Slashdot Mirror
Tom's Looks At The New P-III
Choady McGee writes "Tom's Hardware just posted a review of a new version of the pentium iii. From what they say, it looks like it could threaten the pentium iv and because of that, may not be released any time soon."
But I'm one of the Mac slashdotters so.. um.. nevermind :)
2002: It's the ..er.. Pentium III rev2! Kniht. Eww. Oy...
A PII clocked at the same speed as a PPro was slower (remember how www.x86.org got in trouble for publishing those benchmarks? ) was hotter than a PPro, etc... then people realised that PIIs were becoming eventually faster and cheaper than PPros, and PPros got phased-out (I think retired early by Intel to force people to buy PIIs is closer to reality).
Conclusion? Don't buy the latest and greatest processors in their early incarnation because besides the hype, they don't run that much faster than the previous generation... but eventually will.
---
"Hasta la victoria siempre!" El Comandante
I can second this with the same board. Do not buy a used KA-7 Board for an Athlon.
\\\ SLUDGE
I tried replying to this article, but my computer crashed as I did. No lie.
Anyways, I was saying that I have upgraded this month to the newest version. Failures are still occurring. I'm glad you've had better luck.
\\\ SLUDGE
Yes, but last time Intel had problems AMD wasn't breathing down their neck. While Intel's Pentium 60 was bad, AMD's offerings were worse. That is not the case anymore.
Another major difference is that this time there is a general softness in the PC market, and the best selling PCs are not the $2000 fire-breathers, but the sub $1000 value PCs (where AMD has done remarkably well).
It certainly is true that Intel still has Dell in the bag that could change at a moments notice. If Intel were to have a bad recall now Dell would switch in a moment. They would have little choice.
The fact that Intel has billions in the bank is nice, but nobody seriously thinks that they are likely to go out of business anytime soon. What is far more likely is that Intel will not be able to keep up the revenue growth that has pushed its stock price into the sky.
And when it comes to stock prices, the two for one deal you mentioned certainly can't help things. I don't even know if what you say is true, but if Intel is giving two for one deals then one of two things must be true. Either Intel is purchasing the second stock with some of their billions (basically a stock buy back) or they are diluting the value of the stock that their investors already own. Neither of these scenarios are good (in the long run).
You see, basically what Intel is saying (if this is true) is that their stock is really worth half of what it is currently listed (otherwise they wouldn't be giving stock away).
LOL. That's hilarious. I thought that sounded like a weird deal.
Intel is no longer in a position where they can spend too much time worrying about internal competition, but they have to worry about AMD which has been trouncing them lately. If the new PIII can let them regain some ground they've lost against AMD then you can be guaranteed that they will push it to the market as fast as they can (remember the original P3 1Ghz?).
Hear hear. I started reading this, then closed the shit out of disgust and came back looking for the thread bashing the writing. Thank you, sir for this opportunity to vent!
There's a strong rumor that IBM can make G3s with clock rates that meet or exceed 2 GHz. But as it stands, Motorola is the only PPC manufacturer that can/wants to make G4s -- which are stuck at 733 Mhz. So the overall platform stagnates because Apple doesn't want its low-end systems (based on the G3) blowing the doors off the high-end systems (based on the G4). The situation is getting out of hand... it sucks when marketing concerns get in the way of performance computing.
Just raise the taxes on crack.
11.0010010000111111011010101000100010000101101000
Did anyone else read the article and find the on going metaphor for the chip as a baby both retarded, AND retarded?
It's fine when tech people try to make their articles more interesting, but please, just drop the first paragraph cuteness as soon as possible when talking about new hardware.
And if you look quick, you might even catch the price war over at Pricewatch. I've been debating the purchase of a new computer for the past week, and in that time, the price of the gigahertz Athlon chip has dropped 12%. In one week! So I'm going to get the motherboard and some ram delivered next week, and wait a bit to see how low the chip goes. It's already under 3 digits.
---
I don't think Intel is backing off the P4, despite its many flaws. Intel will continue expensive development projects around the PIII arch, because the P4 doesn't support SMP, and the next few revs of Xeons will be based on the PIII until they are scrapped for the 64-bit Itanium. The Itanium wont be real and ready for prime time for about 2 years.
