First REAL FX Benchmarks

[Doomtrooper]

CPU limited benchmarks are not useless at all, highlights my points exactly in another thread. People assume that a video card can magically add frames if its limited by the platform, sure some driver optimizations may add 2-3 fps over another but still a factor.
So in the event of the 24 player team CTF match on UT2003, both cards are limited by the platform..which SHOULD affect the consumer looking to buy a new card.
If they see they are platform limited then its only a question whats going to give me the best value for my Money.

 

[KimB]

Botmatch scores may not be representative of online play as obviously the CPU is no longer crunching as much AI code as it is when playing in single player mode. Correct me if I am wrong.

That's the idea. One would have to look at the csv's to see just how much CPU time the AI code takes, though (I believe it's in there...).

 

[Doomtrooper]

Playing on a server with 24 players with bots enabled for team balance would be very close to CPU cycles being used IMO.

 

[WaltC]

*nod*

I agree. After all, DK mentioned the GFFX was "optimised" for 4xMSAA performance.

That's seems an extremely odd observation for him to make considering the fact that the only thing we've ever seen in GF FX benchmarks--even the slanted ones released by nVidia to date--are all restricted to 2x FSAA. That would be my biggest question mark concerning these numbers--why 2x FSAA *again*...? What's up with 4x AF? Hate to say it but these sound suspiciously tailored to me. (I also think the hit for 2x FSAA on the 9700P is way too high, as someone else pointed out.) It seems to me the first thing a reviewer would want to see is mode performance different from what nVidia has been handing out--4x, 6x, 8x FSAA, what sort of adaptive AF the drivers do above 4x AF, higher resolution performance, etc.,....why rehash "old" ground? (Which really isn't "old" but already made to seem that way by the nVidia handouts thus far.)

Is there anything about this product that isn't extremely murky?

 

[Tahir2]

Playing on a server with 24 players with bots enabled for team balance would be very close to CPU cycles being used IMO.

I suppose you are right!
I take back what I said

 

[DemoCoder]

hands bottle of Lithium to Walt

 

[KimB]

Playing on a server with 24 players with bots enabled for team balance would be very close to CPU cycles being used IMO.

The bot AI's only done on the server, so it would still depend on how much CPU power the physics and added geometry take up. But you're probably right, a match like that would almost certainly be CPU-limited...I doubt AI could take up that much processor power.

But, a many-player match is also more likely to be driver-limited (since much of the added processing would be due to added geometry) than the botmatches.

 

[nggalai]

Hi Walt,

why 2x FSAA *again*...? What's up with 4x AF? Hate to say it but these sound suspiciously tailored to me.

I mentioned it before--it was Thilos first round of benchmarks: 1024x768 without FSAA/AF, 1024x768 with 2xFSAA and 4°AF. Shortly after, he went offline to get cracking on the rest of the benchmark suite (he only got one to two days left to finish them all, after all). PCGH has, in the past, always been rather thourough with its testing, so I wouldn't look for a conspiracy here. Not as if it mattered; more reviews should pop up shortly, I'm sure.

Apparently, Thilo just felt the need to drop some first couple of scores, last but not least to actually make people still buy the magazine, I suppose. Teasing.

ta,
-Sascha.rb

 

[Ante P]

About 9700 Aniso in UT2003.
There's an insane difference between performance and quality aniso in this title.
I haven't seen more than one or two titles where the difference is this big and of those UT2003 is the worst case.

 

[Typedef Enum]

That's a good point actually. I noticed a large performance hit on the 9700 when you went with performance settings...much more than any other game.

 

[Doomtrooper]

I think you mean the other way around 'quality'.

 

[THe_KELRaTH]

Well here's my UT2 Bench test using a P3-1Ghz

1024x768 32bit: Flyby 86 Botmatch 31.47
w/x2AA x4AF : Flyby 83.87 Botmatch 30.16 (performance)
w/x2AA x4AF : Flyby 79.04 Botmatch 29.91 (quality)

erm.. so 7fps on a P3-1Ghz or 91fps drop using an AtlonXP.

