GeForce FX 5200 Ultra & 5600 Ultra Benchmarks

Very weird !

Look:

http://firingsquad.gamers.com/hardware/geforce_fx_5200_ultra/page10.asp

Overclocking the GFFX 5200 Ultra from 325/325 default to 335/335 increased the default 3DMark 2001 score by 31 points ? That's a bit too low IMO.

The individual test results aren't very good either
http://firingsquad.gamers.com/hardware/geforce_fx_5200_ultra/page11.asp

The game tests after that didn't show much increase.

I got much better increase when OCing the GF3 or GF4 TI. Heck even the GF4 MX I had got better increase than that.

 

Exactly, and that would be $149 MSRP for the Radeon 9600

 

OK, had a bad night after my post, so didn't finish the article at FS until just now.

I am disturbed by how often tests fail to have their resolution specified. Presumably, it shouldn't matter for some tests, but with all the optimization comments nvidia has made it is disturbing that there is no analysis of the viewed results going on, apparently.

I am also disturbed that all the results for the "other" cards shown are 64 MB cards (though I didn't notice the GF 4 MX amount being specified), and the GF 5200 Ultra had 128 MB...really brings some specific test results into question.

I continue to be disturbed by the use of 5200 Ultra (even overclocked!) for benchmarking against the 9000 Pro since, correct me if I'm wrong, the pricepoint for the 5200 Ultra isn't $79, but around $149 (around the same as the 9600 non pro AFAIK), and, for example, the 9000/9200 will presumably be targetted significantly lower at launch (this is the bait & switch I referred to earlier). This comparison regardless of price (AFAIK) is even a key point of emphasis in the conclusion.

We do finally get UT2k3 botmatch comparisons, but, again, the 8500 and 9000 Pro cards used seem to be 64 MB cards...

BTW, is Villagemark a 16-bit only test?

 

I believe the whole idea is to get OEMS to use the 5200 and the 5600 Ultra in their base or add-on products. Getting these cards in the mass market would get game developers to move foward.

At this point I agree that ATI is the leader and NVIDIA needs to come up with something to get the lead back.

But for a company to survive they need to have the OEM market which NVIDIA has and plans to keep.

 

toms review

http://www.tomshardware.com/graphic/20030311/index.html

even has radeon 9200 (non-pro) scores included!

 

Re: toms review

Likely the 5200's (non-Ultra's) competitor, judging from the benchmarks. Definitely a sub $100 card - it had at least better be.

 

Re: toms review

As well, look at those Splinter Cell benchmarks - YIKES.

DX9 capable? It hardly seems like the 31/34 are DX8 capable at this point.

 

Re: toms review

Well I don't really expect it to be more expensive than the R9000 it's going to replace (well probably slightly more expensive at the beginning), and since street price of the R9000 is already down to 60-70$ (even for the dual head cards)...

Yes but everybody gets really bad frame rates - just barely playable on the R9800pro with high quality settings, but without aniso/AA!
However, keep in mind this is a beta. It could get faster (remember the UT2k3 test some sites used, IIRC the release was quite a bit faster)

 

Re: toms review

That's not my point - look at the 5200/5600 Ultra scores in comparison to the 4200.

 

Village mark requires the command switch -32 bpp for 32-bit, I won't comment on the rest of article.

 

Looks like it's clocked at 250/225 (or at least the Gigabyte 9200 NP is).

 

Yay! nvidia "balanced" = ATI "quality", and nvidia "aggressive" (renamed to "performance", apparently? Strange how that monicker makes the comparison seem automatically apples-to-apples...) = ATI "performance"!

Also, he says the 9000 has problems with 4x AA, but he also skipped the no-AA aniso comparisons, and leaves the implication that the reason is the same "ATi's driver problems in UT2003."

Lars doesn't seem to have completely transitioned to thinking for himself, though he seems to be making large strides in overcoming bias, or atleast in overcoming overwhelming the reader with it.

Plenty of other flaws, but I don't really see the point in spending time going into more detail. On a positive note, for a change, only one heavily biased sentence in the conclusion that I noticed, and even that seems within the realm of simply being a conflict-with-my-opinion and possibly a language translation issue.

 

The 9000 uses Supersampling AA the lower performance is because it is applying AA to everything resulting in better AA IMO. Correct me if I am wrong here the reason it performs at lower frames is not a driver problem but the result of using a more complete AA solution (eg. Supersampling handles alpha textures as well.) and it requiers more bandwidth then the more common MSAA (that applies the AA to only primitives.) resulting in lower frame rates. But you would think that this was common knowledge at this point.

 

It is common knowledge (tho I didn't know MSAA applied AA only to primitives...), but Lars isn't complaining about low performance--he's complaining about no performance (thus the "driver problems"). Perhaps he's using 64MB 9000Pro's.

 

[action]pulls foot out of mouth[/action]

 

I think they should rename the Geforce FX the Geforce FiX, it clearly needs alot of work and there shouldn't be any thought in purchasing an ATI board over a Nvidia one atm. Personally I think the terrible performance can be directly related to an inside job by previous 3Dfx employees Revenge is sweet.

 

Huh? I think.....NOT! This is not a very smart comment. I would reconsider your motives before posting this kind of drivel...... I'm sure no one here wants or needs to hear this kind of stuff.....

 

For the love of Tarolli, I think it's pretty obvious it was a joke.

 

another one (5200 Ultra only): http://www.hothardware.com/hh_files/S&V/gffx5200u_benchmarks.shtml