Geforce FX 5900 Ultra VS Radeon 9800 Pro 256 mb

[oldfart]

I am glad to see some close competition again. nVidia was top dog for too long. It was nice to see them humbled by the 9700P and the 5800U flop. It made them work harder. ATi has held the crown for awhile. Now it seems pretty even. This is good news for us consumers. I've owned more nVidia cards than ATi, and will buy either one. Whichever is best. I dont really care who makes it.

Beside top 3D gaming performance, the card has to have sharp 2D, a quiet cooling system (I prefer a variable speed fan). I also like what nV has done with different clock speeds for 2D and 3D mode. Cooling these beasts without a ton of noise is going to be ongoing issue. We've seen what doing it the wrong way caused. The 5900U looks petty good to me. $500 for the 256 Meg version is hard to swallow. The 128 Meg version is more reasonable if it gets down to the sub $350 range. Same with the 9800 P.

 

[Blastman]

I?m willing to concede that the 5900u is a faster card than the 9800pro. Nvidia did a pretty good job with 5900 and fixed the driver performance with AA+AF. The 5800 was getting hammered in AA+AF benchmarks. But considering that the 5900 GPU is clocked 18% faster and its memory bandwidth is 21% faster, I would hope it would be faster. The 9800 did a decent showing and a lot of the testing results seems to depend on what ? ?quality? ? settings were used for each card.

The 9800pro seemed to win 3Dmark 2001, the 5900U won at 3Dmark2003 and generally in the game tests. I sometimes wonder how much Nvidia is optimizing for the Hardware review sites game tests. At ExtremeTech the 9800 won their 3DGG (GameGage) tests - 8 games where the 9800 won 5 out of 8).

My impression also from the Xbit lab review is that running FSAA takes a ton of bandwidth at higher resolutions and in many cases where the 9800 won a benches at lower resolutions it would lose by a larger and larger margin as the resolution was raised.
Here?s a good example

Not only does the 5900 have faster memory but is has a new and better memory controller ? THG ? ?. ?A new memory interface (IntelliSample HCT) improves the efficiency of 4:1 color compression by 50 percent, which should benefit FSAA performance. ?.. ? ?. .?

I think ATI should have put faster memory on the 9800. It?s just not a big of enough jump over the 9700pro. With the same memory bandwidth of the 5900 and running the 9800 at 450mhz, it should be a good match for the 5900 if not better. The 9600pro could use faster memory too. It only has 300mhz DDR compared to the 5600U which has 350mhz. The 9600p currently outperforms the 5600u, but the 5600u is getting a speed up sometime this year too.

It was ATI that single handedly kept nVidia from becoming a monopoly. Matrox couldn't do it. Sis couldn't do it. No one else could do it. I've got a ton of respect for ATI after battling back from the Rage Fury days when they were badmouthed and just downright uncompetitive.

I agree, pretty well just ATI and Nvidia left standing now.

 

[BFG10K]

The 5900 Ultra seems to be a mixed bag, besting the 9800 Pro in some cases and losing in others. And from what I can tell ATi still has the superior image quality with FSAA and quality anisotropic filtering (lets face it, ATi can potentially sample up to double the amount of texels that nVidia can at their respective highest settings).

It's a tough call but I'd say that the 9800 is still the better card overall IMO, though the 5900 probably deserves the speed crown.

 

[BFG10K]

the 256mb 9800pro is a waste, pure novelty.

It's only a waste if you don't use the Catalyst 3.4s which are the only drivers from ATi that can see the extra 128 MB VRAM.

There are games right now that can crush 128 MB cards if you turn up the details high enough.

 

[Mem]

IMHO the FX 5900 Ultra is a better card then the ATi 9800 Pro, for one it has better drivers and they will only get better over the next few months .

Lets face it Nvidia are still top dog in this area,ATi may have slightly better image quality but it`s not a big gap,infact most people probably couldn`t tell the difference,all credit to Nvidia for trying to improve image quality,they have definitely improved IQ since GF2 days,as to performance well 5900 is ahead in most benchmarks.

