation is big, hot and power hungry and is aimed at a niche workstation market; then they do a die-shrink and move it into the mainstream. Thus the fact that P4 systems will cost around $2500 (*not* $4000) at introduction really doesn't mean anything. Even at $4000 they'd still be a full order of magnitude cheaper than an 833MHz Alpha system.
Look at your history. x86 actually did briefly leapfrog the Alpha in SPECint95 when the PPro was first released; Alpha took the crown back with the EV6. In any case, the PPro was Intel's only new core introduction of the past 8 years! Thus, judging from our exactly and precisely 1 datapoint, we can conclusively predict that the P4 will indeed take the SPEC crown away from Alpha for a short while.
I mean, come on--this is engineering, not history. You're not going to get anywhere with a teleological theory of CPU performance over the last 8 years--that's just ridiculous. We don't need to guess or play "history" here; there is plenty of solid evidence about both Intel and Compaq's upcoming designs, how they will perform, and when and at what speeds they will be released. It is a known fact that the P4 is going to be released on November 20 at speeds of 1.4 and 1.5 GHz. It is also a known fact that the Alpha looks like it is stuck at 833 MHz until the release of the EV68 die shrink. It is moreover a known fact that the EV68 is behind schedule and that it will *not* be released before November 20.
Next, we can look at the again well known design specs of the P4 and EV68. The initial EV68s are essentially just a process shrink of the EV67 from
The P4, on the other hand, is a completely new core, full of some really pretty impressive design features. From the point of view of SPEC, the most important are the 3.2GB/s FSB, trace cache, 2-cycle data L1, and larger reorder buffers. In addition, the half-clock-latency ADD will be a tremendous help for much of SPECint and the double-precision SSE2 instructions may allow the P4 to be the first x86 chip to compete or even win on SPECfp. (The much maligned 20-stage pipeline is largely compensated for by the improved larger branch predictors.) In other words, it is very likely that the P4 will achieve higher IPC on SPECint than did the P3. Oh, and it runs at 1.5 GHz.
So, assuming that the P4 will have the same SPECint IPC as P3 (an assumption generous to your argument) and the EV68 the same as the EV67, Compaq would need to release a 1.066 GHz Alpha before the 1.5 GHz P4 (i.e. in a month). Assuming the more likely occurrence that the P4 achieves ~20% better IPC on SPECint, Compaq will need a 1.266 or 1.3 GHz Alpha to win. Very very doubtful. Looking ahead, it doesn't appear that the EV68 will be able to keep up with the P4 in ramping clock speeds over the next 18 months, either; while the lead will likely go back and forth, it will probably be Intel, not Compaq, with the SPECint lead the lion's share of the time. This is all up until the release of the EV8, which looks to kick some serious butt. Note that I'm not claiming the P4 is a better design than the EV6x, just that Intel has access to better fab process than Alpha.
In any case, as you see, this is something that can be analyzed with known facts, engineering principles, and informed industry predictions, not with some blind appeal to (false) generalizations made over an entire 8 year period. In other words, "Alpha always wins SPEC" is not quite on a par with "never start a land war in Russia."
I'd like to see a P4 using LESS power, as small a power supply, and in the form factors that Alpha can get in today. Not to mention price. It'll be interesting to see all the same people who bitch about Alpha not having this or or that or being too expensive when they see P4 systems at $4000 using non-standard power supplies and cases.
LOL! The P4 *does* use less power than an Alpha; around 50W vs. an astounding 100W for an 833MHz EV67. And as for form factors, I truly have no idea what the hell you're talking about. Yes, the initial implementation of the P4 requires a somewhat ridiculous heatsink compared to the typical x86. That's alright, because the initial implementation of the P4 is being positioned essentially like the original PPro was. The P4 won't become Intel's mainstream chip until around 9 months from now, with the Northwood revision and die-shrink. Northwood, being made on a
You think the Alpha can beat the P4 in form factor?? Uh...which do you think is going to come out first, an EV68-based laptop or a P4-based laptop?? Or howabout this: an EV68-based tablet computer or one with a P4?? Sun can make a stab at offering embedded CPUs, but I've never heard of anyone even considering an embedded Alpha. Why?? Well...power and form factor constraints, obviously. This is absurd.
The FUD is so thick, you can cut it with a knife.
You can say that again. Luckily, come Nov. 20 the FUD will be cleared away, for better or for worse. Now, I'm not arguing that the P4 or even Foster (the "P4 Xeon") will be able to replace Alpha for most of its markets. But it looks as if after years of failing to take advantage of their one clear selling point--unambiguous SPEC superiority--the Alpha is going to lose even that. Here's hoping Compaq finally decides to pour the resources necessary into making this excellent architecture prosper as it deserves. But let's not pretend that Intel is standing still just because the Alpha has been.