NVIDIA's GeForce 6800 Ultra GPU

Quote:

---

"NVIDIA have surprised me with NV40. It's rediculously fast compared to R360 in some tests, if you avoid the 8X supersampling mode. But there lies the rub. Shouldn't we expect good high-end antialiasing performance with their latest generation GPU?

DX9 performance, at least theoretically, is very strong. The shading performance of NV40 is mighty in some respects. It's got all the features NV3x should have had (and R360 already does in some respects), like multiple render targets and floating point support for everything.

Overall, I'm appreciative of the features and the performance. It's nice to see NVIDIA get it more right than they have recently. Shader Model 3.0 performance will be nice to evaluate in the future.

A fast wee beasty with excellent features, we now wait for ATI and R420."

---

http://www.hexus.net/review.php?review=747

had a quick scan over it, in 3dmark 2001 it doesnt improve too much, but still a nice boost over the 9800xt

in 2003 - you can tell it was optimsed for direct 9 software it jumps a LOAD

not too sure about with aa and af tho, cos the 9800xt seems overtake to it when there cranked up high (and even the fx5950 beats a radeon 9800xt at high af/aa) so that isnt pleasing

good for low end, bad for high end - id put it like that from just a quick read, will have a longer look over now

Intresting. Cant wait to see how it compares with ati's r420. Seems as though it uses a bit too much juice with nvidia reccomending a 480w psu!, two connectors, seperate rails. I wouldnt mind tesing it out with farcry :D

Some very impressive bench marks from this power house though. 12,000 3D Marks in 3D Mark 2003, on a system that gets roughly 6100 with a 9800XT.

Also runs Halo (well dodgy port) 50% faster than a 9800XT and 57% faster than a 5950 Ultra.

Must read more reviews... :D

those are some pretty damn impressive numbers.

Have a few more then :D : A score of over 60.000 in AquaMark 3. Over 60fps in Halo at 1600x1200 (bearing in mind how dodgy and crap this game runs) and more than 50fps in FarCry with High FSAA and 4tap anisotropic filtering at 1024x768.

Main selling point is the size of the leap in performance. 9700pro - 9800 - 9800pro - 9800XT, all better than the last, but in small leaps. Same with Nvidia's 5900, 5950Ultra's, etc.

All of a sudden, the jump is closer to 50% instead of 15%, in some tests. Need to read up a lot more, but the benchmarks are impessive.

wait for the r420....might not be so impressive then ;)

its quite shocking, how in AF and AA, the 9800XT kicks its ass in most benchmarks, where as the 5950 did the same to the 9800XT

think this shows the 9800XT is better at high power than the 6800, but the 6800 really does have the crap behind it in low res no aa/af

Wait for the final, retail, final, completely final, final version though (and ATi's offering, as you say) before all that. They may (and really should) iron out a few of those problems.

Quote's such as "Compared to the GeForce 6800 Ultra, the former high-end models Radeon 9800XT and FX 5950 Ultra often seem like nothing more than cheap mainstream cards... " from a respected hardware review site, tell me that this card has a bit of grunt under the bonnet.

Now, what can ATi do, the little scamps?

I must say that I found the resolution and AA settings in the Hexus review to be very poor choices to demonstrate this card's true power.

First, the review itself notices how terribly inefficient the 8xFSAA setting is. As such, I find it unlikely many users will be willing to pay such a huge price for a relatively small improvement in image quality. Therefore, I think the 8xFSAA benches are of little practical value. I have no qualms about including 8xFSAA benches for comparison, but by all mean they shouldn't be the ONLY setting tested at any particular resolution.

