Which is better: 1900XT or 7800GTX 512mb?

[werrrd]

You know, the X1900XT (and XTX) are great cards and all, but seeing as I already own a GTX512, I see no immediate need to pick one up right now. I mean, the 1900 in some cases can be 40-80% faster in FEAR, a game which I already beat and has no replay value. GTX512 is still on top in Q4 and perhaps that's a good indication that it will still be on top or very close for ET: Quake Wars, which of course is an important title. As for Oblivion, I'm pretty sure the 1900 will be faster, however, the game is capped at 30fps anyway, I think my "aged" GTX512 will be able to keep up with that, at least, I hope so. So I'd say, for current owners, it doesn't really matter that much unless you love FEAR, and never play B&W2. For people that own neither, the choice is obvious, the 1900 is superior.

 

[ManicOne]

Word Werrrd!!!

 

[BBA]

I think you're the one not understanding something. Namely that it is very CPU-heavy because of its software shadowvolume and skinning algorithm. Which makes it CPU-limited when you put a card in that is 'too fast', like a 9600Pro in an 1800+ system. The same card performs much better on a faster CPU. Ergo, CPU-limited.
.

What I am saying is you put an X800XT in that same PC with the 1800+ CPU and the performance more than doubles that of the 9600 Pro because the 9600 Pro is slow and limiting the Doom3 engine. That means the video is limiting performance. Changing to an FX60 while using a 9600 will still give crap performance in Doom3. If it were CPU limited, the change in performance would not exist.

I don't know anyone who would describe any 9600 Pro as 'too fast' sinice it's pretty much entry level with today's games. Your description by itself takes even more credibility away from you.

Stop being argumentative.

Then again, I think it may be me...I failed to realize that my argument is in futility... some people can not be shown reality even if it runs them over on the highway of life.

 

[Scali]

See the thing is that your opinion (see below for my interpretation of it, correct it if wrong) is not supported by your "arguments", which are mostly opinions themselves, or by actual REAL WORLD gaming results.

Since my argument has been about future games from the start, it's obvious that there is very little real world gaming results. But 3DMark06 is an indication, and so is Doom3.

"The 7800 has 64bit Texture filtering and blending, and UltraShadow II, and those things are SO much more important to the future of gaming than any features the 1900 has that there is no way to ever consider the 1900 a better card."

Oh please, that is not at all what I said. No wonder you're getting so worked up, if you read things as such extreme statements. By the way, you forgot texture fetch, which is probably the most important, since it actually makes new rendering tricks possible.

The other problem with it is that it COMPLETELY ignores any current real-world performance results.

That's another thing you must have read wrongly. I never said the 7800 was or would be the faster card. I said that performance is high enough on both cards that I wouldn't really care which is faster. I just think the 7800 may be able to render some more pretty effects. Well, I know it can, it's just not being done in games (yet?).

Fetch 4

Now that's funny. That's actually their workaround for not having PCF in hardware. NVIDIA has a superior solution, so they will never need Fetch4 at all.

 

[Scali]

What I am saying is you put an X800XT in that same PC with the 1800+ CPU and the performance more than doubles that of the 9600 Pro because the 9600 Pro is slow and limiting the Doom3 engine.

Can you back that up with some proof? My experience was that the game dropped below 15 fps with 2 or more creatures on screen, regardless of resolution or anything. And a faster card also didn't make a difference. A faster CPU did, however.
A faster videocard may improve performance in places were there aren't a lot of shadows and skins to be calculated by the CPU, but that wasn't my point obviously, as I mentioned the shadows from the start.

I don't know anyone who would describe any 9600 Pro as 'too fast' sinice it's pretty much entry level with today's games. Your description by itself takes even more credibility away from you.

Sure it does. Go on insulting. I never said it was 'too fast' for today's games. I said it was too fast for Doom3 when being coupled with an 1800+ CPU or slower. Not that the card itself is too fast for the game, but that the CPU can't supply frames fast enough for the card (which doesn't have anything to do with the actual framerate being high, because it's not).

Then again, I think it may be me...

I'm quite sure it is, you don't seem to understand anything, yet you think you're right and the other is wrong. You'r both ignorant and arrogant. And I'm simply right about the 9600Pro and 1800+, it's easy to verify for anyone. You're just talking out of your arse, because you've never actually tried it.

 

[Fanofnone]

Your still at it Scali? You've singe handidly kept a dead thread alive, congrats. There can be no debate! The X1900XT is CHEAPER, it is FASTER and is AVAILABLE. It has the tri-fecta over the GTX 512mb. Stop arguing for a looser.

 

[ManicOne]

This thread seems to be becoming just an argument. My opinion is that X1900 is far more futureproof, given its 3:1 ALU:TMU ratio. Games have become primarily fragment shader bound, as opposed to basic fillrate limitations of the past. This will only increase with the advancement in shader tech we are seeing from both hardware and software developers.

Furthermore X1 series do have an alternative to VTF, being R2VB (render to vertex buffer). The VTF functionality on nV hardware is also extremely limited, being locked to only 1 (of a select few) ops per clock IIRC. The reason for this is that on non-unified parts full functioning VTF would take up a phenomenal amount of die space, totally out of proportion with the potential use of this feature during these cards expected use-spans.

 

[Scali]

This thread seems to be becoming just an argument. My opinion is that X1900 is far more futureproof, given its 3:1 ALU:TMU ratio. Games have become primarily fragment shader bound, as opposed to basic fillrate limitations of the past.

So far there's only one game, and that's Fear. It's too early to say if this is an exception or a rule in future games.

Furthermore X1 series do have an alternative to VTF, being R2VB (render to vertex buffer).

For the last time: they are NOT equivalent. You can't do the same with both technologies. They have a bit of overlap, but are not interchangeable! Stop pretending that they are!