Video Wars Or Just Vidiocy?

nVidia caught hell a while back for saying they were giving people trilinear filtering, and not. Now ATI is catching some equivalent hell.

Frankly, many if most people arguing about this could use more than a few extra optimizing algorithms programmed into their brains themselves. They need them more than the GPUs.

The Facts of Video Life

The performance of a video card lays on a spectrum. On one end is “How fast do you want it done?” and on the other end is “How well do you want it done?”

On both ends, the tools for measuring how well you’re doing for either goal are highly subjective, each in their own way.

People don’t realize that, indeed they would vehemently disagree with you. They would tell you that you can easily measure how fast something gets done, while measuring how well it gets done is another matter.

There’s the problem. What is really happening is that you can put a simple number in the “how fast” category, while you can’t in the “how well” category.

And that’s not the same thing.

“Objective” FPS?

People have a very bad habit of paying attention to factors that can have a number attached to it, while ignoring factors that are just as important that can’t have a number attached to it.

FPS is one of those measurements. The problem is FPS, as used by people, is a crude and uneven measurement. It’s much like an IQ test. Both do a good, quick job at measuring gross differences. Neither is good at measuring minor differences, but because both are a single number, they give a misleading impression of significance when there is none. A twenty point difference in an IQ test is significant, a two point difference is not.

I defy anyone without a frame counter to tell me the difference between a scene running at 480 as opposed to 440 fps, much less 480 vs. 475. On the other hand, any idiot could tell the difference between 50 and 10, or even 15 and 10.

More importantly, though, there’s a million and one different ways to manipulate FPS, and if that’s the only or main thing people base their buying decision upon is that, that’s just what the sellers are going to do to get the money, whether it actually makes any sense or not for the real task.

What is the victim of such manipulation? Image quality, but since FPS is “real” in a way image quality isn’t (at least in the eye of the average buyer), so long as you don’t grossly sacrifice image quality, you can “cheat” an awful lot.

Is it “cheating” when no one (or just a few) notice it? When does a trade-off stop being an intelligent use of resources, and starts to become a cheat?

If you really want to get deep, isn’t the purpose of the trade-off effectively for a false god your buyers believe in?

“Subjective” Image Quality?

On one level, image quality is completely subjective. We all have to see the image, and we all have different kinds of eyes taking pictures and different brains processing the images.

This doesn’t mean that there isn’t some objective reality behind it, but that often doesn’t matter.

Take a picture and save it in lossless BMP format, then take it and save it in a lossy JPEG format. Print out the two. Objectively, there’s no doubt the BMP picture is “better,” but some will find no difference between the two, and at least a few will find the JPEG-picture “better.”

Factor in the relative size of the files, and many more will find the JPEG “better” for their purposes.

This is not going to change any time soon.

This is not to say image quality hasn’t increased in importance the last couple years. There have been de facto image quality standards established, or at least people think that’s what they are. Call it AA, AF, FSAA, bilineal, trilineal, whatever, these are all supposed to enhance image quality.

But note what these really are. These are just terms for certain kinds of additional processing of a video image for quality. It is no guarantee of better quality. Sometimes they work, sometimes they don’t.

What such techniques do guarantee are lower FPS, simply because the video images are getting extra work whether it makes any sense to do so or not.

Given that situation, a perfectly good, decent person might want to be a little more intelligent about this and only do the extra work when it’s really needed.

Given that situation, so would a perfectly sleazy walking artificial sperm receptacle.

How do you tell the difference between the two? Can you?

The Polarized Political World of Video Cards…

The Polarized Political World of Video Cards

You can see that even a saint with the best of intentions would have a rough time discerning what makes one video card “better” than another, roughly equivalent one.

To put it mildly, this isn’t Heaven.

The situation is more like that of a Martian coming to Earth to find out what the fuss is all about with these guys Bush and Kerry.

If he asks Michael Moore, he’s going to get one picture. If he asks Rush Limbaugh, he’ll get quite a different picture. If he asks both of them, he is going to be one confused Martian knowing less than he did to begin with.

This is what the video card arena is like. Truth ran away a long time ago and didn’t leave a forwarding address. There are two sides. For and against. Us Vs. Them. Truth? There is only one truth: Us is right; Them is not only wrong, but evil. Attack with anything and everything that can be used against Them; ignore anything and everything that can be used against Us.

With such a blindered mindset, any talk about “optimizations” becomes, “Our optimizations good. Their optimizations bad.” End of story.

Not the best in the world.

Even worse, the discussion has become so degraded that even the “impartial” can’t see beyond “Optimizations are cheating. Both sides optimize. Both sides bad.”

This is not much better.

