Slashdot Mirror

• middleware 10%
• content 60%
• and the rest is code, i.e. only 30%

When I want facts on graphics, I go to beyond3d.

P.S. NV3x architectures can't do everything in 8x1 mode. Has to drop to 4 ops/clock with color operations.

Here is a Beyond3d link that has some opcode info. Look around their site for a NV30 vs R300 architecture document that has lots of great stuff. If you are looking for the best s/n ratio, Beyond3d is one of the best. All meat, little fanboyism.

Beyond3D had a shader contest and some genius wrote frogger in a pixel shader. That's some great stuff.

If GPUs are to become general purpose, will the AGP bus problems have to be fixed (fast at delivering data, slow at getting it back from the GPU). Does PCI-X fix this?

ATI has an ascii "Smartshader". Good fun.. for a while..
For OpenGL only.
Check it out here

For 3D, the best site (by far) is Beyond3D.

Missed? Please don't spread the FUD Nvidia produced further.
Fact is, pretty much every review site was at this Nvidia Editors Event where Nvidia accused ATI of cheating. Yet nobody except tom's hardware published it, why do you think they didn't?
Maybe because they checked the facts first?
Ok, I'll address it point for point: 1. The UT2k3 issue: Yes, it's true that detailed textures are missing. To say it's a cheat is greatly exaggerated - it is caused by the (non-standard) default LOD bias (0.8) UT2k3 uses which causes the highest detail level never to be used. This is known since at least december 2002. You can change the LOD back to default (it's a bit unsure why EPIC chose a higher LOD anyway) in the .ini file, you'll lose something like 0.9% performance (btw nvidia loses slightly more if you change the LOD backs, some sites have measured it) and everything will look as it should. It's probably a bug in ATI's driver, but not even close to a cheat.
2. Aquamark: It's true the image rendered isn't entirely correct (neither is Nvidia's, btw), some areas are too dark. However, Tom's states "It seems that ATi's driver doesn't continue to fade out the textures until the blending ends, instead simply cutting out certain textures when a certain degree of fade-out is reached. This obviously saves memory bandwidth, leading to a higher framerate" - which is complete BS.
3. Halo: People tried to look for some discrepancies in the rendering output of ATI's graphic cars, but didn't found anything. Without ANY evidence, this makes this accusation plain FUD.
It's a really sad (maybe desperate) move from nvidia - they more or less openly continue to uhm, "optimize" for certain benchmarks (despite their optimization guidelines which aren't worth the paper they are printed on) and accuse competitors of cheating without any proof (in fact, I'm sure they KNEW those were no cheats, but simple bugs), though they didn't actually accuse ATI of cheating directly (simply saying there are "issues" with ATI's drivers), and let toms hardware do it instead.
You might also want to read ATI's response on these accusations (and some discussion about it) at beyond3d.

IF YOU DIDN'T KNOW ALREADY E3 DEMO HAS BEEN LEAKED K THNX BAI

OH please stop by here if you wanna thank me for the info

http://www.beyond3d.com/forum/viewtopic.php?p=1767 87

NO MODS THERE IS NO WARES LINKS.. ( plz feel free to quadraple check ).

You can't talk about performance of games that haven't been released yet, like HL2

Fine. We'll talk about Tomb Raider then. Sorry, but when your top of the line card has half (or less) of the performance of the competitor's card -- with the difference being between playable framerate and unplayable frame rate -- then your card is indeed slow.

My point is that we're essentially talking about a handful of games here, and these are not games that are particularly well optimized.

A handful of games, yes. But I'm happy to know that suddenly you're an expert on whether or not the games are well optimized. I'm sure that you know far more about this than the people who are actually writing the games, or people with such paltry graphics knowledge like John Carmack.

And while it may be a handful of games, the same could be said for the games that utilized 3D back when it was GeForce2 vs Voodoo vs ATI Rage3D. If you dismissed the benchmarks then as only a "handful of games" and bought a Voodoo or Rage3D then you were going to be cursing yourself when you had to go buy a new card a year later since yours was now utterly useless. And yes, I know plenty of people who got to be in that wonderful position.

If 3D game XYZ was targeted for the Xbox, it would rock and roll and be plenty fast.

