[digitalwanderer]

Originally Posted by R300King!

He says, "Nvidia is scheduled to introduce a new GPU series to target the high-end segment in early March, according to Paul Sun, general manager of Nvidia Taiwan Sales. By February 1, a complete lineup of its GeForce 7-series GPU line will all be available worldwide, said Sun."

Well, according to this article it seems that he is talking about a new series to be announced, the 8-series maybe, to be announced beginning of March. And by Feb. 1st they will have their 7-series cards(7800GTX 512, 256, mid, low end, etc) all widely available. That's the way I read that.

Either that or they will be coming out with a newer series 7 card, but I can't tell which they mean.

[Megadrive1988]

Originally Posted by Redeemer

How can Nvidia make 32 pipes @ 700-750 available? ATI's low pipeline high shader throughput design @ reasonable clocks makes all the more sense now after their R580 launch. Is there anything else Nvidia can do besides adding pipes?. Obviously what good is a product with no availability. Of course G71 specs are not in stone yet or is it?

G71 is not going to have 32 pixel pipelines.

it's 16 full pipelines (since there's only 16 ROPs) and *maybe* 32 pixel-shader units, as G70 has 24 pixel-shader units.

[Chalnoth]

Originally Posted by Joe DeFuria

It would also make the G71 a curious part....it would increase the ROPS and Z and leave the shading power untouched...almost directly contrary to the ATI approach.

And directly contrary to any logic.

[Chalnoth]

Originally Posted by Megadrive1988

it's 16 full pipelines (since there's only 16 ROPs)

No, because the ROP's are decoupled from the shader pipes. I believe they also run at a higher frequency typically (that is, I believe they run at memory frequency).

[Arun]

Originally Posted by Chalnoth

No, because the ROP's are decoupled from the shader pipes. I believe they also run at a higher frequency typically (that is, I believe they run at memory frequency).

On NVIDIA hardware at least, the ROPs run at the normal core frequency. It is the memory controller which runs at memory frequency.

it's 16 full pipelines (since there's only 16 ROPs) and *maybe* 32 pixel-shader units, as G70 has 24 pixel-shader units.

Am I legally authorized to hit the next guy using the term "full pipelines"? jk. Furthermore, any GPU has one "full pipeline", considering you'd have to include a number of other units for such a stupid nomenclature to be correct. I understand the "maybe" though, even if it shouldn't be there.

Yeah, apart from being a very short news piece, even for Digitimes, the two statements appear contradictory, unless the February 1 is a typo.

I don't see how. That would simply relate to 7300 and/or 7600s. The lineup would be "complete" but one part of it would be refreshed in early March. I'm a bit skeptical NVIDIA can properly execute for February 1 (err, where ARE the review samples?) , unless it's 7300-only, which it probably is.

G70 has 16 pixel pipes (ROPS), 24 TMUs, and 48 "pixel shader ALUs".

Per-PS-pipeline, ATI also has 2 ALUs; they are a little bit slower for pure arithmetic, but that's more than compensated by the fact NVIDIA has to use one cycle of ALU1 for each texture operation. As such, if you really wanted to say G70 has 48 pixel shader ALUs, you'd have to say ATI has 96, imo.


Uttar

[digitalwanderer]

The electrician in me really wants to start using the term "conduit" instead of "pipeline"...

[Moloch]

Originally Posted by digitalwanderer

The electrician in me really wants to start using the term "conduit" instead of "pipeline"...

you did electric work eh?
they call you sparky?

[Carl B]

How about putting the G71 discussion in terms of fragment 'quads' - we can all agree on that as far as the architecture goes, right?

When people speculate as to 32 'pipes,' I think it's clear that regardless of how one might define that word, what is meant is eight 'quads' vs G70's six.

[Ailuros]

If there's going to be one unit amount that will most likely stay identical in G71 then it's the ROPs.

[digitalwanderer]

I dunno about calling 'em fragment quads, it makes it sounds like there is math involved.

Originally Posted by radeonic2

you did electric work eh?

My dad owned an electrical contracting place, I grew up earning $0.25/hr sorting returned inventory into the warehouse and cleaning trucks. I was an electrician's helper before I could drive, and when I could drive I alternated as driver/ditch digger/apprentice...whichever the situation warranted.