The bigger problem is, once you discard those 8xFSAA benches, you are left with only 1024 x 768 resolution tests. This is ridiculous! Why not test in 16 color mode while you are at it? It would be fine to test 1024 x 768 all the way up to 1600 x 1200 with the SAME AA and AF settings... but to have ONLY 1024 x 768 tests at a practical AA setting is RIDICULOUS. Running the 6800 at 1024 x 768 is like taking the Concorde from Los Angeles to San Diego! You are NEVER going to stress this card in any meaningful way at this resolution.... and worse, the R360 isn't stressed either which means we are given no clear idea of the performance difference between them. Nobody is EVER going to run this card at 1024 x 768 in the real world. I really doubt 1280 x 1024 will even get much use in the real world on this card.

I know you've put a lot of hard work into this review, so I'm sorry to be so critical, but I just feel like all that had work was wasted by a couple poor choices on the test configuration. Test all resolutions at the same AA and AF settings. Or at the very least, if you are going to change the settings between resolutions, recognize settings that customers are going to find impracticle and don't make them primary data points in your graphs.

Quote:

---

Originally Posted by Swafeman

its quite shocking, how in AF and AA, the 9800XT kicks its ass in most benchmarks, where as the 5950 did the same to the 9800XT

think this shows the 9800XT is better at high power than the 6800, but the 6800 really does have the crap behind it in low res no aa/af

---

I said it all above, but your comment highlights the poor choices in this review. The graphs in the review portray the 6800U performing badly with AA and/or high res. The fact of the matter is that the 6800U does extremely poorly with on one particular setting: 8xFSAA. That setting is an anomoly not the rule. With 8xFSAA, rather than using the typical (and efficient) multi-sample approach, the 8xFSAA does most of its work using the techincally superior but very costly super-sample techinque. In other words, at 1280 x 1024 the card is actually rendering the frame at 2560 x 2048 (!!!) then taking that frame and doing its normal 2xAA multisample AA on each pixel. Finally, it takes 4 of the pixels from the oversized scene and combines them into one pixel in the final 1280 x 1024 scene. As you might expect, though this technique produces excellent FSAA results, it is just not practical for gaming. My guess is that this techinque may have been included for 3D applications that do not depend so heavily on framerates.

The bottom line is, as the Hexus review will tell you, the 8xFSAA performance is an anomoly as far as performance. The review's own numbers show you that 4xFSAA runs 3.5 times faster (!!!!) than 8xFSAA. With that kind of performance delta, would you run your games with 8xFSAA?

The truth is, with 4xAA, high AF and high resolutions, this card smokes... often 100% or more faster instead of just 50% as indicated in this review.
Look at these benchmarks (and notice the biggest performance delta occurs at 1600x1200 which this review fails to consider, for some reason):

(3DMark03: up to 120% faster)
http://www.hardwareanalysis.com/content/article/1708.5/
(Tomb Raider: up to 112% faster)
http://www.beyond3d.com/previews/nvi...index.php?p=25
(Halo: up to 110% faster)
http://www.anandtech.com/video/showdoc.html?i=2023&p=13
(Far Cry: up to 105% faster)
http://www.tomshardware.com/graphic/...e_6800-42.html
(Splinter Cell: up to 90% faster)
http://www.firingsquad.com/hardware/...tra/page16.asp
(Serious Sam SE: up to 83% faster)
http://www.techreport.com/reviews/20.../index.x?pg=10
(Comanche 4: up to 62% faster)
http://www.hothardware.com/hh_files/S&V/nv40_debut(9).shtml

The LCD I used for testing doesn't do 1600x1200, hence its ommission. I show the 1024x768x4AA results too, so readers get to see 4AA performance, along with 8AA. You saying I shouldn't tell our readers what the maximum AA mode on their brand new £400 acclerator is like?

ATI manage to do high performance, high quality AA, NVIDIA still can't get it right, using supersampling for their 8AA mode which is unchanged from GeForce FX.

"The fact of the matter is that the 6800U does extremely poorly with on one particular setting: 8xFSAA. That setting is an anomoly not the rule."

Rubbish. On the fastest graphics accelerator in the world, it's definitely the rule. That their new chip can't do their highest AA mode at speed isn't really forgivable. Here's hoping they change it in future.