The strong AA/AF speed improvement was also good to see.
My verdict is Nvidia are on the right track , they just need to keep improving in all areas and don`t do any silly mistakes like they did with the NV30.

The battle and War is never ending between the two companies.

 

[touchmyichi]

Has anyone here heard of how the NV 30 overclocks yet? So far on all of the reviews I've seen they haven't gone into overclocking the cards.

 

[slackerinabox]

Question from the casual gamer- has nvidia's 2D quality issue been fixed? I'll consider getting the value version of the 5900 (whenever it's released) if their 2D can finally rival ATI or Matrox.. I had a geforce4MX and really hated how the 2d was fuzzy and noticely less clear than my matrox or ati setups...

 

[Mem]

freedomsbeat212,
2D quality should have improved on th FX models over the older Nvidia models,if you go with the top brands(Nvidia wise) for best image quality ie Leadtek,Gainward etc you should be happy.

 

[Rogozhin]

Well

It seems from the latest info that the doom 3 beta build was leaked to nvidia for driver optimizations without a word to ati.

Therefore the beta build was rendered on the nv35 using the dx 9 non-compliant nv30 path and not the agreed upon arb2 path. The nv35 also defaults to fp 16 while the 9800 runs it at a fp24 increasing even more the discrepency in this psuedo benchmark..

This doom 3 fiasco is meaninless. because.

1.Doom 3 won't be out for 4-6 months
2.the 5900 can't be purchased.
3.by the time this game is released ati is promising a new product (and definitly new drivers).
4.The cat 3.2s were used (and these don't register the other 128mb of memory on the 9800), and neither do the cat 3.4s thereby validating the proposition that ati never had a chance to optimize its drivers while nvidia did.
In the rest of the benchmarks I've looked at the 5900 does poorly when comparing AA. The AA performance at 8x on the nv35 and 6x on the 9800 is not comparable the radeons looks MUCH better and is about 60% faster.

I would like to see the nv35 do well and so far it is a top performer with the 9800 pro BUT to use a beta as a benchmark is just asking for trouble especially if it can't be done apples to apples.

Rogo

 

[Tab]

I am abstaining to make a decision on the cards because...
1)Its not offically out yet
2)ATi's cat's aren't up to snuff
3)Its a reference card (You never know what Gainward will do )
4)Its phuqing Expensive! I'll be waiting for my GeForce FX 4200 Thanks!

 

[BenSkywalker]

Rogo-

Where are you getting your information? Some of it appears extremely suspect.

It seems from the latest info that the doom 3 beta build was leaked to nvidia for driver optimizations without a word to ati.

It wasn't 'leaked' to nVidia, id gave it to them. Last time id handed over a build of the game to ATi it ended up being available for download to the general public- that is a leak.

Therefore the beta build was rendered on the nv35 using the dx 9 non-compliant nv30 path and not the agreed upon arb2 path.

???? Where are you reading this? First off, the game runs under OpenGL, not DirectX. The ARB2 path is non DX9 compliant. Second, who agreed to run the game under ARB2? Carmack has been saying that the NV30 should be run under the NV30 code path for several months now, why should he change his tune now? Should he state things that aren't true because ATi looks bad?

The nv35 also defaults to fp 16 while the 9800 runs it at a fp24 increasing even more the discrepency in this psuedo benchmark..

And oddly enough FP16 is what Carmack was stating he wanted for Doom3 a couple of years ago. Let's take a look at what JC has to say about the quality difference-

The Nvidia card will be fastest with "r_renderer nv30", while the ATI will be a tiny bit faster in the "r_renderer R200" mode instead of the "r_renderer ARB2" mode that it defaults to (which gives some minor quality improvements).

Moving from INT8 and a DX8 level feature set to FP24 and a DX9 level feature set(this in no way indicates compliant with DX9) results in some minor quality improvements. That is a quote from Carmack in Anand's review.

