[Dave Baumann]

Well, I know a number of you have been looking forward to it, and here's Beyond3D’s relatively exhaustive Reference GeForce FX 5800 Ultra Preview. We go over the gaming performance, Anisotropic and FSAA performance and quality settings, look at Shader Speeds and throw a number of other tests at it to see what this has in store for us. We compare the board against NVIDIA’s GeForce4 Ti 4600 to see how far they’ve come along in a year since their prior generation, and also clock the card to 400/400 speeds to give a rough feel of the performance of the non-Ultra boards.

"Comparing the different boards with 4X FSAA we can see that the GeForce4 Ti, which has no colour compression abilities, just Z compression, becomes bandwidth limited very quickly. GeForce FX, on the other hand, has plenty of performance left in it, due to its higher bandwidth and compression routines. We can see that at best, the 5800 Ultra is in excess of twice as fast as the GeForce4 with 4X FSAA enabled, and even running at 400/400 it is still twice as fast."

Read the full preview here.

Note, that while I say this is a fairly exhaustive investigation, we do still have more to run for an upcoming article

[epicstruggle]

second post.

[Kristof]

You are one minute late... granted I was 2 minutes late with my post in the other thread...

[Brent]

third post!

and i'm still up

gonna read the preview now :P

[Luke Philpot]

wooo! Reading now!

[Brent]

oops nm

[Anonymous]

Nice Article and Very informative.. but..

YAAAAAAWWWWWNNNN...

Bring on the R350 previews...

[Anonymous]

he board arrived in tin, resembling that of an old cinema film reel casing

When I saw a photo I tohught it was a garbage container and the photo was a practical joke 8)

[epicstruggle]

Ok read it a quick typo at the bottom of page 4:
"Note, the 42.68 drivers are note available"
should be:
"Note, the 42.68 drivers are not available"
Didnt see anything else, but im freaking tired.

I am curious as to whether or not there were any restrictions to how you could (p)review this card. Or if you can even tell us that there were restrictions? I for one would have loved to have seen it against the competition.

Still the article was great, whished the card was too.

good job to all those involved,
later

[Luke Philpot]

epicstruggle: afaik NVIDIA aren't letting [p]reviewers say 3DMark03 results.

All in all, awesome preview! Thanks!

[Anonymous]

epicstruggle: afaik NVIDIA aren't letting [p]reviewers say 3DMark03 results

Which is Complete Bullshit.

So now we have the precident set that any card company can deny reviewers boards unless they ONLY bench what Nvidia wants them to. How wonderful. They wont even allow it to be mentioned. Whcih is just underhaned.

This Site should have told them to shove their POS up their @$$. Then made a public statement as to why. This is not good for the industry. It only opens a can of worms for the future. Whats next.. WE suck at this game.. so no one is allowed to bench with it or else..

[Jima13]

"Now, finally, we get to see what NV30, formally known as GeForce FX"

any relation to the artist formally know as 'Prince'?

[tEd]

about the 6xS AA mode which i would say is rather 4xMS with 1.5 horizontal SS.


anyway great preview!

[Lezmaka]

Originally Posted by Jima13

"Now, finally, we get to see what NV30, formally known as GeForce FX"

any relation to the artist formally know as 'Prince'?

No, that's a different word. Formally as in the the official name, as opposed to formerly, as in an old name.

[Dave H]

Well I wasn't really expecting to learn anything new with the traditional B3D test suite, but there were certainly enough new pieces of information (by far the clearest info yet on what exactly each AF mode is doing!) and curious anomalies (the persistent 1600*1200*4MSAA plunge...not to mention RTCW ) to keep it plenty worth staying up for!

And how could I forget this one:

During testing the fan did spin down quite frequently, presumably because the board was overheating, when running 3D applications, also reducing its 3D performance.

I assume these never affected the results you got, but...perhaps for kicks you could run a little "stress test" for the next article, and let the clock throttling chips fall where they may?

Of course I think I speak for many when I say I viewed this preview more as an appetizer for the DX9 (and DX9 equivalent) shader-fest that will hopefully be coming soon. pocketmoon_ and others are doing a great job putting up some intriguing data points, but I'm counting on you to synthesize all the numbers and make some sense of that mess!