I show that it's fast at 4X, I show that it's slow at 8X. That's what all reviews of the card should do.

Rys

I have a question about this review. Your's is one of many reviews i've read on the 6800 and what i'd like to know is this. Why is it that your's is the only review that I've seen that shows the R360 actually beat the 6800 in any benchmark? All the other reviews I've show the 6800 beating the R360 by either a very little to completely dominating the R360 as well as the 5950. What gives? I'm not a fanboy to either side in this, but I like the absolute cold hard trueth when it comes to reviews and benching. How is it that your benches end up so different than several from tech site I had thought were reputable?

The LCD I used for testing doesn't do 1600x1200, hence its ommission.

You're hardware is not the reader's problem. Can't you borrow a monitor? I just find it shocking that a review of an ultra-high end card does not include 1600 x 1200 tests. It's like reviewing a Hummer based only on city driving.

You saying I shouldn't tell our readers what the maximum AA mode on their brand new £400 acclerator is like?

No, I clearly said you should include benchmarks showing 8xAA performance. I said that shouldn't be your ONLY setting for your higher resolution tests considering you know the performance on the setting is impractical. What would you have done if the 8xFSAA setting didn't work at all?

ATI manage to do high performance, high quality AA, NVIDIA still can't get it right

It is correct of you to inform the readers of what the 6800U does very poorly. But you should also inform the reader of what it does very well. You left that part out.

Rubbish. On the fastest graphics accelerator in the world, it's definitely the rule. That their new chip can't do their highest AA mode at speed isn't really forgivable. Here's hoping they change it in future.

That's fine if you have looked at all the facts and that is your opinion. But just because you think this card is unforgivable, doesn't mean you should withhold from the reader the tests on which the card does well. If want to impress on to the reader that this is a poor card by withholding facts, do an op-ed piece, not a review.

Look, I'm not a fanatic, but I have generally been "rooting" for ATI the last couple years. I own a 9700NP. I'm not writing this out of my non-existant love of Nvidia. I writing this because think the the data you left out of this review is more "unforgivable" than the 6800U's 8xFSAA performance is.

BTW, thanks for replying.

Peace.

Funny how all the other sites actually omit the 8XAA and 16X AF tests *completely*..

If I was buying a the supposed "top end" card I'd want to know how it performs with the settings properly maxxed out, most people with 9800pro's or XT's are used to slapping up the AA and AF settings to maximum and still getting half decent frame rates.
Those are the people nVidia are targetting with the 6800, and those are the people that are going to want to know what the card is like when you turn the eye candy up to maximum.

Rhys never pretended that this was a fully exhaustive review, that will come when the retail boards arrive.

Quote:

---

Originally Posted by rogerw99

Why is it that your's is the only review that I've seen that shows the R360 actually beat the 6800 in any benchmark? All the other reviews I've show the 6800 beating the R360 by either a very little to completely dominating the R360 as well as the 5950. What gives?

---

It only beats it with 8xAA, that is what they were trying to show. The worlds new uber card cant use 8xAA verry well at all, not even as good as last generations cards... i think thats the point that they were making, and for £400 it should do that without second thought...

but apart from that flaw it whoops the cheese out of the other cards...

On a lighter note. here's a review that's been equally well-researched and compiled.

http://img.hexus.net/TS/Battle1.jpg

<GeForce 6> I'm gonna open a can of whoop-ass on you, f00.
<9800 XT) Bring it on, honey.

http://img.hexus.net/TS/Battle2.jpg

<GeForce 6) 'ave it, 'ave it. You know you want it.
<9800 XT> Pot Noodle in NVIDIA flavour, lovely!.

Seriously though, I think Rys' take is comparing the two cards at their respective maximum settings. It's valid enough methodology.

Quote:

---

Originally Posted by Spystar

Seriously though, I think Rys' take is comparing the two cards at their respective maximum settings. It's valid enough methodology.