This doom 3 fiasco is meaninless. because.

If you believe that then you are delusional The Doom3 benches did exactly what they were supposed to do for both id and nVidia. id had a huge round of press for their game leading in to E3 while nVidia got to show off its board looking extremely good in one of the biggest titles coming to the PC this year.

1.Doom 3 won't be out for 4-6 months
2.the 5900 can't be purchased.

I had to put these together. If you are buying a board right now looking to play Doom3 what do you think non rabid fanATIcs are going to buy?

4.The cat 3.2s were used (and these don't register the other 128mb of memory on the 9800), and neither do the cat 3.4s thereby validating the proposition that ati never had a chance to optimize its drivers while nvidia did.

The Cat3.4s recognize the RAM just fine. Look at {H}'s review and see how badly the R9800Pro gets throttled to hell and back when they use those drivers for the game. The reviewers were doing ATi a favor running the Cat3.2s instead of the 3.4s. As far as the validation in terms of optimizations, we saw the same thing in the alpha build running launch drivers from the FX 5800U looking at comparative performance. The drivers used are available to scrutinize for anyone who wants to, at least for nVidia. Can't say the same for ATi, although looking at the issues with the Cat3.4s I'm not sure they really should be either.

In the rest of the benchmarks I've looked at the 5900 does poorly when comparing AA. The AA performance at 8x on the nv35 and 6x on the 9800 is not comparable the radeons looks MUCH better and is about 60% faster.

Comparing SSAA to MSAA, could you point me to who does this with a straight face? 8x is there to enable SSAA in older titles who need AA for alpha textures, something ATi can't do right now. As far as looking MUCH better, try playing a game @1600x1200x4x AA on the R9800Pro and the R5900U and talk about the difference.

I would like to see the nv35 do well and so far it is a top performer with the 9800 pro BUT to use a beta as a benchmark is just asking for trouble especially if it can't be done apples to apples.

List off three apples to apples benches that have come out in the last six months.

 

[slackerinabox]

What's the point of this silly "Doom 3 benchmarks were rigged!" discussion? The new card beat out the 9800 in almost EVERY benchmark, not just doom 3.. Stop with the silly fanboy propaganda.. Nvidia has the upper hand now.. ATI will top them with their next card.. Then Nvidia will top ATI.. Then....... The good thing is that each of these cards will have value lines that will be hopefully faster than the previous generation's high end card

 

[TypeM]

I have an R300 myself, but I must ask how it was rigged in NVIDIA's favor (Quoted from John Carmack):

We were not happy with the demo that Nvidia prepared, so we recorded a new one while they were here today. This is an important point -- while I'm sure Nvidia did extensive testing, and knew that their card was going to come out comfortably ahead with the demo they prepared, right now, they don't actually know if the demo that we recorded for them puts them in the best light. Rather nervy of them, actually.


Sounds like it was an even playing field to me.

 

[nRollo]

Frek:
"You'll be eating those words by the time DIII does come out. I give at least 80% odds that ATI will have the top benchmarking card for DIII when it actually comes out."
LOL You base this "80%chance" on what, since you have ABSOLUTELY NO WAY OF KNOWING WHEN DOOM3 OR ATIs NEXT CARD ARE COMING OUT?!?!?!
I give at least a 99% chance you're talking out of your a$$.


"What they do say is that its too close to call the 5900 vs. 9800. All the reviews I have read all say the same thing, its too close to call because the benchmarks flip back and forth between the 5900 and the 9800. The 5900 by no means OWNS the 9800"
Please share your drugs with me- it's been way too long since college for me. Well, gee, Frek, let's see:
Anands review, 5900 wins 18/21 benchmarks
Hmm, I see what you mean. The 9800 won Commanche...and...er..Commanche...and....Commanche! LOL, does anyone even play this game?

More "too close to call" at Toms, the 5900 only wins 35/42 game benchmarks!
It would be preferable to own the card that wins 7/42, right Frek?

