Radeon 9700

I have some mixed feelings about the Radeon 9700 due to my experience with the 8500.

I bought that one shortly after introduction, paid a bit less than $300. I basically had to shell out because I needed something that could actually handle 3D, unlike the Matrox cards. There were some initial driver problems. nVidia came out with GF4. Then the price dropped. A lot.

Now I find myself essentially in the same boat again, and probably will go through the same experience. Talk about deja vu.

Why am I going through the same thing again? To some degree, it’s society’s fault. This particular computer hardware society’s fault.

3dMark has become one of the universal benchmarks. Problem is, it’s a terrible universal, standalone benchmark. It’s a very good video card benchmark, and may be an OK comparative benchmark in some cases, but that’s it.

In other words, in a video card comparison, 3DMark2001 is a very good comparison. In a comparison of motherboards tested under equivalent conditions, it may be OK. But looking at five websites and just looking for the highest 3DMark score you can find is not OK.

If you look at Mad Onion’s databases, you’ll see that scores pretty much separate themselves by video card, with little mix. You’ll also see very similiar 3DMark scores from machines with the same video card, but significantly different CPU speeds.

3dMark is essentially a “do you have the latest video card” benchmark, which makes the competitions now surrounding it a little strange. You just don’t stand a chance being at or near the top unless you have the latest and greatest.

Now there’s nothing wrong or evil with that. That pattern is just what you should get from a good video card benchmark. The problem occurs when you don’t know what the game is and you think it’s something else.

To play serious football, you need shoulder pads. So long as everybody walking on the field knows that, no problem. If you don’t know that, big problem.

I suppose some might equate it to buying the best athletic equipment, and others might view it like those who run in a marathon, but couldn’t compete to win if they got a lift for half the course. Instead, you have people competing in a “best-of-category” group.

Now if you want to compete in those kinds of games, fine, it’s your money. Just know the rules of that game.

What Game Do You Play?

If that is your game, in all likelihood, once the numbers get posted, that will probably become top dog, and you’ll have to shell out about $400 if you want to play the 3DMark game seriously. If you’re in that boat, the Radeon 9700 will almost certainly be a must-buy.

For most of you, though, that isn’t your game. If you buy a 3D gaming card to actually play 3D games; the rules of those games are quite different.

You don’t play Quake by just hooking into a network and measuring each others’ FPS. You’ve never shot somebody just to be told, “You can’t kill him because he can get more FPS than you.” You win by doing, not measuring.

The problem with benchmark mania for you is rather more indirect.

If someone looks at other equipment, and that person doesn’t have the latest and greatest video card to do it with, that person is going to pull a (relatively) low 3dMark score. Many people looking at that review aren’t going to realize why, and think there’s something wrong with the other equipment or the tester rather than the actual cause.

The only (somewhat) valid use of 3DMark in testing other equipment is in multi-whatevers comparisons where the “whatever” is the only thing that changes. But again, the video card is the eight hundred pound gorilla in this comparison.

If you have two or three or twelve mobos, and you run 3DMark a dozen times, the only difference being the mobo, that could be a valid comparison (then you have to look to see if settings are either equivalent or at least set to the max the system can stand).

So you might be able to say that mobo A does better with video card B than mobo C, but you can’t say that mobo A will always do better than mobo C in video, or that the benchmark scores are indicative of what you’ll get with another video card.

Reviewers generally try to deal with this by having the latest and greatest video card, but that’s precisely what most of you won’t do, so any video card benchmark will be meaningless to your personal situation.

I realize that for many of you, this is a huge “DOH,” but it isn’t a “DOH” for many others.

ATI: A Grand Coach Until The Clock Strikes Twelve; Then It’s Pumpkin Pie Time

If 3DMark isn’t a sport for you, the buying track looks quite a bit different, and illustrates ATI’s core weakness vis-a-vis nVidia.

The card will probably be and stay more expensive than most of you will want to pay until nVidia is close to being ready to trump it. Then it will drop a bit. After nVidia trumps, then it will drop a lot.

This is another game altogether, and one that seems to be rigged in nVidia’s favor, at least insofar as most of you are concerned. Given your preferences, nVidia usually gives you what you want at a price you want. ATI really doesn’t; they seem to play (or be forced to play) Cinderella; early adapters pay through the nose for a coach, then watch it suddently turn into a pumpkin.

Of course, in this fairy tale, everything eventually turns into a pumpkin, but at least nVidia gives you some options as to what you pay for the coach. If you pay them less, the coach may turn into a pumpkin at 11:30 rather than midnight, but that’s perfectly fine for most people.

I think ATI is going to have problems successfully competing against nVidia until they start offering the same variety of options.

Conclusion: One Size Doesn’t Fit All

This audience is not monolithic. There is no one answer that fits everyone. For those into 3DMark competition, this will probably become a must-buy.

For anyone with a sizable degree of cash consciousness, it isn’t, and probably not even a good purchase, given your priorities. Same video card, different people and motivations and values.

Benchmarks are not monolithic, either. They don’t have value in-and-of-themselves. They only have value in the context of what they’re meant to test, and how well they do that. They don’t work equally well in all situations.

In the fairly near future, we hope to provide some sort of guide to the various benchmarks out there, and what they’re good and not good for, when they should be used, and when they shouldn’t.

Review sites have the habit of giving you a lot of fish. We think it’s better to learn how to fish, or at least be able to recognize the good fish from the bad ones.

Ed