Someone you trust is one of us.
*Scoff* *Scoff*: But who in his right mind would waste such a beautiful machine on Windows 2000 anyways? And which motherboard doesn't need a service pack to run Windows?
The Pentium 4 is MUCH faster than Athlon on FP. The 1.7 GHz P4 scores a SPECfp of 598 and the 1.4 GHz Athlon scores a laughable 426. FYI, the P4 scores higher than Alpha, PA-RISC, and every other processor, making it the fastest CPU in the world. It also is #1 on SPECint, scoring 575 vs. 495 for Athlon. SPEC benchmarks are considered to be considerably more scientific and reliable than the toy-type unsophisticated benchmarks you see on peecee hardware review sites.
They are still benchmarks, though. If you look at a broad range of benchmarks, the pattern is exactly what you would expect, considering the technology.
SPEC benchmarks are based on real applications. The integer suite consists of real applications such as GZIP, GCC, etc. As does the FP suite, which consists of real kernels.
In broad-looped, unpredictable, or multi-tasking situations, that deep pipeline and RDRAM prove lethal as the processor burns off an insane number of CPU cycles getting its act back together after each misprediction.
For starters, P4 performance is by no means affected by the number of pipeline stages. To say so is quite a pedestrian claim. Although the branch misprediction penalty measured in cycles is higher than other processors, the P4 burns through cycles so much faster than other processors that it all evens out.
Second, Rambus is the reason why it is so good, particularly at getting the fastest floating point performance in the world. The P4 has a faster bus than any CPU on earth (at 3.2 GB/s it is faster than 266 MHz Athlon at only 2.1 GB/s and 200 MHz Alpha at 1.6 Gb/s). The FP benchmarks are mainly driven by memory bandwidth which P4 excels (thanks to Rambus). DDR-SDRAM is considerably slower, and would not be able to saturate the P4 bus. Athlon systems only show a 4% bandwidth increase due to DDR, while P4/Rambus shows a 300-400% increase over Athlon/DDR. Whether that is Athlon's problem or DDR's problem, I'm not sure.
The this is this: Floating point is used very seldomly in modern applications, and the intensive application of it is in extremely rare scientific apps (which the vast majority of us don't run), or games.
Are you new to computers? Engineers and scientists spend many tens of billions of dollars per year on technical computing, a market which is driven mostly by FP performance and memory bandwidth. The fact that a $1500 Pentium 4 PC made from commodity parts outperforms the fastest $20,000 Alpha's and HP's on these types of applications is a very big deal.
Why the hell would you want to compare performance vs. frequency, when one part is avaiilable at a substantially higher frequency?
You must be inexperienced with SPEC benchmarks. The P4 wins @ the base benchmarks, and the others win with peak benchmarks. The baseline benchmarks are for using standard optimization, which is what ISV's do when developing software. The peak number means that you are allowed to use several different optimizations throughout the suite, but nobody does that in real life. So, the P4 wins at the ones which are reflective of real software.
The lead vehicle for a new process (in this case 860) is always a tried-and-true CPU instead of something brand new. So, of course PIII will have a shrink before P4 because it's better understood how to shrink it. The P4 shrink is only a couple of months behind it. Remember the transition to 854 when the fast MMX Pentium's came out way before Klamath? This is the same deal all over again.
:-)
Of course, the speedup from Northwood over Willamette will be substantially higher than the speedup from Tualatin over Coppermine. The P4 microarchicture is a lot more scalable; for example, the bus is triple the bandwidth of PIII/Athlon, so it can scale that much more without memory being a problem.
Sounds to me that Tom is just making a big fuss since he got Tualatin samples instead of Northwood's.
Sound like what happened with the Celeron A whipping the P2's ass?
I am !amused.
Now that the Athlon MP and dual-CPU mainboards have been released, it just restates what we knew a year ago: AMD is slowly but surely beating Intel at Intel's game.
I think that both you and your friends need to try developing (assuming that you develop) or gaming (and I'm sure that you game) on AMD boxes for a week.--
--
I like to watch.
From Tom's article:
Of course we wouldn't let this CPU suffer out in the cold, far from its mum in Satan Clara!