No clear winner at FS, only 24/36 won by 5900!
(I didn't consider the botmatch UT2003, it's pretty well known it's not a good VGA benchmark, although a 9700 MAY be faster than a 9800

)

"In my opinion(as well as Anandtech) its way too close to call."
So how does Anand show it way too close to call when the 5900 is winning 18/21 benchmarks, some by a LOT?

"Will I be buying a 5900? Hell no, my 9700 Pro plays every game I have at INSANE frame rates"
Good for you. I sold my 9700 Pro today for $225 and will be buying a 5900 Ultra. Then I'll play at PSYCHOTIC framerates! LOL

 

[nRollo]

Gururu:

I agree that the NV30 was late, I meant that the NV35 may have been rushed to make-up for the NV30

Where do you get this "rushed" stuff? You have NO IDEA if it was "rushed"?

But I still ask whether it is too little too late, because Nvidia has lost video card domination. This is good for the industry, but not for Nvidia investors. But, if I were in any way involved with Nvidia (employee, shareholder), I'd have to feel like so much ground has been lost.

HUH?!?! You do realize that nVidia totally OWNS ATI in terms of market share and profitability, don't you?

s the 5900 too little, too late? Technologically its late, because we've had
similar performance to it for nearly a year already.

What the? A 9700Pro offers "similar performance"?!

 

[Wurrmm]

Originally posted by: BFG10K
The 5900 Ultra seems to be a mixed bag, besting the 9800 Pro in some cases and losing in others. And from what I can tell ATi still has the superior image quality with FSAA and quality anisotropic filtering (lets face it, ATi can potentially sample up to double the amount of texels that nVidia can at their respective highest settings).

It's a tough call but I'd say that the 9800 is still the better card overall IMO, though the 5900 probably deserves the speed crown.

Some cases??? As Rollo has so elquently shown, the FX5900 Ultra has won most tests. Anand has also show that image quality is practically even and one would be hard pressed to tell the difference. The 5900 Ultra even has somewhat better IQ. For a game such as myself, this is the card to go with.

 

[gururu]

Rollo, you seem to have a wonderful time looking at individual words and not context.

In regard to the possibility of NV35 being rushed:
did I say it was rushed? no. I said I thought it was rushed. Now, since I am obviously not free to voice opinion without hard facts, which would essentially de-opinionize my statement, let me elaborate.
With reference to the NV30, Nvidia claims to have made a mistake (read Anand's review). IMO, it was an EMBARASSING mistake that Nvidia wanted cleared up FAST. No company worth its salt sits on a mistake like that. Jen-Hsun Huang cracked down, put on the pressure. They're back in the game. Enough said.

In regard to Nvidia marketshare:
did I mention marketshare? no. I said Nvidia lost video card domination. Its clear in the rest of my post that I was obviously talking about the speed crown. Nvidia has for years taken great pride in the performance of their products and constantly boast about it in shareholder meetings. This is not unusual, all the companies do it. But with the NV30, nvidia, like so many others before them, had to blow smoke out of their ass. But it was clear, whether spoken or not, that they had to hustle on the new product. I'm sure these things hurt investor confidence (hurt mine) and shook worker morale. But its not so much that they fumbled the NV30, its that they have a REAL competitor now. Who can dispute? The future is not certain.

In regard to 9700/5900 performance:
UT2K3 (Anandtech, maxed out)
9700pro: 53 fps
5900ultra: 68 fps

Now, seriously, 15 fps improvement a friggin year later... give me a break.

 

[nRollo]

Gururu

" No company worth its salt sits on a mistake like that. Jen-Hsun Huang cracked down, put on the pressure. They're back in the game. Enough said. "
The card has issues, but with the new drivers it's arguably the second best card you can buy today. That isn't an "embarassing mistake". Like I said, TSMC made the mistake, not nVidia. nVidia can't control supplier performance.