 

[Ante P]

Flyby btw

From my Gigabyte 9700 Pro review..
(and yeah I actually verified that 2x performance aniso was faster than simple trilinear filtering, I've seen the same effect in other benchmarks too for that matter)

for the test rig specifics look here:
http://www.nordichardware.com/recensioner/grafikkort/2002/GV_R9700/index.php?ez=3

and btw, the difference in IQ between Performance and Quality isn't really noticable at all in UT2003 IMHO
so I don't see no reason whatsoever for anyone to use the quality mod ein this title

got some pretty nice results when I combined 16x Performance Aniso and 6x FSAA on the 9700 Pro and 8x FSAA and 4xS FSAA on the Ti4600

btw if anyone is wondering that's 75 fps on the 9700 Pro and 8 fps on the Ti4600
(with "normal" 4x FSAA the Ti4600 got 21 fps)

 

[Doomtrooper]

Performance mode is the same as the 8500's, bilinear filtering only mode...there is one reason why.

 

[Ante P]

Performance mode is the same as the 8500's, bilinear filtering only mode...there is one reason why.

yeah I know, so I'll rephrase it:
'8 tap aniso is faster than ("8 tap") trilinear'

 

[Doomtrooper]

No people won't notice a difference playing UT 2003...yet there was some websites that made a huge deal about it on the 8500..X-bit being one of them...AND STILL calling it ripmapping

 

[DemoCoder]

Hmm, I seem to remember making a similar point awhile ago about AA during play. But of course, if people screenshot the bilinear AF to death and nitpick pixels that don't look right, I'm sure Doom will be happy to say that people won't notice. But we if we can see any marginally flawed pixels in an NV30 screenshot, I predict everyone will say that are blatantly obvious and stand out like a sore thumb.

I agree with Doomtrooper, most people won't notice during play. I only hope a similar open-minded approach to IQ will be available during the NV30 reviews.

 

[Typedef Enum]

Oops...I said it backwards. Meant to say that Quality showed a significant penalty in UT2003, as opposed to Performance (Duh!).

 

[Nagorak]

I'm confused as to why the GF FX is slower in standard 1024*768*32, but faster with AA and Aniso? I thought the feeling was the GF FX would be faster stock, but slower with AA on. This seems to be the exact opposite of that situation?

CPU limited benchmarks are not useless at all, highlights my points exactly in another thread. People assume that a video card can magically add frames if its limited by the platform, sure some driver optimizations may add 2-3 fps over another but still a factor.
So in the event of the 24 player team CTF match on UT2003, both cards are limited by the platform..which SHOULD affect the consumer looking to buy a new card.
If they see they are platform limited then its only a question whats going to give me the best value for my Money.

With 4X AA and 16x Aniso, there's no such thing as CPU limited though.

Based on price/performance, I'd still say you're better off with a high end graphics card and a high mid-range CPU (since beyond that CPU prices just sky rocket) and with all the goodies turned on you're still not going to be CPU limited anyway. (I'm referring to completely FPS & FPRPGs only, though, I realize that this might not be the case in games like DS and WCIII.)

No people won't notice a difference playing UT 2003...yet there was some websites that made a huge deal about it on the 8500..X-bit being one of them...AND STILL calling it ripmapping

Personally I think that Quality is just a way to waste performance. You're not going to notice it while playing, unless you're not really playing. I guess it's just as well that it's there for those who want it.

 

[Windfire]

I don't understand why anyone would want the FX to fail. Even if you are an ATI fan, you've got to want competition. Sure, the FX is definately stumbling out of the gate, but as long as it delivers, WHO cares? Take a couple steps back and recognize that you're stance is often forced beyond where you would ideally want to be.

I hope the FX is awesome. In fact, I hope kicks ATI into more action too. Prices will go down, better models will continue to be announced (by boht companies) and hopefully some of this new tech will even trickle down to some real games we can play.

hehehe. I think it is funny/ironic how everyone gets into a tizzy and the sad fact is we won't really see any games that take advantage of these cards. Oh well, at least FSAA and AF will be nice.

My TI 4400 is doing great, but I'm starting to get that itchy "need to buy something" feeling. Maybe the FX or R350 will be enough to push me over the edge.