I knew some folks felt the whole newfangled technology thing to be evil, but jeez...
--
Er, Slashdot is the contraception.
I mean, how many of us have been fucked by something other than Unix?
III.IIVIVIXIIVIVIIIVVIIIIXVIIIXIIIIIIIIVIIIIVVIII
I think that Intel will release it just because they'll be able to continue to sell to the ppl with Socket 370 mainboards, and the manufacturers who already have an investment in building computers with those boards, which are probably more stable than the newer mobos. Its a way of keeping the revenue stream wide...there are still plenty of ppl that will buy the P4 just because of the 3 blue guys -K.
I need a TiVo for my car. Pause live traffic now.
It is clearly not easy to rank these processors in any order, especially when you usually get limited information, be it clock speed or certain benchmarks or whatenot. But I would just like to clarify a few things.
First of all the P4 is a larger processor with more overhead so that it can easily reach faster speeds. The estimated maximum clock speed is 10Ghz for the P4, and it will be undergoing a size reduction, it will come out on DDR, and most likely w/ copper interconnects. It also has SSE2 and other special features that compilers are not yet optimized for. When that optimization comes out, you will see a big performance increase. The P4 is also designed to promote maximum bandwidth vs speed. Something everyone will probably start to appreciate when the clock speeds get up there.
The p3 is similar to the Athlon in that they are both have aging core designs and it is becoming increasingly difficult to speed them up. Thats why Intel redesigned the core of the microprocessor to allow for faster speeds. There is a lot more overhead but that is to be expected. There is a lot more overhead when you try and fly a plane instead of taking your bicycle somewhere. I personally favor Intel over AMD, but I have biased reasons to do so. The p3 (imho) beats out the Athlon, and optimized software for the p4 beats out the Athlon (even if it doesn't the Athlon will soon be completely overwhelmed by increasing clock speed.)
isn't microsoft supposed to be the one humping our ass? :)
--- Metamoderating abusive downgraders since my 300th post.
Intel would wind up looking rather stupid if they released Tualatin, that's the problem. It would be very much like the situation they had trying to position the Celeron against the Athlon, and we all know how well that went over.
The situation Intel has right now is that it's becoming obvious to all that the P4 was a mistake. If they try to market the P3/Tualatin as downmarket (the P3 seems to have pretty much replaced the Celeron these days) nobody will pay attention. If they put it up against the Athlon where it belongs, it's lights out for the P4 and the Intel marketing department will wind up guzzling Alka-Seltzer after blowing big money on commercials and such.
Either way, it works to AMD's advantage -- all they need is an ad agency with the stones to tease Intel about pulling their punches.
/Brian
/Brian
Ick. P4SX maybe, but I think they probably need a new name for it altogether.
That's a fun thought, though: the P6 as the pinnacle of Intel's chip design capabilities (and don't tell me Itanium's not a boondoggle). And they've been milking it for what, six years now?
I think it's rather interesting, though, that Intel seems to have hit the limits of its collective ability to do anything interesting. FWIW, I think Sledgehammer will get a lot more mileage than Itanium myself just because there's no tricky business in the instruction set. But the fact remains -- Intel's day in the sun definitely seems to be over. They used to be the most dangerous fish in the tank; now they're just the biggest.
/Brian
Did anyone else see this initially as:
Tom's Looks At The New Pill
I though it was going to be a Matrix II reveiw or something.
Oh well, my mother did tell me it would effect my eyesight.
All of this is true. However, it is much easier than what was done even just going to PIII coppermine, I believe. As CPU designs go, I think it was very cheap. I imagine the pipeline structure is the same, with tweaks appropriate to the new process. And the new process will buy some speed on its own, just not that much. It will give the P4 a bigger increase -- taking it from 2 GHz to 4 GHz last I heard; it will probably take PIII from 1GHz to 1.5GHz, a much smaller (though not insignificant) increase. But we'll never see 1.5GHz PIIIs is my guess. And while you can't just stuff in more cache, I don't think it's that hard either.
not really. There is still a PIII market, and Intel is selling to that market. It really doesn't take as much time to move the PIII to 0.13 as it would to do a redesign. I think the real problem is the Duron -- the celeron is underperforming, and so the PIII is competing in the mid-low end with slow Athlons and Durons. Hence, they needed to up the clock speed. A slight tweak to the chipset, and it's ready to go. I think the PIII will hold the mid-low market segment, and slowly ramp in clockspeed.
