More info about RSX from NVIDIA

[nAo]

mpeg4 removed
a SPE faster than 8 vertex shader removed too

32 pixelshader+4 redundancy pixelshader will be fine

I love you version, cause you keep mixing dreams with reality, go on dude, don't stop please

 

[Fafalada]

EE can decode 1080p in software?

Yep, I think we even have a poster around here that wrote the software for it

mpeg4 removed
a SPE faster than 8 vertex shader removed too
32 pixelshader+4 redundancy pixelshader will be fine

Fog ALU removed and replaced with FP32 normalize?

 

[Titanio]

Titanio
Well, novelty is a good feature right, and undeniably, G70 is a fast chip, actually it's a top performer that is beating everything to a pulp at this point, BUT, as a forward-looking, novelty-seeking product, it undermines all of rumours about it and the RSX.

Huh? What rumours were you hearing, exactly? This one came out pretty much in line with the rumours, but with a few surprises here and there.

It doens't seem to incorporate any new ideas that ATI have already included in Xenon which is why i'm asking if by chance the chips are pretty much identical (hightly doubtful).

They're not identical, their architectures are completely different. I'm not sure where you got the idea that they'd be of similar designs?

 

[-tkf-]

Dunno if this is usefull:

http://graphics.tomshardware.com/graphic/20050622/nvidia_7800_gtx-01.html

Hopefully

 

[Shifty Geezer]

Huh? What rumours were you hearing, exactly? This one came out pretty much in line with the rumours, but with a few surprises here and there.

A 'helper' chip for one thing, providing shadowing and illumination functions. In RSX anyway.

 

[Titanio]

Huh? What rumours were you hearing, exactly? This one came out pretty much in line with the rumours, but with a few surprises here and there.

A 'helper' chip for one thing, providing shadowing and illumination functions. In RSX anyway.

Oh good lord, anyone who took that seriously...

 

[j^aws]

Jaws, those instruction counts are not max's across the entire chip as you suggest. Clearly they relate to the maximum number of execution units active per clock, but which units those are hasn't been clearly defined. The 136 likely includes things like norm, fog, etc. Surely you aren't suggesting that ALU's must sit idle for fog or vertex fetch because of a lack of instruction slots.

No, I wasn't suggesting that but the contrary. What I'm saying is that when comparing only the '32bit' per component execution units, then 'peak' Xenos ~ 480 Flops/cycle and G70 ~ 464 Flops/cycle are comparable. However, the G70 has 72 instructions still 'free' and available per cycle to do whatever useful work on other execution units etc., but Xenos has already used up all it's instructions for that cycle to attain that 'peak' number.

If your're including texture ops for Xenos in addition to it's 96 instructions/cycle, then you might aswell include what G70 is doing with it's other 72 instructions per cycle, even if they're not texture ops...otherwise ~ 30-40 Biilion instructions per second are being ignored!

 

[psurge]

Seems like you could cut 8 of the ROPs (and if they are really capable of fp32 blending, then hopefully they'll also be extended to handle 4 Z/stencil samples per clock). Also, couldn't one drop support for most of the weird AA modes (like quincunx, 8xS, etc..) and offer just 0/2/4x?

 

[Farid]

Also, couldn't one drop support for most of the weird AA modes (like quincunx, 8xS, etc..) and offer just 0/2/4x?

8xSSAA is a "free" feature transistors wise.

 

[alexsok]

They're not identical, their architectures are completely different. I'm not sure where you got the idea that they'd be of similar designs?

Now where did you hear that they are COMPLETELY different? The only company that went on-record and dispeled all the rumours are ATI who explicitly stated that the architecture of Xenos has been built from the ground up (which pretty much speaks for itself - what's with the unified ps & vs shaders and other technological innovations, which pc accelerators will catch up on in 1/1.5), while NVIDIA have confirmed that they have no intentions in the near future to rework their architecture and unify the dissociated units and the RSX will be based on a more traditional architecture (ergo G70 assumption). Besides, the next-generations specs are longhorn oriented and all the new features would have to adhere to that reworked framework, seeing as RSX isn't compatible to it, i surmise it will offer pretty much the same features as G70, perhaps with subtle improvements. Now i didn't say that X360 is more powerful than the PS3 so please no flame wars here, but, to tap into all the next-gen tech, NV will have to adapt a new approach and start from scratch, and after seeing G70 in action and reading NV's comments on RSX, i HIGHLY doubt they will actually differ by all that much (higher clock speeds, some cash optimization, perhaps more ps & vs the usual polishing).

