NVIDIA's Goldman Sachs Webcast, Summary/Transcript (info G71, G80, RSX, etc.)

[Richthofen]

Sorry, but who gives a rat's ass about ATi's dominant market share in segments with piddling profits?

my words mate...
just look at their terrible margin numbers. Its almost a 10% difference between them and Nvidia.

 

[Jawed]

Indeed, it is amazing to me how some people can think of success as anything other than making $$$...

Well, as the ATI guys said recently, NVidia has been promising great great things for the laptop sector for years.

Handheld is a complete wipeout for NVidia. IGP isn't much better. No presence in consumer TVs.

Yeah, that's a recipe for success.

Jawed

 

[Jawed]

You're really reaching there. I can't believe you're actually trying to cast Nvidia's current financial and market position in a negative light after their performance over the last few quarters. Sorry, but who gives a rat's ass about ATi's dominant market share in segments with piddling profits?

One word: Overbought.

Jawed

 

[trinibwoy]

But the patent has nothing to do with unification and everything to do with programmable texture units, and accessing portions of their functionality.

Unification isn't material to that patent. Unification as a concept has been around for years...

Jawed

So because the patent isn't named "UNIFIED ARCHITECTURE" unification is immaterial? On the first page of the patent is a picture of a unified shader with no texture unit in sight - that's pretty material in my book. You've apparently taken a very negative and narrowminded view on anything that Nvidia has done or can possibly do in the future - guess I should know when I'm talking to a wall.

Edit: Actually, having just re-read the patent it seems the idea of scheduling vertex and pixel threads on each of the processors is fundamental to the description and a lot of text is devoted to it so I'm not exactly sure how you can claim it's immaterial. Oh well......

 

[trinibwoy]

Well, as the ATI guys said recently, NVidia has been promising great great things for the laptop sector for years.

Handheld is a complete wipeout for NVidia. IGP isn't much better. No presence in consumer TVs.

Yeah, that's a recipe for success.

Jawed

Jawed, ever hear about the bottom-line? That's the measure of a company's success, not whether it has chips in your TV. It's really weird seeing you argue that a company making truckloads of money isn't successful, it's downright surreal.

 

[Geo]

Jawed, ever hear about the bottom-line? That's the measure of a company's success, not whether it has chips in your TV. It's really weird seeing you argue that a company making truckloads of money isn't successful, it's downright surreal.

Tho to be fair, financial and market share tend to be trailing indicators.

I do think _xxx_ comment above about what has R&D been doing is a valid one --we know (or at least think we do!) what they *haven't* been doing, and that's spending great gobs of whiteboard time on G71!

 

[Jawed]

Which claims in this patent refer to a unified architecture?

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=/netahtml/search-adv.htm&r=1&f=G&l=50&d=PTXT&p=1&p=1&S1=6,987,517&OS=6,987,517&RS=6,987,517

None. Therefore unification is immaterial.

A multi-threaded processing unit isn't unified, you know.

Jawed

 

[Jawed]

Jawed, ever hear about the bottom-line? That's the measure of a company's success, not whether it has chips in your TV. It's really weird seeing you argue that a company making truckloads of money isn't successful, it's downright surreal.

Making loadsa money now doesn't equal making loadsa money in the future.

Success isn't just a backward-looking concept.

I suppose I could draw a parallel with 3DFX or something, but why bother - I don't think NVidia's going to disappear or anything, but there are great chunks of the market (e.g. laptops, which have overtaken desktops in sales) which NVidia isn't playing in successfully.

Jawed

 

[trinibwoy]

Come on man, now you're just being superficially anal. Here's one excerpt -

In Multithreaded Processing Unit 300, a thread execution priority may be specified for each thread type and Thread Selection Unit 315 may be configured to read thread entries based on the thread execution priority assigned to each thread type. A thread execution priority may be fixed, programmable, or dynamic. In one embodiment the thread execution priority may be fixed, always giving priority to execution of vertex threads and pixel threads are only executed if vertex threads are not available for execution. In another embodiment, Thread Selection Unit 315 is configured to read thread entries based on the amount of sample data in Pixel Input Buffer 215 and the amount of sample data in Vertex Input Buffer 220. Specifically, the thread execution priority may be tuned such that the number of pending pixels produced by processing vertex threads is adequate to achieve maximum utilization of the computation resources in Execution Pipelines 240 processing pixel threads.

A unified approach is the underlying theme of this entire document whether you choose to acknowledge it or not. The diagrams demonstrate it. The text demonstrates it. I'll give you this - if you can point out to me where there is evidence of dedicated pixel and vertex shaders in this patent I'll concede

 

[trinibwoy]

Making loadsa money now doesn't equal making loadsa money in the future.

