Dissappointed : No NEW features introduced with nV30

[sabeehali]

I was hoping to see atleast a few new features introduced with the nV30 but it seems like it does nothing beyond R300 (feature wise). Parhelia had Glyph antialiasing, fragment antialiasing etc R300 had Video shader etc.

So my question is are there a few feature thats haven't been disclosed yet?

 

[megadrive0088]

disappointment - 128-bit bus (4x 32-bit memory controllers) - expected
disappointment - single TMU per pipeline - unexpected
disappointment - same method of FSAA as Nv25 - unexpected


500 Mhz core clock is more than I expected though.


regardless, it would be stupid to fork over the money for a GeForce FX
when you can almost smell the refresh coming in 2003 with much better performance (256-bit bus, 2nd TMU per pipe, additional VS engines, etc)

 

[Typedef Enum]

I just saw a blurb about Dual-Link DVI...This would be a very welcome feature for LCD users. At present, only the Gainward GF4's provide dual-DVI, and ATI's FireGL does as well...

 

[Typedef Enum]

I think the fact that all of these other guys...Matrox/ATI have come out before nVidia takes a bit of the wind out of nVidia's sail.

Clearly, there's only a few features that you could really "wow" somebody with @ this point...New Antialiasing technique, new filtering, totally new way to game (Surround), etc.

What is really needed are more in-depth white papers explaining specifics on their AA implementation(s). As of right now, we're hearing that it's a similar one as found in previous products...but I don't think that has been confirmed yet.

Bascially, I think there's far too little information known about this chip/product @ this point in time. I was very surprised @ just how very little information was found in the pre-release PDF's that nVidia sent out.

 

[Testiculus Giganticus]

NV30 has videoshader as well.

 

[Joe DeFuria]

To be fair to NV30, it does offer some notable advancements over R-300...just not really for gamers AFAIAC.

The advanced shaders with control flow should be a benefit for "cinematic rendering". Should be a nice toy for programmers to fool around with and try to exploit.

But for gamers, the only thing possible that NV30 appears might offer over the R-300, is performance. And obviously the jury's out on that one...

 

[Brent]

NV30 has videoshader as well.

its marketing name is: VMR (Video Mixing Renderer)

 

[Testiculus Giganticus]

Correct.

 

[Fred]

But for gamers, the only thing possible that NV30 appears might offer over the R-300, is performance. And obviously the jury's out on that one...

What about better image quality? The jury's still out on that one too. Since, we don't know anything yet.

 

[Joe DeFuria]

What about better image quality? The jury's still out on that one too.

What leads you to believe any of the technology in NV30 would lead to better image quality relative to the R300 for games?

 

[BoddoZerg]

It seems that NV30 will have better image quality than R300 - if only because of 128 bits internal precision vs 96 - but whether this difference is ever visible remains to be seen.

On the one hand, nVidia sold tons of TNT2s based on having 32 bit support that the VooDoo3 lacked, even though the TNT2 couldnt run 32 bit at anything resembling a playable framerate. Then again, Radeon's 40 bits internal precision didn't win it much acclaim. =/

 

[Sage]

remember, the NV30 can make two 64-but texture reads in the time it tkaes to make one 128-but texture read, whilst R300 takes just al long to make a 64-bit. and as for the 128-bit memory bus, i think nVidia wont have a problem due to efficiency. its not exactly easy to use up most of a 16Gb or 20Gb bus and i'll bet the NV30 is more efficiency optimized than the R300. remember how horribly the real-life performance of the Parhelia's bus was?

 

[Dave Baumann]

remember, the NV30 can make two 64-but texture reads in the time it tkaes to make one 128-but texture read, whilst R300 takes just al long to make a 64-bit.

No, it can't.

The FP16/FP32 speed differential only applies to exectuion of shader instructions, not reading in or pixels out. Whether the input texture map is 32 int, 64 FP, 128 FP it still only samples from one map at a time.

 

[BoddoZerg]

i'll bet the NV30 is more efficiency optimized than the R300. remember how horribly the real-life performance of the Parhelia's bus was?

So R300 has the efficiency of Parhelia now?!?

 

[RussSchultz]

i'll bet the NV30 is more efficiency optimized than the R300. remember how horribly the real-life performance of the Parhelia's bus was?

So R300 has the efficiency of Parhelia now?!?

Yup, that's exactly what he said. 2x

 

[andypski]

its not exactly easy to use up most of a 16Gb or 20Gb bus

Where did you get that idea?

 

[Nagorak]

I was hoping to see atleast a few new features introduced with the nV30 but it seems like it does nothing beyond R300 (feature wise). Parhelia had Glyph antialiasing, fragment antialiasing etc R300 had Video shader etc.

So my question is are there a few feature thats haven't been disclosed yet?

They already announced their new features months ago with CineFX (although not as if it will do anything for games). This is what happens when you're running severely behind schedule. I guess a big part of it is the 0.13 move, but I still have my doubts as to whether it's "all TSMC". I'm sure that's what they told Anand, but that doesn't mean it's the complete truth.

 

[RussSchultz]

its not exactly easy to use up most of a 16Gb or 20Gb bus

Where did you get that idea?

I heard it on the internet. It must be true.

 

[PiNkY]

If both memory systems have similiar cycle latency nvida's solution with 4*32bit access @ 500mhz could well offer more usable bandwidth than ati's 4*64bit @ 310mhz, especially for shading. The point though is in my opinon that r300 in its current form will not be the direct competitor to nv30, we will probably at least see a speed bumb within 2 months of nv30's actual shelf availability, perhaps even an updated core. Also, ati should be well ahead in dx9 driver developement by then.

 

[Fred]

Well, for one we don't know if they tweaked the internal math/precision of the AA or anisotropic (its very possible with a new architecture since buffer sizes/restraints can change).

We don't know anything about the grid selection algorithms (Anand mentioned it was adaptive.. What this entails precisely is anyone's guess. Foley and Van Dam have a section about adaptive filters).

We haven't seen the gamma tweaks in action.

Nor do we know anything about signal integrity.

Like I said. We don't know anything until a card is in hand.