 

[mckmas8808]

Here is something from the 1Up website.

Ujesh Desai also mentioned that while this very technology is being developed for the PS3 as well, since there's plenty of time to improve the performance gains on what they've already accomplished, that the PS3's performance will surely outpace what was demonstrated at this event. In fact, the key to Nvidia's SLI strategy is to show, in future generations, what will be capable with single chipsets, using dual graphics cards running in parallel now. Meaning, the next gen of graphics technology can likely be gauged using two of the current cutting edge tech. Based on what we saw today, it's a promising glimpse into the future of each console's visual potential, perhaps finally fulfilling the promise of Hollywood-quality CG in the context of real-time, interactive gameplay.

So the RSX will be comparable to 2 G70s running in SLI? Man thats huge.

Link http://www.1up.com/do/newsStory?cId=3141621

Also Derek from AnandTech wrote:

Overall, we consider this a successful launch. Aside from the performance of the 7800 GTX, we can infer that the PS3's RSX will be even more powerful than the G70. As RSX will be a 90nm part and will still have some time to develop further, the design will likely be even easier to program, faster, and full of more new features.

I know this is expected but its just makes me feel better reading it from them.

 

[PC-Engine]

Here is something from the 1Up website.

Ujesh Desai also mentioned that while this very technology is being developed for the PS3 as well, since there's plenty of time to improve the performance gains on what they've already accomplished, that the PS3's performance will surely outpace what was demonstrated at this event. In fact, the key to Nvidia's SLI strategy is to show, in future generations, what will be capable with single chipsets, using dual graphics cards running in parallel now. Meaning, the next gen of graphics technology can likely be gauged using two of the current cutting edge tech. Based on what we saw today, it's a promising glimpse into the future of each console's visual potential, perhaps finally fulfilling the promise of Hollywood-quality CG in the context of real-time, interactive gameplay.

So the RSX will be comparable to 2 G70s running in SLI? Man thats huge.

CGI demo...

 

[Titanio]

They're not identical, their architectures are completely different. I'm not sure where you got the idea that they'd be of similar designs?

Now where did you hear that they are COMPLETELY different?

I think it's pretty obvious just looking at them, no?

while NVIDIA have confirmed that they have no intentions in the near future to rework their architecture and unify the dissociated units and the RSX will be based on a more traditional architecture (ergo G70 assumption). Besides, the next-generations specs are longhorn oriented and all the new features would have to adhere to that reworked framework, seeing as RSX isn't compatible to it, i surmise it will offer pretty much the same features as G70, perhaps with subtle improvements.

RSX is a modified version of the G70. That much is agreed on by pretty much everyone. Were you expecting differently before or something? I may be missing your point.

to tap into all the next-gen tech, NV will have to adapt a new approach and start from scratch, and after seeing G70 in action and reading NV's comments on RSX, i HIGHLY doubt they will actually differ by all that much (higher clock speeds, some cash optimization, perhaps more ps & vs the usual polishing).

I'm very confused by what point exactly you're trying to make.

I can only guess at what you're trying to get at here, but in general we're seeing two different approaches being taken. MS/ATI's approach offers no inherent guarantees about performance just because its different and newer. They're arranging their power in one way, nVidia in another.

Again though, I'm not sure what point you're trying to make, so it's perhaps pointless to continue until you clarify.

 

[Rockster]

If your're including texture ops for Xenos in addition to it's 96 instructions/cycle, then you might aswell include what G70 is doing with it's other 72 instructions per cycle, even if they're not texture ops...otherwise ~ 30-40 Biilion instructions per second are being ignored!

