UT2003 GPU Shootout

http://www.anandtech.com/video/showdoc.html?i=1647&p=1

The results seem rather expected...But there are some pretty interesting quotes...

-He claims that next-gen parts will use 350-500 MHz. DDR. It would seem reasonable, given the expected releases, that R300 would be 350-400, while NV30 would be closer to 500 MHz.

-He claims that R300 will end up laying some serious smack in this benchmark.

The striking thing of note is how brutal the Ti4600 does in comparison to the 8500. In both benchmarks, when comparing HQ details @ 10x7, the Ti4600 darn near doubles the performance of the 8500. Of course, he also fails to apply any additional GPU-specific features too...

The other thing I didn't quite get was the "bandwidth test" benchmark...Is this a UT2003 benchmark, or is this something they cooked up themselves? In either case, the GF4...heck, even the GF3...really did a number on the competiiton.

Also...Does anybody believe, at this stage in the game, that UT2003 ends up shining on the GeForce4 due to...Well, think 3dfx...That is to say, that Epic really developed the engine in/around a specific chip, and it ended up making life very difficult for the competition? It would seem a little more far-fetched, given the fact that this thing was developed with DX8 in mind...But the numbers sure seemed favorable to the point of near disbelief.
 
I'm glad to see from these results that UT2003 isn't going to be totally CPU limited like the original UT. The Radeon 8500 line seems to face off pretty equally with the GF3 Ti500. The two cards trade a few wins/losses, but both end up being essentially equal. The R8500 buries the original GF3 though. I guess after all the driver optimizations, it really is a GF3 Killer. ;)

As for the GF4 performance...I really doubt Epic would stupid enough to base their game around a single video card architecture (again). They made that mistake with 3DFX, and then they went under... Although I don't think Nvidia is likely to go out of business, it seems unlikely they'd set themselves up to be burned a second time. Also, if you look at the scores, the R8500 is pretty close to the GF4s in medium quality, while in high quality it gets left badly behind. That leads me to believe that the GF4s can just handle some setting in high quality mode better. Maybe high quality automatically enables FSAA or something?

Either way, if the game was "Nvidia optimized" it would run way better on the GF3 Ti500 than the Radeon 8500, and you can see they are essentially equal. My understanding is there is not much difference between the GF3 and GF4, so if you optimized for one you'd help the other out to. So, I really doubt there's any showing of favoritism on Epic's part (wouldn't really make sense either, their goal is to have the game run well on as wide a range of systems as possible).

Maybe they just haven't gotten around to really tweaking the older cards yet? It sounds like the test is sort of out of date too. They have the 3DFX line of cards running fine in the game, yet they won't run the test. So, maybe take the results with a grain of salt.

One thing that is impressive about this benchmark is how high the scores already are. High resolution gaming is definitely becoming a reality. You can already run at 1280 in high detail with a GF4. If the R300 is faster than you should definitely be able to run in 1280 or even 1600!
 
Well at every benchmark at medium texture quality the R8500 is neck to neck with the TI4200. Once the texture quality is at max the R8500 falls behind.

If you take the first "bandwidth" benchmark into account, the R8500 did really bad and it is propably this low "memory performance" that causes the R8500 to fall behind. But i didnt think the R8500 was so bandwidht limited, I thought thier Z culling beat the GF4's hands down. And how much more bandwidth does the Ti4200 have???

Maybe it is some shitty mem controler??
 
From the article :
Although the game looks great at 640x480, in order to combat the aliasing effects of a high polygon-count game like UT2003 you'll want to move to a higher resolution.
Eh? Aliasing with hi-poly game should be lower at low res than aliasing with low-poly game at low res.

Otherwise, another Anand article that is PR material for the fact that they work together with Epic re UT2003 troubleshooting. :)[/quote]
 
Huh Rev? Is that a typo? What's your reasoning behind this? More polies = smaller polies == higher frequency == more edge aliasing. A game at 320x200 with 1 million polygons and no AA will suffer way more aliasing artifacts than a game with 1 million polygons at 1600x1200. A game with 4 polies at 320x200 will not have more aliasing than the scene with 1 million at the same resolution.
 
More polies = smaller polies == higher frequency == more edge aliasing
You are correct.
A game at 320x200 with 1 million polygons and no AA will suffer way more aliasing artifacts than a game with 1 million polygons at 1600x1200
You are correct again.

However, jaggies will be more pronounced at lower rez in your case but jaggies will be more in terms of "numbers" at higher rez.

