Radeon 9700 Benchmarks...

[DGhost]

in the "i can be like Kyle and talk lots of shit about a product that hasn't been released yet" fashion, i have to post a link to this on gamespot, showing an alienware P4 system w/ a Radeon 9700 being benchmarked.

its kinda funny how it does so well in 3dMark but in UT2003 performance test it even can hardly beat a Ti4600 - only 12fps max, even w/ FSAA enabled. Wolfenstein 3d shows a more realistic picture, with it failing to beat a 4600 at 1024x768 unless FSAA is enabled, beating it by a decent margin at 1280x1024 and then failing again at 1600x1200.

kinda limited benchmarks, so it is a little early to tell, but it looks like without FSAA enabled it really doesn't smoke the competition.

suprise suprise.

[Dr Mordrid]

3DMark tests mean nothing if it isn't optimized for the card being tested, and even then its value is questionable given their connections to nVIDIA and the cheats their cards are routinely given in its routines.

Dr. Mordrid

[Indiana]

Re: Radeon 9700 Benchmarks...

Originally posted by DGhost
...but it looks like without FSAA enabled it really doesn't smoke the competition.

suprise suprise.

Yeah, surprise, surprise. Ever thought of CPU-limitations?

A GF4 4600 has a really hard time to manage to outperform a GF4 4200 at those lower resolutions and/or without AA. And this even on a >2GHz CPU - so what exactly do you expect from any other card that is faster than the GF4 Ti4600?

Take a >3GHz Pentium4 or a Hammer in there and you'll see. Besides the newest (unfortunately not yet publicly released) 6143 drivers do make a massive difference to the 9700s performance and GameSpot doesn't even mention the driver versions used AT ALL...

EDIT: BTW, shouldn't this rather go in the Benchmarks section?

[DGhost]

ahh, see, the beauty is that it doesn't smoke the GF4 at a higher resolution, either.

i just think its a little bit hypocritical to defend this now if you bashed the P before it was even released....

edit: possibly benchmark section, thought general hardware was good as its not a Matrox card and this is a Matrox forum...

[Greebe]

Not in my opinion. Nobody (Joe Public) has the card, the drivers used aren't mentioned and the results cannot be verified.

[Indiana]

So let's put it in easier words: 1024 is heavily CPU-limited with anything >GF4 Ti4200. 1280x1024 is CPU-limited with a GF4 Ti4600 more or less. Even 1600x1200 is partly CPU-limited on a GF4 TI4600 as long as you don't turn on quality-improving features like anisotropic filtering and AA.

To clean those results up a bit:

UT2003:
1280x1024 (partly CPU-limited):
GF4: 40.31fps
R9700: 49.8fps
That's a 24% advantage for the 9700.

1024x768, 4x AA:
GF4: 32.6fps
R9700: 42.4fps
A 30% advantage.

1600x1200:
GF4: 27.6fps
R9700: 38.6fps
That's already 40%.


RtCW:
1280x1024 (CPU-limited):
GF4: 125,1fps
R9700: 142,3fps
Only 14%.

1600x1200 (with this game, even this is CPU-limited):
GF4: 110,5fps
R9700: 123,2fps
Only 12% advantage, which is not really what one would expect. A driver-quirk maybe? (remember, this is an ATI card...)

1024x768, 4x AA:
GF4: 80,2fps
R9700: 140,2fps
A 75% lead for the 9700.


There are no more benchies but the 9700 does really look well in this very limited (and as Greebe said very unproven) overview, so there surely wouldn't be any reason to defend the 9700 here. Why should anyone "defend" a gfx-card anyway? Stuck in fanboy thinking? Just take the facts and put a little logic on those scores.

P.S.: I did not bash the Parhelia before it was released, I dind't even "bash" it afterwards. I'm a bit disappointed majorly by its poor anisotropic filtering, yes. Where does the Parhelia come into this thread anyway?

[MadScot]

I will wait for it to be released before I make any conclusions about this, or any other card. Does anyone even know yet the final core/mem speed?

MadScot

[borat]

whichever way you split it the 9700 has the most horsepower and arguably features plus aniso should work very nicely (MARTOX are you listening?) its not a surprise that it doesnt work much better than a gf4600 without fsaa as no body can use 200 fps anyway, its going to be faster than parhelia just face it. parhelia has the best 2d quality and will be most stable and through its clever faa will keep apace with gf4 when aaing but aniso does not work properly and its sheer horsep[ower is lacking, still a lush card though and i am getting one in my new system next week(p that is!)

[az]

Wow, don't you have a full stop on your keyboard?

AZ

[thop]

the 9700 manufactured by sapphire (who made all BBA's) will likely also have excellent image quality. on par with the P? probably not, but still very good.
and i'm curious to see how FULLSTREAM will be accepted in the industry.

[az]

BBAs?

AZ

[Indiana]

BBA = Built-By-ATI

EDIT: BTW, I would not be so sure about the image quality. The Radeon8500 has worse signal quality than the old Radeon and the 7500 and this includes the BBA models. The Hercules R8500 card according to tests has decent 2D, but has no real dual monitor capabilities. Crappy, but that's unfortunately just the way things are.

So we really have to wait until real cards appear before making assumptions on signal quality.

P.S: @MadScot: ATI has officially presented the Radeon 9700PRO (the "PRO" is another marketing BS, probably meant for leaving space for lower clocked non-Pro versions) and it will have 325MHz GPU and 310MHz RAM clocks.

[isochar]

Here's an official (German) review of the 9700 with benchmarks:

http://www.chip.de/produkte_tests/unterseite_produkte_tests_8788088.html

[Indiana]

DOH! This has to be one of the worst tests made EVER.
They're using totally different quality-settings between the candidates, e.g.:
- Simply setting anisotropic levels and not regarding the differences in naming-conventions between different manufacturers.
- Comparing 6x FSAA on the R9700PRO and 16x FAA on the Matrox Parhelia to the crappy, lousy looking Quincunx of the GF4 is meant as a joke, or?

What this test shows, however, is that there still are some driver problems at ATI - surprised, anyone? - see e.g. the SS TSE benches and some other strange results, IF the tester was able to setup his machine right, what's questionable considering the quality of his review. This beats THG by miles

A really well writen article concerning R300/NV30 (but without any benchmark figures) can be found e.g. here Finally an author that seems to know what he's writing about.

[Indiana]

It appears Gamespot have drawn the R9700 results? (Maybe they got a nice email from ATI....)