---

If that was the methodology, it is even more ridiculous that 1600 x 1200 wasn't tested. It wasn't Nvidia's fault Rys didn't have a monitor capable of 1600 x 1200... why punish them for Rys lack of hardware in the review? And again, I ask... what if 8xFSAA didn't work at all? Would those points on the graph just have been left blank rather than testing at 4xAA?

As far as I am concerend, 8xAA is non-functional right now. At about a third of the performance of 4xAA, I don't know how anyone could disagree. So make a note, chide Nvidia about it, and continue testing with 4xAA. I understand Rys was trying to make a point, but he did so in a subversive way that (I feel) was intended to mislead readers (such as Swafeman).

I think the point is, that at the sort of res that this card is actually going to be used at, i.e pretty much always 1600x1200, it wipes the floor with the current front-runners, the 9800XT and 5950Ultra.

12,000 3D Marks and a 57% performance increase in Halo over a 5950Ultra is very impressive. If the 9800XT can beat it at a certain setting, or keep up with it at low res, fair enough.

At the business end, high res, eye candy, and 60fps+, the card seems to do the job very nicely indeed.

I wouldn't go as far as to say the review was intended to mislead anyone, but I agree that when testing a 6800Ultra, you really should be testing it on a set-up that can do it justice, and having a monitor that can do 1600x1200 would be part of that.

One more thing bothers me, namely the results of Call of Duty benchmark as compared with the results that guys at Tomshardware got:

Call of Duty - 4x FSAA & 8x Anisotopic
http://www.remedium.republika.pl/pcp/toms6800.gif

http://img.hexus.net/v2/graphics_car...images/cod.png

Look at the framerates - at toms tests Radeon 9800XT achieves 2 to 3 times fewer FPS in all resolutions, FSAA, ANIZO, whatever.
In Hexus test there is considerably smaller difference between the scores.
This may be proving the fact that the nVidia card was not tested adequately - i.e. it did not had the chance to show it's real potential.
But such a huge difference in scores on a single game with similar settings is strange anyhow.

BTW. I am not nVidia fanboy - I'm a happy owner of R9700@PRO :)

EDIT:
BTW2: I have nothing against 1024x768 - thats the highest res that my VDU will run @ 85 Hz :( (Phillips 107T)

What's the big deal about 1600x1200? It's not the maximum 3D resolution either tested board supports, so testing at maximum settings isn't something any site does.

1280x1024 is a perfectly valid maximum testing resolution, stop taking aim at that.

Your beef is that I didn't do any 1280x1024 4AA 8AF numbers, which I'm endeavouring to do as we speak and I'll augment the graphs with them as soon as I can.

The conclusion as it currently stands will be no less valid with the extra numbers, than it is already.

I show time and time again that their 4X RGMS mode is fast. If you can't see that in the review then you're not seeing it on purpose to make a point. Look at the 1024x768 @ 4AA 8AF data points and watch the 6800U whoop the 9800XT's arse. It does it every single time and I mention it over and over and over again.

I then test the maximum possible 3D settings at a perfectly valid resolution (ooh look, it's the most common LCD resolution in use today, something the board is targetting with two DVI ports) at the highest possible IQ settings each board can manage. The 6800U gets beaten because it has a slow maximum AA mode.

You can't say 8AA is non functional. It clearly is, just it's slow. If you don't like it, don't run it, I'm not forcing you to. But a great many people, having sucked in the hype about it being the fastest graphics card on the planet, will want to crank the settings right up, and that means 8AA.

Take your beef to NVIDIA, not to my highlighting the fact it missed the boat in producing a usable high quality AA mode that gives ATI's 6AA mode a run for its money.

I won't mislead our readers by conveniently missing out the fact that their maximum AA mode is slow.