Success isn't just a backward-looking concept.

I suppose I could draw a parallel with 3DFX or something, but why bother - I don't think NVidia's going to disappear or anything, but there are great chunks of the market (e.g. laptops, which have overtaken desktops in sales) which NVidia isn't playing in successfully.

Jawed

Oh Jesus man, now you're telling me a company can't be successful now if they fail in the future? So 3dfx was never successful? This argument is getting downright stupid. You get grades for exams you've done, you gaid paid for work you've done, you get praise for past accomplishments. Success is a measure of things in the past!!

Lol, even webster agrees http://www.webster.com/dictionary/success

 

[_xxx_]

Well, as the ATI guys said recently, NVidia has been promising great great things for the laptop sector for years.

And delivered it as well. Their current laptop lineup trounces ATI in every regard.

Handheld is a complete wipeout for NVidia. IGP isn't much better. No presence in consumer TVs.

And they also don't make cars and fridges...

EDIT: I'm not pro-nVidia here, but you seem to be quite biased against them as of late. I wonder why?

And my prediction is that both companies will offer products which are within 10-20% both price and performance-wise. And that will continue being so for quite a while. The company which manages to do more stuff for less money will make more cash.

 

[Jawed]

A unified approach is the underlying theme of this entire document whether you choose to acknowledge it or not. The diagrams demonstrate it. The text demonstrates it. I'll give you this - if you can point out to me where there is evidence of dedicated pixel and vertex shaders in this patent I'll concede

No. The unified stuff is after the fact.

This architecture would equally support a non-unified architecture. It makes no difference.

All NVidia has done is conceptually land-grab a unified embodiment of this patent. If you can't tell that this patent doesn't depend on unification, then I can't help you.

It's a texturing patent.

Jawed

 

[Xmas]

Because the focus of D3D10 is the addition of geometry shading. An important aspect of the enhanced geometry/vertex shading capabilities of a D3D10 GPU is to remove these tasks entirely from the CPU - and a side effect is that the more advanced rendering algorithms will put an extremely high load on GS/VS, in rendering passes that do no pixel shading.

A conventional architecture with a low GS/VS capability (say 1/4 or less of total) will be severely bottlenecked in these algorithms.

As long as the overall workload distribution remains similar to what we have now, you could also try to decouple pixel and vertex processing if necessary. Stream output comes to mind.

 

[_xxx_]

As long as the overall workload distribution remains similar to what we have now, you could also try to decouple pixel and vertex processing if necessary. Stream output comes to mind.

Please explain, streaming always sounds interesting but I have no idea what you're talking about.

 

[trinibwoy]

No. The unified stuff is after the fact.

This architecture would equally support a non-unified architecture. It makes no difference.

All NVidia has done is conceptually land-grab a unified embodiment of this patent. If you can't tell that this patent doesn't depend on unification, then I can't help you.

It's a texturing patent.

Jawed

That I will certainly give you - decoupled texturing doesn't depend on unification. But since the patent also allows for coupled texture-units I think it's more of a fleshing out of the unified architecture that somebody at Nvidia is thinking about. You don't think they went through all that trouble just to describe decoupled texture units and throw it away afterwards?

 

[_xxx_]

OT: to clarify what I meant above with patents (I'll use cars, since that's my job): we develop and test a tremendous amount of stuff which never gets into series and much of that stuff also never gets patented unless we really see it as useful to do so. We have a HUGE pile of interesting things in the "storage" and some of it gets used years later. Just have that situation right now with one of my toys. (EDIT: from 2 years ago)

 

[Xmas]

Please explain, streaming always sounds interesting but I have no idea what you're talking about.

D3D10 allows writing the output of the geometry shader to memory. This is called "stream output".

If there is only a small FIFO between VS and PS, slow pixel processing can stall the VS and slow vertex processing can starve the PS. Now if you have a huge FIFO instead (i.e. video memory) both units can work on completely different parts of the scene, using the FIFO to smooth out workload differences.

 

[Jawed]

And delivered it as well. Their current laptop lineup trounces ATI in every regard.

Except sales - boy you fell for that.

EDIT: I'm not pro-nVidia here, but you seem to be quite biased against them as of late. I wonder why?

Prolly because of a lack of criticality.

Jawed

 

[_xxx_]

Except sales - boy you fell for that.

I dunno 'bout current numbers, but it's not like these are carved in stone forever. And I see next to none ATI chips in high-end notebooks right now. Let's see how it works out this and next year.

 

[Geo]

Laptops is where the stark difference in the rhetoric is. No hedging on either side, no real caveats. Entirely different view of the reality over the next year. It's going to be interesting.