NV30 only as tech preview Monday...

Vince,

Why do you think the only way to extract preformance is threw linear increases in (a) Clock Speed, (b) Memory Speed, (c) Memory bus-with
Click to expand...

I do not think this. the quote was however specifically related to clock speed and Types reply.

However.. Nvidia is claiming 48GB bandwidth which is exactly 3x their supposed raw bandwidth of 16Gb. Which 3x is a common assumption for advanced bandwidth saving tech.. even Hyper Z 3.

I get the feeling some of you really are expecting some form of Gigapixel tech.
 
jjayb said:
Reverend said:
Personally, if I was known in a public forum as an employee of any one IHV, I'd stay out of arguments where it is especially about products other than those from my company (i.e. I'd stick to posting stuff about my company's products, be it in the form of defending incorrect posts about my company's product(s) and/or software). I'd refrain from "downplaying" another company's products/paper-specs/etc.

But that's just me.
Click to expand...

Where did he "downplay" another company's product? Where did he get involved in an argument about products from another company? All he did was call another user on his credibility. And backed it up with examples.
Click to expand...
OpenGL guy's first quote-and-answer to Username is "downplaying" the NV30 and NVIDIA. It is an "indirect downplay" of both NVIDIA and its NV30 "product". IMO, of course.

I am sorry if OpenGL guy feels slighted or annoyed by this but here is a simple example and is what I refer to as "don't argue if not necessary, especially if it has to do with a competitor's products, and you are a known employee of a IHV" :

OpenGL guy works for ATI

sireric works for ATI

The way of posting : OpenGL guy is unlike sireric

I will leave personalities aside. I would include responsibility as well as professionalism when you are a known employee of a IHV. There isn't even a need to call in question the "credibility" of a stranger. OpenGL guy is not a stranger. I have not found similar type of posts from known IHV employees such as Kristof, Ghost Of Envy, sireric and a few others compared to the emotion-influenced-and-driven posts by OpenGL guy.

OpenGL guy, this is not meant as an attack, just my opinion, which may or may not matter to you (or to many others). You are not alone (as an employee of a IHV browsing, if not actually participating, these forums)nor specifically picked-out by me... it's just the way things have progressed in these forums (and not just this thread).

I personally would not like to see "Dave-Barron-when-he-was-with-3dfx" type of public-forum posts here at B3D :) :) :) :) ;)
 
Hellbinder[CE said:
]I get the feeling some of you really are expecting some form of Gigapixel tech.
Click to expand...

Move away from the idea of "GigaPixel" or "3dfx" and into the idea that their [nVidia] starting over, from scratch. Alot has changed since their designed their previous core architecture back in 1997. Stating "this" or "that" former IHV is the wrong way of sayng it, but rather they as a collective have many talented thinkers who, if allowed, could do much to change the way their future architectures do the things they do.

If an IHV could claim a 10X reduction in necessary bandwith for their architectural plans 3 years ago, a company with the resources and talent of nVidia can do alot in the time that has followed - even if they choose a less radical departure from the traditioanl IMR appraoch.

I will be very disapointed if their isn't some sort of advanced underlying architecture that allows for larger gains in computational and bandwith effeciency than a linear advance such as HyperZ3, or a Crossbar, et al. They will have truely wasted a golden oppertunity, when they're starting from scratch.

Reverend said:
I personally would not like to see "Dave-Barron-when-he-was-with-3dfx" type of public-forum posts here at B3D
Click to expand...

Bummer, I miss him, was good [funny] times - this board is so much tamer than it once was... well, not exactly, but anyways. :LOL:
 
Reverend,

Call me stupid, but where in my post did I say anything about nvidia or NV30? So you're saying I can't disagree with someone without putting down whatever company they support or work for?

The quotes from Username I chose were mainly ones where they downplayed R300 and the impact it would have. I would say they guessed wrong, wouldn't you? The last quote I chose just showed that they haven't been so correct in guesses regarding nvidia's products either.

My personal opinion is that Username works for nvidia, but that doesn't matter to me. What matters is that the information coming from them are fact and not fiction.

Sorry, Reverend, for not fitting in with your version of the "ideal" poster, but you can at least explain to me how I can accept someone as being credible when their obvious (to me) lack of credibility is all I have to go on.
Reverend said:
I would include responsibility as well as professionalism when you are a known employee of a IHV.
Click to expand...
So if someone happens to mention that they work for, say, IBM, would you then hold them to a higher standard? What if they said they worked for Sony? How about McDonald's? If I make a post, I am posting as me: Take it, or leave it.
 
Hellbinder[CE said:
]Type seems to be indicating that the Nv30 is going to have considerably higher clock speeds. At least.. thats what i think he is saying.
Click to expand...

You could also read it as Type was indicating that the performance advnatgae over the R300 is more than 30%.
 
Vince, not all problems are solvable ... what GigaPixel promised was only possible since they did not restrict their solution space.

IMO the only thing where NVIDIA can truely innovate as far as bandwith is concerned is anti-aliasing ... for the rest they can just try to take some rough edges off compression and hierarchical Z-buffer (the crossbar is probably as fine grained as it is going to get). There might be a lot of gain to be had from those rough edges, but it seems unlikely.
 
They will have truely wasted a golden oppertunity, when they're starting from scratch.
Click to expand...

Tell me, if they really started 'from scratch' would you still design a part with or without register combiners?
 
DaveBaumann said:
They will have truely wasted a golden oppertunity, when they're starting from scratch.
Click to expand...

Tell me, if they really started 'from scratch' would you still design a part with or without register combiners?
Click to expand...

Tricky question here, Mister Baumann! I would say without and instead just do a clean sheet design as the R300, but as we know they haven't. Why? Is the performance of the 'old' register combiners faster than a whole new shader design - or is it just easier on developers since they don't have to re-program? You tell me. ;)
 
DaveBaumann said:
They will have truely wasted a golden oppertunity, when they're starting from scratch.
Click to expand...

Tell me, if they really started 'from scratch' would you still design a part with or without register combiners?
Click to expand...
Starting 'from scratch' doesn't mean you forget all prior knowledge. ;)

I guess the main difference between designing a refresh and starting from scratch is the question at the beginning: "What do we change?" as opposed to "What do we keep?"


PS.
Dave, got some more Parhelia shots? ;)
 
Is the performance of the 'old' register combiners faster than a whole new shader design - or is it just easier on developers since they don't have to re-program?
Click to expand...

If your drivers do the work via the shader then its transparent to the developer...
 
If silicon space is paramount (isnt it always?) the dump the old Register Combiners in a new architectureDX9 part IMO (but hey what do I know :) )
 
Interesting, it appears the NV30 will be flip chip as well. According to this source it will be a part of the geforce 4 family.....??

I am looking around for more info but seems there is nothing yet.
 
The only logical reason to continue to use the GF4 name is if the initial volume batches of NV30's are crippled, i.e. only partially functional.

They can turn off broken chunks and ship a less that inspiring product using the GF4 name.

So I doubt the story!
 
You must log in or register to reply here.
Back
Top