You ask if 8AA didn't work at all, what would I do. I'd lambast NVIDIA for screwing up even more than they have done. To not produce a working AA mode over 4X on their flagship product is infinitely worse than producing a mode that does work, but is just slow.

Get over 8AA being poor and use 4AA if it bothers you so much, it's a rubbishrubbishrubbishrubbish hot card at those settings, both in terms of speed and image quality.

Rys

Quote:

---

Originally Posted by Simplex

One more thing bothers me, namely the results of Call of Duty benchmark as compared with the results that guys at Tomshardware got:

Call of Duty - 4x FSAA & 8x Anisotopic
http://www.remedium.republika.pl/pcp/toms6800.gif

http://img.hexus.net/v2/graphics_car...images/cod.png

Look at the framerates - at toms tests Radeon 9800XT achieves 2 to 3 times fewer FPS in all resolutions, FSAA, ANIZO, whatever.
In Hexus test there is considerably smaller difference between the scores.
This may be proving the fact that the nVidia card was not tested adequately - i.e. it did not had the chance to show it's real potential.
But such a huge difference in scores on a single game with similar settings is strange anyhow.

BTW. I am not nVidia fanboy - I'm a happy owner of R9700@PRO :)

EDIT:
BTW2: I have nothing against 1024x768 - thats the highest res that my VDU will run @ 85 Hz :( (Phillips 107T)

---

You assume that our remit is to show the card to its full clearly visiblepotential at all times. That only comes in the theoretical analysis, which I show clearly. Check the ShaderMark or RightMark scores.

In games testing, we use our stock benchmarks. If they show certain amounts of CPU limiting, that's great. It shows GPU performance, just in a way that's not obvious. If you're CPU limited, it means the GPU is far faster than it needs to be. That's a good thing, just not as obviously as a 10000fps point on a graph.

Tom's use a different CoD demo than us, ours is more CPU limited (and arguably a more realistic measure of real world performance, in my opinion) than not. All there is to it really :)

Rys

Thx for instant reply :)
Your explanation is logical and convincing. I think both types of tests are useful - those cpu limited - because they simulate real world situation, and those on very fast CPUs because they demonstrate the maximum (or near maximum) power and potential of the card :)

I guess I didn't notice that Rys went as far as 8xAA & 16xAF and all those other tech sites did not. My apologies to you Rys. I really don't understand why those other sites would not have used those settings unless they were trying to hide the fact of the 6800's dismal performance at that setting. I was under the impression that Anandtech, Techreport & a couple of others were more reputable than that.
So, does anyone have any insight to this? Is anyone trustworthy anymore or are those other sites waiting for R420 to do a really thorough review?

Hexus really, most of the others seem biased on some way or another :(

They are reputable. I personally hold sites like Tech Report in the highest regard. We all had about 4 or 5 days with the boards and with the amount of testing that needed to be done (I personally ran over 200 benchmarks for our article), you can't really fit in all AA and AF settings at all resolutions, plus get the writing done and the images prepared.

With two weeks or so, we'd all have tested at all the major display settings. I chose, after initial testing, to highlight their shortcomings with their MS+SS 8AA mode. Other sites didn't, but their approach is no less valid than mine.

They're not hiding anything, their focus was probably just elsewhere with a limited amount of time to spend with the hardware. It's exciting to focus solely on the giant speed increases, I was tempted to do it too.

Rys

The thing is, nobody even mentioned it in passing, which gives doesn't really give the whole picture. Your review does, it shows that the card is absolutely rubbishrubbishrubbishrubbish hot on most settings, but there is a performance hole at maximum settings which ATi could possibly exploit.

The other (p)reviews give an impression that it's performance is *always* rubbishrubbishrubbishrubbish hot, and it isn't - that's more than a little misleading, at least it is for me personally.

The question is one of valid comparison.

Does NVIDIA's card at settings of 4x antialiasing and 8x anisotropic filtering produce comparable or better results than ATI's. The consensus is that it does. In that case, performance from NV's newest card is comfortably ahead of ATI's present best.