What BS. Console freaks who say this (and yes, I own a console as well) don't have a damn clue about what resolutions they're playing. Standard NTSC video is all of 640x240 resolution. You don't have to render anything but every other line, and you better believe that the consoles don't. Xbox has a maximum resolution of 1280x720 or 1920x525 (the former is far more intensive graphically, and is why virtually no Xbox games support 720p widescreen -- if they support 720p it's 4:3 only, giving a paltry resolution of 960x720). Yawn. And on top of that the most advanced platform of the three out there -- Xbox -- only supports a subset of DX8 level graphics. We're talking about DX9 level now and the quality differences involved are not insignificant.

The sad part is that high-end PC gaming is moribund, except for a small class of fanboy games. Everyone else has either given up or bought a console.

You know, someone has said this every year for the past 15 years and they've been wrong each and every year.

Mass market PC gaming is growing like crazy, though.

And mass market PC gaming is merely what high end PC gaming was a year prior.

I said that it's a damn nice card, whose only offense isn't being as fast as a newer card. But that's like arguing that one fighter jet is 200Mph faster than another.

The only offense is that the FX is a fighter jet that can't even get off the ground, but costs more to buy and use than the one that actually flies. That's a pretty damned big difference.

Who cares, especially if that jet sits in the hangar 90% of the time anyway?

Well if you don't care then why the hell are you even buying a new graphics card? The FX line provides virtually nothing over the Ti line of cards. If your jet is only going to sit in the hanger, why even buy it now? Buy it when you'll actually use it -- you'll get it for less to boot.

Which is why, I'm sure, that every single real DX9 benchmark has shown nVidia falling far, far behind ATI.

The quotes from that second link are particularly damning -- and they're from a variety of companies, including id Software, not just Valve.

I've never owned an ATI card. My last 5 or 6 cards in all my computers (and my wife's) have been nVidia. My next card is almost certainly going to be ATI though because they're currently the performance leaders. I have some reservations about drivers still -- not with performance or stability but with long term support since ATI has still failed to deliver a unified driver architecture -- but I'm unwilling to sacrifice that much performance while still paying a higher price.

Frankly, at this point anyone who is still wondering about the validity of the benchmarks is deserving of the title "nVidia fanboy".

Toms Hardware
FiringSquad
Tech Report
Gamers Depot
Beyond 3d

as to why synthetic benchmarks are useful please see the following:Beyond3D. This website is probably the best site for info on 3d hardware.

later,

Prolly nobody is going to read this but never the less I will post it.
(The internet will converve it forever ^^)

DX8 was designed around the NV20 (NVIDIA) specs, DX9 was designed around the R300 (ATI) specs.

clicketiclick for more detailed info

If anyone is interested in publicly communicating with Tero Sarkkinen, he has been known to monitor this forum. (He's also posted there...)

Beyond3d is reporting on the ATI part of this issue.

ATI's official statement:
The 1.9% performance gain comes from optimization of the two DX9 shaders (water and sky) in Game Test 4 . We render the scene exactly as intended by Futuremark, in full-precision floating point. Our shaders are mathematically and functionally identical to Futuremark's and there are no visual artifacts; we simply shuffle instructions to take advantage of our architecture. These are exactly the sort of optimizations that work in games to improve frame rates without reducing image quality and as such, are a realistic approach to a benchmark intended to measure in-game performance. However, we recognize that these can be used by some people to call into question the legitimacy of benchmark results, and so we are removing them from our driver as soon as is physically possible. We expect them to be gone by the next release of CATALYST.

Cg was developed, designed, and created by nvidia. While one of their claims is that it can be made to run on any card and is multiplatform, don't let that fool you. Cg is, at its worst, a thinly veiled attempt to convince developers to produce optimal code for nvidia cards at the expense of broad hardware support. ATI has already said that they will not be supporting Cg (in order for it to work best on ATI cards, someone needs to create profiles for it) and will instead be supporting HLSL. I doubt S3/Via or SIS have the resources to commit to 2 different projects, so I bet they're going to go with HLSL.