"But it was clear, whether spoken or not, that they had to hustle on the new product. I'm sure these things hurt investor confidence (hurt mine) and shook worker morale"
Where do you get this stuff? How many workers do you know? How many investors? Don't you think the workers know what's in the development pipeline? Or that nVidia is PROFITABLE? I'll tell you what Gururu: If I'm working at very profitable huge market share nVidia, I feel my job is a lot safer than working for losing millions every quarter, tiny market share ATI. It's making money that keeps my job secure, not whether my company has a card that's a little better in a tiny market niche.

"Now, seriously, 15 fps improvement a friggin year later... give me a break."
A. It's 9 months later
B. It's almost a 30% improvement
C. It's one benchmark. There are others in the review where it's 25-50% faster. That is not inconsiderable. But I'm sure at you're job your always bumping your productivity 25-50% every nine months?

 

[gururu]

Rollo,

I'm seriously not disregarding your views. I just enjoy arguing because the subject is interesting.

Nvidia chose to go with 0.13u for the NV30. Though commendable, it was a mistake. It was Nvidia's choice. ATI is already producing on 0.13. I concede that these things can't be predicted. Everyone takes a bad road now an then. TSMC may have the role you suggest, but do you really think Nvidia nailed all the first designs, considering it was newer tech?
NVIDIA's CEO Jen-Hsun Huang said, "We can't make the same mistake twice. When you're in a tough spot, you focus on getting the job done."

You keep bringing in ATI into this marketshare argument. For all I care, ATI can crash and burn. I'm focused on Nvidia. The future of Nvidia. Stiff competition compromises present and future security.
Nvidia's not going to burn out. At least not tomorrow, I know that. But do you really think they are not taking their competition seriously?

Of course its an improvement. And I realize that 'similar' is a subjective term. It is one benchmark, sure.
IMO its the most relevant. The numbers to me are similar enough or at least not different enough to
justify calling the 5900U a major improvement on a card that came out 9/02. We are looking at 9.5+ months here.

Again, I enjoy reading your posts, particularly when you and Rogo get into it. Respect your opinion too.

gururu

 

[cmdrdredd]

Originally posted by: Rogozhin
Well

It seems from the latest info that the doom 3 beta build was leaked to nvidia for driver optimizations without a word to ati.

Therefore the beta build was rendered on the nv35 using the dx 9 non-compliant nv30 path and not the agreed upon arb2 path. The nv35 also defaults to fp 16 while the 9800 runs it at a fp24 increasing even more the discrepency in this psuedo benchmark..

This doom 3 fiasco is meaninless. because.

1.Doom 3 won't be out for 4-6 months
2.the 5900 can't be purchased.
3.by the time this game is released ati is promising a new product (and definitly new drivers).
4.The cat 3.2s were used (and these don't register the other 128mb of memory on the 9800), and neither do the cat 3.4s thereby validating the proposition that ati never had a chance to optimize its drivers while nvidia did.
In the rest of the benchmarks I've looked at the 5900 does poorly when comparing AA. The AA performance at 8x on the nv35 and 6x on the 9800 is not comparable the radeons looks MUCH better and is about 60% faster.

I would like to see the nv35 do well and so far it is a top performer with the 9800 pro BUT to use a beta as a benchmark is just asking for trouble especially if it can't be done apples to apples.

Rogo

Carmack codes in OpenGL though

 

[Keysplayr]

Quoted from Rollo: "Will I be buying a 5900? Hell no, my 9700 Pro plays every game I have at INSANE frame rates"
Good for you. I sold my 9700 Pro today for $225 and will be buying a 5900 Ultra. Then I'll play at PSYCHOTIC framerates! LOL

LOL Rollo....

 

[Rogozhin]

Fellas

sorry about the dx-opengl confusion, didn't even look at it after i wrote it until now.