We have a bad problem with worse judges.

The Mailman’s Dilemna

You have a mailman. He needs to deliver mail. Sometimes he has more, sometimes he has less. No matter how much or little mail he has, he needs to deliver it within a certain period of time, and the quicker, the better.

The people on his route get different quantities of mail. Sometimes they get a lot, sometimes none at all. Some mail is more important than other mail; an unsolicited offer to clean your chimney is unlikely to be considered as important to the recipient as a Social Security check.

Let’s say there are some days when there is just too much mail to deliver on a certain day, and other days when there’s slack.

What does the mailman do?

Some would prefer that the mailman visit each and every door, if only to tell each and every person that they don’t have any mail today. You could do that, but you’re going to end up with a lot of undelivered and late mail that way.

The mailman can decide to optimize his route by just visiting those people who really have mail.

If push comes to shove on especially heavy days, the mailman might even decide to let commercial junk mail sit for a while to be delivered on slower days, and just deliver the higher-priority stuff.

Or he may decide to deliver mail to people every other day. Or he may decide to simply trash excess mail, maybe just junk, maybe more than just junk.

The mailman has a broad spectrum of options to “optimize” his route. Some cause little if any harm, while preventing greater harm. Others cause much more harm, or even violate the whole purpose of the exercise.

In short, there are good optimizations, and bad optimizations. There are trade-offs. You look for the trade-off that gets you the most good for the least bad.

You certainly wouldn’t find the mailman not knocking on your door to tell you you don’t have mail to be the same as him throwing your mail in the trash. Yet that is just what people who denounce all optimizations are doing.

It’s simplistic, if not moronic.

It would be even dumber if people said, “If the FedEx guy does it, it’s good; when the UPSS guy does it, it’s bad,” but how common is that when you attach different names like ATI and nVidia?

But even excluding that vidiot view, insisting that cards be tested only a certain, set way is like insisting that all mailmen stop at every house every day and tell you he or she has nothing for you today. No real mailman does this.

Consider the possibility that if you do come up with a smarter, better way to deliver the mail, you might not necessarily want to be too forthcoming about it, since you’ll catch hell from both the ideologues who hate you no matter what you do and those who think the only honest way is the dumb brute-force approach, regardless of any other negative consequences.

These attitudes do not help innovation, they block it.

Dumb Demanding Dumber

Some would say that the only way one can fairly compare competing products is to equalize the conditions between the two.

While this is a good idea sometimes, it’s ridiculous here. These are two very different product lines with much different design strategies behind. The point of these optimizations is to make the products better. Demanding that any competition be dumbed down to a low standard is like lobotomizing everyone before taking the SAT test.

It’s even sillier when the competition is all about an essentially meaningless objective.

Maybe the best way to describe this is that you can’t easily apply two-dimensional rules to a three-dimensional race, and you certainly can’t apply simple two-dimansional rules to it.

In a way, both sides have decided to run on downward inclines due to optimizations. You can insist on them running on flat inclines, but that’s not the real race. The better approach is to judge the inclines. Does the optimization get the job done better and faster, or just faster?

So long as video speed and video quality are related factors (and that’s certainly not going to change anytime soon), you rely solely on numbers. You have to apply judgment.

Much Ado About Little?…

Much Ado About Little?

Quick, who’s the better writer? Mark Twain or Charles Dickens? Now prove it!

Pretty silly question, isn’t it? Any answer or explanation you come up with will reflect you more than it will the writers.

Quick, who’s the better writer? Mark Twain or you? Unless you’re extraordinarily talented or extraordinarily self-centered or conceited, odds are everyone could agree on that answer.

Measuring video cards is much like deciding who the better author is. When there is a gross difference between the two, simple crude tests work pretty well for everyone. When there isn’t, they don’t.

When the crude measurements indicate it’s close, what ends up happening is that Card A does some things better, and Card B does other things better. What then matters is what you find most important and valuable, and there is no single simple answer to that. What you think is best is not necessarily what someone or even anyone else will think, and it’s not a matter of one person being right and another being wrong.

So many people do not comprehend that, and the answer to such legitimate differences of opinion is not to yell louder.

And let’s face it, we don’t need a single right answer here. No one will die, or likely even notice, if they buy what you think is the “wrong” video card in a certain class, much less you.

This is not exactly like fighting terrorism.

Yes, it is better to argue more intelligently about these matters rather than apply simpleton rules to a complex situation.

Nonetheless, no matter how brilliantly you do this, there comes a point where you’re just arguing who is better: Twain or Dickens.

As a body, we don’t have to choose. One doesn’t have to win. We’re better off with both.

Think about it.


Be the first to comment

Leave a Reply