By Michael Ahlf
Intel, known to many by the not-so-flattering name of Chipzilla, has recently sent out its Pentium 4 to a few select review sites, to see what response they would get. The big name right now for news on this new chip is Tom's Hardware, run by Dr. Thomas Pabst. Pabst himself has had a colorful past, running the gamut from being a devout 3dfx lover, to being flamed when he posted not-so-flattering (and later proven flawed) Quake 3 benchmarks between the NVidia and 3dfx chip offerings. Be that as it may, it seems even he can't make up his mind, as he continues to benchmark the Pentium 4 day in and day out, with every test he can find, trying to make heads or tails of it. Well Tom, if you don't mind, allow me to add my two cents to the mix.
Let's take a look at Tom's Hardware for a minute. To the untrained eye, they seem confused, unable to make up their mind about the P4. First, their benchmarks showed Intel kicking butt in Quake 3 and Mpeg authoring, while reaching back in some other benchmark software designed to simulate commercial applications. Then, a surprise -- it turned out that Mpeg encoding using floating point rather than MMX routines, which can arguably look a LOT better due to the loss of rounding errors, cut the P4 down to size quickly. The next turnaround came just a day later -- mysteriously, an Intel employee sat up all night and programmed the same mpeg encoding software to use Intel's SSE2 routines, and lo and behold, the P4 came roaring back. Intel got to crow about how "easy" SSE2 was to use (nevermind that this was most likely one of the SSE2 developers, or an in-house writer who's had access to it for well over a year), and how quickly the changes were made (rather trivial, I would imagine, as most of the work would involve substituting the semi-identical SSE2 command for a floating point operation at the right points). Then, the benchmarks showed Intel on top of the performance crown, with their 1.5GHz chip beating down a 1.2GHz Athlon. One has to wonder, however, what the difference would be with a 1.5GHz Athlon? By the time the P4 has hit the market, this kind of accurate comparison will be available, as the 1.5GHz Athlon will be for sale to the general public.
Still, Intel doesn't deny that their P4 processor has a really weak floating point line. Their claim, quite the opposite, is twofold: first of all, you'll never miss the speed on older apps because the chip is so lightning quick. Second, all future apps will rely heavily on SSE2 anyways, because it's so easy to use, why wouldn't they?
Wait! Hold on to that phone, Intel. You may be Chipzilla, but the market has changed a lot from the times when FUD (fear, uncertainty, doubt) tactics would scare people from buying AMD processors. You're facing what's essentially an even battle right now, and you're doing it with your chips costing a lot of money compared to the competition, for less and less performance advantage as the days go by.
The bottom line of the Pentium 4 is this: it is NOT, as Intel will claim it is, an all-around workhorse. Unreal Tournament shows otherwise quickly: anything that uses floating point operations will doubtless do the same. The Pentium 4 is designed so that Intel can try once more to exert market control. Why not buy Intel? With SSE2 on chip, you'll be able to run everything those silly Athlons won't, because everyone will be programming SSE2 by years' end. Or at least, this seems to be their thinking. Right now, Intel seems determined to ram SSE2 down our throats just like they did with MMX, forcing AMD and Cyrix to license the technology to get anyone to buy their chips, even though MMX is to this day something of a waste of silicon compared to what could go in that space. Why would Intel do that? One reason: market dominance. Since the market has changed so much, Intel is running scared from AMD, and have been trying desperately to take control of the market again. To do that, they need an ace in the hole, and both Rambus and the 1.13 GHz P3 were severe busts.
So what do you get with the P4 if you buy today? Let's sum up.
First, you get a soon-obsolete chip mounting system (next time you upgrade, expect to buy a new motherboard... AGAIN).
Second, you get high-powered Quake 3 performance.
Third, you get an anemic floating point unit: don't expect Unreal Tournament, or much other current software, to be seriously improved compared to your Athlon.
Fourth, you get SSE2... if Intel can convince anyone to use it seriously.
Bottom Line Time: buying into the P4 right now won't do anything for a serious gamer, and serious gamers recognize SSE2 for what it is: Intel's attempt to change the rules of the game since they've had such a hard time competing with the Athlon going toe to toe. They're hoping to clobber enough designers to use SSE2 (thus becoming Athlon-incompatible) or else have to use both SSE2 and Floating Point to get decent performance on most home systems, thus increasing development time on titles trying to do this. More likely than not, we'll get a hundred and one "SSE2 patches" for various games, most of which won't do much at all. Given current market dynamics, Intel can't really claim they're leading the industry by implementing SSE2, since it's not an open or even at this point licensable instruction set. The bald truth is, SSE2 is an attempt to clobber the industry to falling back in line with what Intel wants, not what the consumers want.
Face it, Intel, you're going to have to work a LOT harder to get this fish to bite on the worm that is SSE2.
All respective copyrights named or used in this article, and images, are copyright of their respective copyright owners. All opinions expressed in this article are the opinions of the writer, and may or may not express the view of Glide Underground, Inc. Please don't hurt me.