Slashdot Mirror
P4 3.2GHz Reviews
Nathan writes "The Intel 3.2GHz Pentium4 has passed its NDA with reviews coming out over the net, including this one at MBReview, This one at HardAvenue, This one at TweakTown and this review at HotHW." Yay. Benchmarks. Wowee-zowee.
(Yes, fellow pedants, I am aware that "Pentium" was used for the chip following the 486, as Intel couldn't copyright a number and stop their competitors using the term "586".)
Seriously though, how long have successive generations of Pentium technology lasted? Is it just me, or was the PIII the primary product line for longer than the PII, and when will the P4 break the PIII's record?
When I am king, you will be first against the wall.
And the other 95% of computer users will pick the cheaper 32-bit Intel chip running Windows. What's your point? You're willing to pay an enormous premium for very little gain? The average consumer isn't going to see a difference between a 32-bit CPU and a 64-bit CPU other than one is going to be more expensive and perhaps run a bit faster. In 6 months the 32-bit CPU will trounce it yet again.
We've been going over this for over 10 years now. If 64-bit CPUs were some kind of panacea then the DEC Alpha would have become the dominate desktop chipset. Now it's just pleasant history. Mac users will continue to buy Macs no matter what CPU is in them if history repeats itself. They're even willing to lose almost all binary compatibility in the switch if necessary. Heck, if I remember correctly they've done it twice so far going from Motorola 68k chips to PowerPC and then from OS 9 to OS X. Most people aren't willing to make such a sacrifice in the name of platform advocacy which is why Windows still runs old DOS programs.
I find it irrelevant whether the speed of an existing type of processor has increased by less than ten percent, although looking at the price compared to the 200MHz lower clocked variant, maybe this would fit under "It's funny, laugh".
;)
However, this processor does seems very suitable for overclocking (4GHz, yikes!). Did anyone manage to come close to that with the 3GHz model, or has Intel increased the therapeutical window of their processors slightly?
Now this is really pure FUD, I'm afraid, but it does make me laugh.
Yeah, it's true that the masses will probably stick to what is cheaper. It's what they're always gonna do, and that's fine, because most people just want Office and maybe the occasional game. Apple will never really penetrate that market.
But this is Slashdot. We demand more from our machines here. We want high speed UNIX boxen and game stations that we can frag at 150 fps on, and if we're lucky, both at the same time.
The bit about binary compatibility shows that you know nothing about Macs. The PPC 970 _is_ backwards compatible with all the old software - everything will run! And the best thing is, as has always been the case with Macs, backwards compatibility is unrivalled. Macs of today still feature Motorola 68k emulation so that they can run software written for those chips, for OS 9 and for OS X.
Windows XP (the equivalent of OS X in terms of consumer accessibility and reliability), on the other hand, has terrible backwards compatibility, and I find that many, many, many old DOS or even Windows programs will not run...
I rest my case.
iqu
Something is missing, oh yea, the Intel vs AMD benchmarks, WTF, how can you compare your own CPU's to each other, I wanna see how they hold up to AMD
If there is nothing left worth living, what are you willing to die for?
If you need that kind of toast-making performance, you're luch more likely to either build a toaster-farm with dozens (or maybe even hundreds) of inexpensive run-of-the mill toasters, or splurge for a big, heavy-duty continuous-feed made-to-order beltway toaster.
Sort of like getting either a cluster of cheap middle-performing x86 boxes, or a big-iron type machine from Sun or IBM, come to think about it.
I mean, how many apps really critically need that 2% parformance increase, but do not benefit from a dual or quad-cpu machine, a cluster, or a big non-x86 Unix machine?
Trust the Computer. The Computer is your friend.
when money is available, you buy the top-of-the-line computer. You may be using it for the next ten years.