[Anonymous]

Originally Posted by Anonymous

So now we have the precident set that any card company can deny reviewers boards unless they ONLY bench what Nvidia wants them to. How wonderful. They wont even allow it to be mentioned. Whcih is just underhaned.

When Anand first previewed the 9700, ATI only let him post percent improvments over the GF4 Ti4600. Same kind of thing here.

Now if B3D had gone out and got themselves GFFX, then they could post whatever they wanted. Nvidia sent them a board for testing, just like ATI let Anand use a board for testing, because in both cases, neither is available to buy.

[Jima13]

Originally Posted by Lezmaka

No, that's a different word. Formally as in the the official name, as opposed to formerly, as in an old name.

hehe, yea I realized it was a mistake, but I just copied Dave's lead

EDIT> after I reread it several times, the original comment started to make sense.......time for some shuteye

[kyleb]

a quality preview for sure Dave. i must agree with the guest that i would rather you told them to stuff it than work under restrictions. regardless, when it comes down to the path you took, i feel you handled the situation in a most dignified manor. on the other hand i feel compelled to ask; might we expect an unrestricted review of a retail product shortly?

[GreenBeret]

The performance difference between the TI4600 and the GFFX in some cases is so small that it seems that the GFFX is a considerably slower card clock for clock.

[davefb]

shame the radeon review uses a 2.53ghz machine . perhaps that needs a retest using 3hgz

-dave-

anyway jobs-a-good-un....

[LeStoffer]

Very good job, Dave! 8)

I liked this bit:

You’ll note that missing from that list is the number of colour values that are written to the frame buffer – which is the principal value that is required for calculating theoretical fill-rate, unless the criteria for defining a pixel is being redefined.

Taken straight from nVidia's 3D Glossary (before they change it!):

Pixel
Shorthand for "picture element." A pixel is the smallest element of a graphics display or the smallest element of a rendered image.

Hehe.

Just one thing that some people might get wrong from the preview:

... so this would appear to be operating more as a pipelined 4 pixel pipe card, that can be logically working on 8 pixels at any one point in time, but only 4 can be written per clock. However, ATI’s R300 architecture features 3 shading units per pipeline which are also pipelined ...

Most of us know what you getting at here, but wouldn't it be more fair to state that R300 only provide one 'true' colour operation per cycle which can be extended into two ops per cycle if you divide it into both a scalar and a vector operation?

I guess I just doesn't like the term of 3 shaders units in the context of discussing NV30 inability to render (colour) more than 4 pixels per cycle.

[Anonymous]

Great (p)review!

I'd still like to note something.

Balanced 1X AF: 89.6
Aggressive 8X AF: 93.4

So, if you compare the respective screenshots...

http://www.beyond3d.com/previews/nvi...bal_1x_col.jpg
http://www.beyond3d.com/previews/nvi...agg_8x_col.jpg

You clearly see that the balanced mode uses trilinear ( although not everywhere ) and that the agressive mode doesn't.
But the "agressive" performance numbers are, still, quite impressive IMO.

The idea behind Aggressive & Balanced seem quite easy to understand.
Aggressive puts AF *everywhere* , and never any Trilinear
Balanced puts AF at many places, but sometimes uses Trilinear ( instead or at the same time, but it sounds like it's often instead ) if it seems it would give higher quality.

Balanced seems like a good idea to me. Aggressive just sounds too aggressive...
But the problem with balanced is that while it gives quality on par with the Radeon 9700 Pro, it's a lot slower...


Uttar

[jvd]

Originally Posted by Anonymous

When Anand first previewed the 9700, ATI only let him post percent improvments over the GF4 Ti4600. Same kind of thing here.

Now if B3D had gone out and got themselves GFFX, then they could post whatever they wanted. Nvidia sent them a board for testing, just like ATI let Anand use a board for testing, because in both cases, neither is available to buy.

The cards they used for percent only tests were not the finished product and they had yet to settle on a clockspeed. Nvidia should have these boards out by now . Esp considering they will not be sold on store shelves.

[cho]

i think there is a mistake on the gf fx 5800ultra fillrate score in page 6.

[Colourless]

Something to note about the 8x agressive shot is that the first miplevel is bending backwards. From what I understand really shouldn't be doing that...