If your purchasing decision is based on PREVIEWS with hardware and drivers that aren't availiable then something is wrong with you, maybe to much grass in college eh rollo

Plus, even anand says this:

"From the ATI camp the $499 Radeon 9800 Pro 256MB, just like the NV35, is a difficult purchase to justify; even more difficult in this case because the GeForceFX 5900 Ultra does outperform it in a number of tests. The benefit of 256MB of memory is quite small right now, and the price premium is definitely not worth it. We'd also honestly hope that the issues we encountered with our Radeon 9800 Pro 256MB board were limited to our board alone, and aren't representative of ATI running their memory out of spec.

There are rumors of an answer to NVIDIA's NV35 in the works for an introduction in the relatively near future based on the R3xx architecture. We'd assume this NV35 competitor to remain a 0.15-micron part, but with higher core clock speeds and maybe a few micro-architectural tweaks/enhancements to gain the performance edge on NVIDIA.

The next-generation of games are finally on their way, and it looks like we'll have an extremely powerful set of cards to choose from when titles like Doom3, Half-Life 2 and Deus Ex2 hit later this year."


He's basically saying 'unless you're a mad man or x-pot smoking college grad' there is no reason to buy EITHER of these cards when they come out (the 9800 is already out), since the games that will take advantage of this power won't be available for at least 3 months and by that time the prices for these cards will be lower, and the drivers will be optimized and stable.

Rogo

 

[nRollo]

Rogo:

He's basically saying 'unless you're a mad man or x-pot smoking college grad' there is no reason to buy EITHER of these cards when they come out (the 9800 is already out), since the games that will take advantage of this power won't be available for at least 3 months and by that time the prices for these cards will be lower, and the drivers will be optimized and stable.

Easy for Anand to say- he gets the cards free, and long before we get a chance. I'm guessing that if the entrepreneurial bug wouldn't have bit Anand and he was still just a "normal gamer" like the rest of us, he'd be buying these cards as well. I don't think you starting working your tail off in high school reviewing computer hardware if you don't like to play with hardware a whole lot.

 

[chizow]

My main concern with ATi has always been related to compatibilty first, and performance second. Recently, ATi's significant performance lead over nVidia persuaded me to give them another shot despite my concerns, but now there's simply no reason to choose ATi over nVidia with the release of NV35. The news of compatibility issues in next-gen games from the various NV35/DooM3 previews is *very* disturbing to anyone who has been following the gaming/video card industry. The next generation of games and engines from which the mainstream will be built upon shine on the NV30/NV35, but suffer on the R3XX boards. The R3XX has numerous issues with Splinter Cell (lighting, shadows, AA, etc.) which is arguably the best showcase available today of what the future of PC gaming holds. DooM3 (or should I say its engine) on the R3XX boards suffers in comparison to nVidia's next-gen cards, which is signficant since we all know the impact of a Carmack engine on the PC gaming industry.

I'll be looking to pick up an NV35 as soon as its available. The 9700pro was certainly a significant card as it raised some of my gaming expectations, but I still find myself frustrated with ATi compatibility issues (PlanetSide atm) and reactive driver fixes months later. In hindsight, the 5800U wasn't a poor performing product, the major problems with it were related to noise/heat, image quality, and its sub-par performance with AA and AF enabled compared to R3XX. With new drivers, its performance and IQ don't even look so bad anymore. On the other hand, the NV35 leaves no doubt in claiming the performance crown in all the resolutions and settings "someone buying a $500 card would care about", while correcting all the problems that made NV30 a failure.

Chiz

 

[Rogozhin]

Hm

Rollo I don't quite understand what you're saying in that last post.


chiz

I have no problems with splinter cell on my 9700, shadows run at highest and they work, check out this url for the 9800 pro pulling ahead of the 5900 in THAT game-http://www.hardocp.com/article.html?art=NDcyLDEw

Anand, hard, digextreme, b3d all state that ati's drivers are ROCK solid, so it's hard to take what you say as viable when the major sites and myself are concluding that ati's drivers are very strong therefore compatability too.

rogo