That's a statement of fact.

What if NVIDIA had implemented a 16x AA mode?. The sheer number of reads would cripple any card.

Compared on a like-for-like basis, the NV40 is the fastest gaming card on the planet, and by some way. Compared with respective cards set to maximum image quality, NV's settings are just too harsh to be used in conjunction with modern games.

This is one monster, monster card. What would you be thinking if NVIDIA had limited the AA settings to only 4x?.

Quote:

---

Does NVIDIA's card at settings of 4x antialiasing and 8x anisotropic filtering produce comparable or better results than ATI's. The consensus is that it does. In that case, performance from NV's newest card is comfortably ahead of ATI's present best.

---

but the fact of the matter is that a previous generation of cards beats this one at 8xAA that is what Rys is saying! If you payed £400 for a video card would you want it to be playable at 8xAA??

He mentioned that its a hot and uber card (not in thoes words) but just fails in one setting.

he showed this where others did not! WELL DONE!

The 8xS settings is a token gesture from NVIDIA. If you understood the demands imposed by 8x Rotated-Grid Multisampling and 4x Supersampling, you'd soon realise that it's not going to work at any great speed.

Imagine that on a 1600x1200 setting. It's just a very strange choice of AA. You'll need a theoretical NV60 to run those settings at any kind of acceptable speed.

Bottom line - it's not a truly usable option right now. You have to understand what the card's being asked to do at 8xS before slating it. It's a driver fault, if anything.

Quote:

---

Originally Posted by Applecrusher

but the fact of the matter is that a previous generation of cards beats this one at 8xAA that is what Rys is saying! If you payed £400 for a video card would you want it to be playable at 8xAA??

---

Wouldn't you also want to know how the card performed at 1600 x 1200? How can you applaud the decision to use 8xFSAA settings but ignore the choice of resolutions?

Also, I feel my comments are constantly being construed that I didn't think 8xFSAA should have been included. This is NOT the case. I also applaud Rys for including it. All I am saying is that he graphs should ALSO show what the card can do with 4xFSAA at high resolution.

Quote:

---

Originally Posted by Applecrusher

He mentioned that its a hot and uber card (not in thoes words) but just fails in one setting.

---

He said that in writing, but by just flipping through the graphs, it leaves the impression that the higher the resoultion is and the higher the AA and AF settings are, the worse the 6800U does. With one exception, it is the complete opposite. IMHO, the graphs should have included 1024 x 768 up to 1600 x 1200 at an AA setting of not more than 4, then one data point at the end that showed the 6/8 AA settings. This is would highlight the 6800's downfall at 8xFSAA as well as its dominance at any lower AA setting.

Quote:

---

Originally Posted by Maui

Wouldn't you also want to know how the card performed at 1600 x 1200? How can you applaud the decision to use 8xFSAA settings but ignore the choice of resolutions?

---

Rys already replied to that one! GO READ THE THREAD!
or i could shpw you here...

Quote:

---

I then test the maximum possible 3D settings at a perfectly valid resolution (ooh look, it's the most common LCD resolution in use today, something the board is targetting with two DVI ports) at the highest possible IQ settings each board can manage. The 6800U gets beaten because it has a slow maximum AA mode.

---

Quote:

---

He said that in writing, but by just flipping through the graphs, it leaves the impression that the higher the resoultion is and the higher the AA and AF settings are, the worse the 6800U does. With one exception, it is the complete opposite. IMHO, the graphs should have included 1024 x 768 up to 1600 x 1200 at an AA setting of not more than 4, then one data point at the end that showed the 6/8 AA settings. This is would highlight the 6800's downfall at 8xFSAA as well as its dominance at any lower AA setting.

---

but if you were just flipping through the graphs youd not realy notice what the different settings are (i know i only read the first one) and you'd still see the card doing realy porly at the end therefore the same result??