Isn't this the age old argument that higher rez will reduce aliasing (specifically jagged edges), which is wrong?

I could be wrong however since I typed all of this after downing a few beers :)
 
Seems like my KyroII will be OK for UT 2003 :D!
At this test the KyroII is between a Geforce2 Ultra and Geforce4 MX440 :D! Not bad at all :)! There are playable framerates at 1024x768x32bit and High Detail Settings :)!

CU ActionNews
 
Yea,

those K2 scores in some settings on par with a GF2 Pro? Yea for a card with the specs of a TNT2!!!!

I was at the UT mod summit last weekend. Mark handed out a bunch of GF4 and 8500s to people there that needed them. He seem to have an even mix of both cards so I think UT2003 is not just nVidia optimized. However I did happen to snag a UT2003 T-Shirt and notice it had a nVidia logo on it :) :) :)

The mod teams will get access to the tools and I will see if I can post my results here. The new game offers a ton of possibilites. As well as some increadible IQ offerings.....
 
I will reserve my judgement on this until I have the benchmark myself and tested it myself, since Anand is the only website that shows a Ti 4200 beating a 8500 in Jedi Knight out of countless others that show the 8500 even tying a Ti 4600, I'm skeptical of these results.

Either he doesn't know how to configure a Radeon card or....
 
either way this is much better perfromance than I'm seeing on build 927 with my 8500, and the 8500 beats the Gf3Ti500 most of the time bearing in mind in the original Flyby Test the 8500 won the firts outing, then the Gf3 started to go past it significantly.

I'm as amazed as Type tho at the beating the Gf4 gives to every other card out there (inc. the Gf3Ti500).
 
Randell said:
either way this is much better perfromance than I'm seeing on build 927 with my 8500, and the 8500 beats the Gf3Ti500 most of the time bearing in mind in the original Flyby Test the 8500 won the firts outing, then the Gf3 started to go past it significantly.

I'm as amazed as Type tho at the beating the Gf4 gives to every other card out there (inc. the Gf3Ti500).

The last interview with Sweeny and Epic they stated they had not implemented Shaders or even decided on what shaders to support, I am wondering what specific card features are being supported by UT 2003.
 
AFAIK.

Upping the resolution is the most natural way of reducing spatial aliasing. It doesn't however solve temporal artifacts, and it doesn't give the distinct look of anti aliasing.

As far as spatial aliasing is concerned, AFAICS higher resolution >= antialiasing with the same samples. But thats in terms of mathematical signal clarity.

Antialiasing gives a distinct look that many people seem to subjectively prefer however.
 
> "The striking thing of note is how brutal the Ti4600 does in comparison to the 8500."

Interesting, as here where I live in Vancouver Canada, the Ti4600 card's are 3 times the price, and it is well known that the Ti4600 image quality is not as good as the 8500.

$179 ATI Radeon 8500LE 64MB
$239 ATI Radeon 8500LE 128MB

$485 Abit TI4600 128MB
$559 ASUS V8460 TI4600 128MB
$599 Gainward Ti4600 128MB
$509 Microstar G4Ti4600 128MB
$589 VisionTek TI4600 128MB

Prices in Canadian dollars. Above prices from a local store here, www.anitec.ca

Note that the Ti4600 is not of the same generation as the 8500, as it is a half generation ahead.

Comparing the two, would be like comparing the R300 with the Ti4600. Which people will go to lengths pointing out, that it is not fair.

Note: nothing above to neccessary dispute what you said, but wanted to put things in perspective.
 
Yes and thats after the Ti4600's have dropped recently, I check all my old wholesalers on their debut in Canada and some were over $700.00.
There is something strange going on in this test on high quality settings as the 8500 is competetive with the Geforce 4 cards until high quality is enabled and then the bottom drops out of the 8500.
I will be running a R300 anyways but with Medium Detail settings a 8500 owner will still get 107 FPS @ 1024 x 768..not bad for a $ 239 dollar card :)
 
I don't really see the point in buying a Ti4600 when the Ti4400 is so much cheaper, especially since it is easily overclocked to Ti4600 speeds.

Over here (in the UK) you can pick up a Ti4400 128Mb for £227 whilst the Radeon 8500 64Mb (built by ATi) is still £210, and you can grab a Ti4200 128Mb for as little as £162, and the 64Mb version for £133, all from dabs.com.

A price difference of £17 for an extra 64Mb of memory and what is basically a Ti4600 after overclocking. Makes no sense at all.
 
Back
Top