If you don't understand why nvidia might be looking for code that works best only on its cards (it's almost a "duh" question), look at it a different way. Look at the GFFX. In almost every instance, it's a failure. Sure, it can stick to 32-bit precision, but it runs really, really slow when you do (just look at the 3dmark03 scores recently released and john carmack's .plan comments). When it runs at 16-bit precision, it's still damn slow, almost always losing out to the Radeon 9700/9800s, but it's a little more competitive (DX9's minimum spec appears to require 24bit precision, but rumor says the jury's still out on that). It's in nvidia's best interest to make the FX appear to be fast (which it isn't), and so they're relegated to make Cg code that optimizes for nvidia cards their best interest.

Sorry I don't have links, but the beyond3d.com forums have a lot of information on this subject.

Carmack has said a few things in regard to this:

1. He can increase performance by 50% by optimizing for a fixed plattform (Xbox)
2. Doom3 was designed with NV20 (GeForce 3) level features as the baseline--the NV20 rendering path has all the eye-candy of the ARB2 or NV30 rendering path (see this interview with JC over at Beyond3D).
3. A NV20 (GeForce 3) will be able to run Doom3 at 30FPS with max detail at 1024x768. It's a single-player game as well, so 60FPS+ isn't that big of a concern. With R300/350/NV30/35, you'll be able to get much higher framerates and use of anisotropic filtering and antialiasing. BUT, the game will be playable on a GF3.
4. The only thing that would have to be cut for an Xbox version would be some of the higher-res textures because of the Xbox's memory constraints--however, those textures would also be completely unnoticeable because of the much lower resolution on a TV compared to a PC (unless you're playing on an HDTV, of course).

So yeah. Xbox IS a PC, and not an altogether bad one at that. Plus the NV2X core in the Xbox is faster than a GF3, or so most people believe (but not as fast as a GF4).

there is also a thorough one at Beyond 3D

topic at beyond3d's forums

It seems the GeforceFx previews made some mistakes when comparing the relative merits of the NV30 and 9700.

For a good read check out this comparison of the new chips. Even this gets a few things wrong

The vertex shader 1024 instructions limit of the 9700 is mostly misreported.

The Geforce vertex shader is more powerful, supporting dynamic branching, but the 9700 can do a static branching per primitive with instruction lengths of 65,026. I'm sure such shaders will be equally unwieldy for gaming on both cards.

It seems some of the reviewers only looked only at what NVidia told them and what DX9 exposes and not what the 9700's actual silicon supports.

The reviewers barely understand the full capabilities of the 9700 months after its release. I'm sure the FX will provide similar surprises (good and bad) when they actually get their hands on one.

It seems the GeforceFx previews made some mistakes when comparing the relative merits of the NV30 and 9700.

For a good read check out this comparison of the new chips. Even this gets a few things wrong

The vertex shader 1024 instructions limit of the 9700 is mostly misreported.

The Geforce vertex shader is more powerful, supporting dynamic branching, but the 9700 can do a static branching per primitive with instruction lengths of 65,026. I'm sure such shaders will be equally unwieldy for gaming on both cards.

It seems some of the reviewers only looked only at what NVidia told them and what DX9 exposes and not what the 9700's actual silicon supports.

The reviewers barely understand the full capabilities of the 9700 months after its release. I'm sure the FX will provide similar surprises (good and bad) when they actually get their hands on one.

Apparently, nvnews don't accept poor GeForce results. Author "typedef enum" wrote an article showing how Matrox Parhelia-512 beat nVidia GeForce Ti 4600 in benchmark after benchmak. The review was pulled for rewriting and some tabulated comparisons were never allowed as mentioned by the author at Matroxusers forum. The results can now be seen at Beyond3D.

Can I recommend the review at Beyond3D? The reviews there generally dig a bit deeper into the technology than most of the run-of-the-mill sites.

Simon

SharkyExtreme is loading fine for me (Midday Shanghai, China time) but here is a link to a earlier story by nvmax.com (including a couple of screenshots).

NVIDIA NV30 Sneak Preview

Some Beyond3d forum discussion as well as screenshots and more info on the NV30.

NV30 Screenshots

One more link.... to nV News with further NV30 details

nV News

- HeXa

The rumored kyro 3 offering soon may to have to compete with nvidia's new tbr patent dated the 5th of febuary. Let's hope the best video card wins and nvidia doesn't just crush them with dirty tactics (warning pdf file linkage).