This is the rationale I hear for buying expensive hardware from Sun or SGI (and I agree, for the most part). I've never heard it used to justify buying Intel's latest offering - PCs are retired quicker than any other platform. If you really need to make a crappy PC workstation last for ten years, you're better off buying a cheaper box, like a 2.4Ghz P4 (which isn't slow by any means), and use all the money you save to purchase spare boxes or parts. You'll definitely need them if you want to keep the system going for ten years.
I know from experience that there are few things more annoying than trying to squeeze the last bit of life out of PCs that have been obsolete and off warranty for two years. . . sometimes, when the moon is out, I can still hear those IBM Pentium 90s calling my name.
I have. Read the line about integer performance and you'll see why Apple will still be playing catchup with Intel and AMD. Most people are going to be doing integer and not floating point calculations when they are running their systems. Those that do benefit from floating point are likely not "Switch" candidates anyway. Either way, it's difficult at best to just drop one system and replace it for another when it comes to FP calculations as you not only need to purchase new hardware, you have to purchase new software and even with Adobe allowing crossgrade licensing, it's going to be a big hit to the wallet.
Modern chess engines represent the board as several 64bit bitboards, one for the white queen, one for the black queen, one for the white pawns etc.
This as opposed to the good old days with a 64 byte array containing 1 for the white queen, 2 for the white pawns etc.
Bitboards really benefit from 64bit registers and 64bit (integer) arithmetic.
"There is no substitute for thinking" - Bjarne Stroustrup
While those reviews are more than adequate, I am surprised that Tom's Hardware review is not mentioned. While I would not mention it blindly just because THG was one of the first sites to offer in-depth reviews, after reading it I gained more insight than from the other "here are the benchmarks, mam" sites. Here's the synopsis:
"Intel launches the last P4, with 3.2 GHz for FSB800 and Dual DDR400. Its rival AMD fights back with the Athlon XP 3200+ and Dual DDR400. With the Pentium 5 and Athlon 64 waiting in the wings, it's a historic duel." [tomshardware.com]
Must-not-watch TV!
Escher was the first MC and Giger invented the HR department.
I agree that this article wasn't that interesting, but I did find the following passing interesting
Next up, weâ(TM)ll be taking a look at FutureMark's 3DMark2001SE. With the recent debacle surrounding NVIDIA and FutureMark, I have chosen to exclude 3DMark2003 from our benchmarking suite for those of you wondering why you arenâ(TM)t seeing any results for it. (from here)
We've all read how NVIDIA fiddled with the results and how FutureMark became complacent with it. Now here's the result.
-no broken link
Someone somewhere else on this thread sarcastically suggested that I was simply assuming 64bit to be twice the speed of 32bit.
Well, at one point you said "64 bit processors can process twice as much data as 32 bit processors", or words to that affect. Which is a pretty meaningless statement without a time reference.
However, 64bit quite clearly is the future, and whether x86 or PPC is your architecture, it's where we're going.
Sure. Eventually. In much the same way that ipv6 is the future.
surely the ability to fetch 64bits of data at a time rather than just 32bits is going to speed things up?
I'm pretty sure that the rate CPUs read from memory is actually limited by things like memory bandwidth and speed. I think most CPUs already fetch memory speculatively in chunks of 128bits or more, so I doubt that'd make much difference.
You also have to remember that the size of the pointer type doubles. That can actually decrease efficiency - as pointed out in the article linked to in a sibling post, a lot of computation involves linked list traversal. The increased pointer size would cause greater amounts of data to need to be processed.
It is quite clear from your posting history that you do not like Macs
Well this is the interesting thing. I don't have much against Macs themselves, other than a general dislike of proprietary platforms (but the same is true of Windows or Solaris for instance). It's more the attitude of some (unfortunately the most vocal) Mac users that annoys me. A lot of, well, to be frank inaccurate things are said about Apple and their products, and it's a big turnoff.
It's especially annoying when people work themselves into a frenzy then treat a corporation and its product almost like a religion. So that's where a lot of my "anti-Mac" viewpoint comes from, not in fact the technology or even the company themselves (though apple have done their fair share of shady things) - just the blind loyalty of its users.