Jaws, I got you now. If you know what those other 72 instructions are, I would be more than happy to. Sadly Nvidia isn't giving us that info. In any event, I was counting flops, for which thankfully Nvidia has helped us out and given us totals we can derive. So, the best available info is the basis for my comparison. Texturing is noted, because it is "known" to affect the advertised flop count in one architecture and not the other. If any of those 72 instructions are "known" to have a similar effect, then they should absolutely be considered. There are obviously lots of factors influencing performance, but as I said before, this was trying to focus on potential shader execution speed.

 

[KimB]

But designing a GPU for "peak" is clearly not working.

In all these reviews, the best case we're seeing is a 50% speed-up over 6800 Ultra in shader-limited cases. That 50% speed-up can be entirely explained by increased pipelines and clock.

One of the shader benches here at B3D showed 119% speedup over the 6800 Ultra. Two of the Shadermark shaders showed less than 50% speedup (which can likely be attributed to either an immature compiler, or the limitation not being in the shader). Most were much higher.

 

[aaronspink]

Xenos can only issue 96 instructions/cycle ~ 48 vec4 + 48 scalar. It doesn't have any further instructions available per cycle for any other execution units? :?

Xenos ~ 480 Flops/cycle + NOTHING ELSE?

G70 can issue 136 instructions per cycle.

e.g. 64 instructions on 56 vec4+ 8 scalar ~ 464 FLOPS/cycle AND it still has 72 UNUSED instructions per cylce.

G70 ~ 464 Flops/cycle + 72 instructions/cycle on further operations.

Actually its probably broken down closer to this:

(PS)24*2 ALUs each of which can issue 2 instructions in co-issue = 96 inst +
(PS)24 Misc Ops (aka 16b NRM) = 24 inst
(VS)8 VALUs each of which can issue 1 instructions = 8 inst +
(VS)8 SALUs = 8
Total = 136

Aaron Spink
speaking for myself inc.

 

[aaronspink]

Both the EE and certainly the CELL are FP monsters, which is exactly what decoding needs.

Yep that floating point does wonders considering that all video is integer data.

Aaron Spink
speaking for myself inc.

 

[Titanio]

Both the EE and certainly the CELL are FP monsters, which is exactly what decoding needs.

Yep that floating point does wonders considering that all video is integer data.

Aaron Spink
speaking for myself inc.

In Cell's case at least it shouldn't matter, since it's also should be an int op monster IIRC.

 

[mckmas8808]

Another point risen from Anandtech

Even though features haven't been added to the vertex and pixel shaders directly, the increased power will allow game developers more freedom to generate more incredible and amazing experiences. Though not seen in any game out now or coming out in the near term, the 7800 GTX does offer the ability to render nearly "Sprits Within" quality graphics in real-time. Games that live up to this example (such as Unreal Tournament 2007) still have quite a ways to go before they make it into our hands and onto our hardware, but it is nice to know the 7800 GTX has the power to run these applications when they do come along.

WOW!!! Do you guys believe the conclusion that they have arrive to?

 

[ralexand]

Here is something from the 1Up website.

Ujesh Desai also mentioned that while this very technology is being developed for the PS3 as well, since there's plenty of time to improve the performance gains on what they've already accomplished, that the PS3's performance will surely outpace what was demonstrated at this event. In fact, the key to Nvidia's SLI strategy is to show, in future generations, what will be capable with single chipsets, using dual graphics cards running in parallel now. Meaning, the next gen of graphics technology can likely be gauged using two of the current cutting edge tech. Based on what we saw today, it's a promising glimpse into the future of each console's visual potential, perhaps finally fulfilling the promise of Hollywood-quality CG in the context of real-time, interactive gameplay.

So the RSX will be comparable to 2 G70s running in SLI? Man thats huge.

Link http://www.1up.com/do/newsStory?cId=3141621

Also Derek from AnandTech wrote:

Overall, we consider this a successful launch. Aside from the performance of the 7800 GTX, we can infer that the PS3's RSX will be even more powerful than the G70. As RSX will be a 90nm part and will still have some time to develop further, the design will likely be even easier to program, faster, and full of more new features.

I know this is expected but its just makes me feel better reading it from them.

Maybe they got rid of all the needless trannies and added a extra g70 core.