They design a new chip using an established process. This gives a stable process to debug the new logic. At high speeds one thing that needed fixed was how to get the clock signal to arrive at all parts of the chip at the same time. This debugging and design was done on a known working process. Later they migrate the chip to the new process that that was developed. Making new logic and a new process at the same time is the same thing as going up the stairs 10 steps at a time. Few people make it even after many failed trys. Working out the bugs and tweaking to perfection a step at a time will get you there faster. The latest migration is from the 858 to the 860 process which uses copper interconnects. They are testing it on an established processor (PIII). I also think the marketing department noticed the problem with Rambus and decided the need for fast chips that use cheap memory may sell better. After all they do produce to fill the customers orders. Right now in a stagnet market, faster and cheaper is what sells. Faster cutting edge at any price is a slow market.
The truth shall set you free!
P6 architecture on which Pentium III is based has 6 10-stage pipelines while Pentium 4 has 9 20-stage pipelines, for out-of-order instructions execution.
Sometime the out-of-order execution would fail due to excessive branching and mis-prediction, in this case Pentium processor will flush all the stages in pipelines. Pentium 4 has then lost more instructions then Pentium III on average, thus lost more execution cycles. Chances that a program which has a lot of conditional branching would run faster in Pentium III.
That explain why some benchmark tests show Pentium III out-perform Pentium IV in some cases.
P.S. FYI, Athlon only flush 1/2 stages on average and that explain why Athlon run faster then Pentium III&4 at same Mhz.
I thought the P4 was supposed to have better branch prediction to minimize the pipeline flush penalty. Does anyone know which CPU has a better branch prediction algorithm?
I think its major improvement is to make each stage comparatively simple by lengthening pipelines, but it brings other problems as I said before
There are improvements in other aspects, but they are mostly useless, some reviewers said.
Simple: They still have some stability and compatability issues. Now perhps the new 760s boards have cleared this up, I haven't use one, but I HAVE had problems with Athlons in the past. It continues to get better, but when for some it's still an issue. When I got my mobo it was really bad. I bought an Athlon 700 and an Abit KA7 (VIA KA133 chipset). I could not get my system to work. In addition ot having random crashing problem, etc, as soon as I installed my video drivers the whole thing went to hell. Took the board back, got another, same thing. Canned the Athlon and got a PIII 700 on an Asus CUBX and had no problems getting it to work straight off. Yes, I know that the KT133 is significantly better, however you can see how issues like that scare people off. AMD is going to need to prove themselves for a little bit before some people will buy them. Remember: Many people are willing to trade some speed and cost for the gaurentee of stability and compatibility.
Any and every motherboard based on an Intel chipset. The 440BX, 815, 840, etc all work with all version of Windows straight off, no patches required.
Most Asus motherboards are based on Intel chipsets.
...that Intel is slowly backing away from the Pentium IV.
.. anything .. to save their collective asses when the bomb goes off. Their 64-bit offerings are not ready to go; there is very little software support. That leaves the Pentium III.
If that isn't the spin that you take away from this story, it doesn't even make sense. Why in the world would a bellweather tech company that is having severely difficult financial times all of a sudden dedicate a whirlwind of time and energy to a previous-generation processor? Unless the Pentium IV has serious problems, this expenditure of resources doesn't even remotely make sense.
The Pentium IV has already had a troubled history. It underperforms. It runs way too hot. Machines that use it have had some high-visibility, bad-publicity recalls. Some have estimated that as many as 70% of P4-based machines experience intermittent hardware problems. My guess? Some engineers at Intel have discovered something even worse about the P4. They've discovered that the P4 is a ticking time bomb and are looking at something
Personally, I bought a new machine five months or so ago, and instead of buying a P4, I went with a 1GHz Pentium III. A lot of my friends ridiculed me. Well, they're not looking so damned smart now, are they?