When he found out I could type, office work was included until office manager could have been aptly fit into me job description too.

I don't think I've ever been as happy as when he finally closed that place down about 12 years ago.

they call you sparky?

Everyone calls you sparky!

[Moloch]

Originally Posted by digitalwanderer

I dunno about calling 'em fragment quads, it makes it sounds like there is math involved.


My dad owned an electrical contracting place, I grew up earning $0.25/hr sorting returned inventory into the warehouse and cleaning trucks. I was an electrician's helper before I could drive, and when I could drive I alternated as driver/ditch digger/apprentice...whichever the situation warranted.

When he found out I could type, office work was included until office manager could have been aptly fit into me job description too.

I don't think I've ever been as happy as when he finally closed that place down about 12 years ago.

Quite a story you got there mate

Everyone calls you sparky!

Especially your wife huh?

[Ailuros]

Light travels faster than sound; that's why some people appear bright until they open their mouth

Back on topic availability will depend on final frequency and the combination of the chip's complexity. I'd love to be proven wrong but if I am to trust those rumours that speak of 8 quads with =/>700MHz frequencies, I'd be very surprised if availability would be acceptable. Not the first time I'd be surprised though...

[Jawed]

I'm with Joe.

And G70 has more capable ROPs than NV40, so I see no reason why G71 shouldn't go further. Especially if NVidia was anticipating GDDR4.

CJ, if you can get your friend to produce the CATZ for the Go series parts, that would make things a lot more definitive.

Jawed

[neliz]

Why would G71 be architecturaly different from G70?

I mean will it be so much different? I don't think that's a wise step to take right? changing architecture, upping clocks etc. on a high end part with a new proces..

[Ailuros]

Originally Posted by neliz

Why would G71 be architecturaly different from G70?

I mean will it be so much different? I don't think that's a wise step to take right? changing architecture, upping clocks etc. on a high end part with a new proces..

Did ATI just increase the amount of ALUs in R580, or where there additional minor improvements also necessary?

[Jawed]

If NVidia is making G71 in approximately the same timescale as G72, G73, G74 - the supposed "easier to build" parts that lead NVidia's way into 90nm, then it seems to me that G71 can't really benefit greatly from the smoothing of the 90nm path from G72 et al. There's not the 6-month gap that the "smoothing of the path" argument implies.

In other words, I expect G71 to be more like a G70 refresh, i.e. the same capability as G70 on 90nm with perhaps slightly faster clocks and more ROPs.

Although there's an argument that says that NVidia's refreshes are normally a little more significant than that. My history's rusty, though: how many high-end NVidia refreshes have coincided with a new process?

On the other hand, if you believe that 90nm G7x is around 4 months late, and G71 is appearing 6-months after G72 should have appeared... (Then again, G71 should have appeared before Christmas - GTX-512 was clearly shoved on stage as an emergency fallback.)

Jawed

p.s. I vote to have this thread merged with the other G71 thread

[Neeyik]

Threads merged and renamed to be all cute and togethernessy...

[neliz]

Originally Posted by Ailuros

Did ATI just increase the amount of ALUs in R580, or where there additional minor improvements also necessary?

Minor improvement, bugfixes etc. they're allways necessary but G71 sounds like a new architecture.. a completely new level .. according to rumours.. it wouldn't be a refresh, but a complete rehaul.

I'd have to go with Jawed that the g71 is running into "problems" as well and that there is a plan "B" on the way now..

[Razor1]

Hmm I think nV was/is having issues with the .9 process, so far all .9 chips have been delayed to some degree.

[Ailuros]

Originally Posted by neliz

Minor improvement, bugfixes etc. they're allways necessary but G71 sounds like a new architecture.. a completely new level .. according to rumours.. it wouldn't be a refresh, but a complete rehaul.

I doubt it'll be a new architecture in the sense you see to be putting it in; the transition from 110nm to low-k 90nm shouldn't have been as easy as 130 to 110nm since the libraries aren't as close as in the second case for one and what I was hinting at is that when you increase a specific number of units you have to come to minor changes too in order for the architecture to sustain or increase it's efficiency and not losing any in the end.

A complete "rehaul" would be something like G80, a G7x refresh on a smaller manufacturing process w/ or w/o additional units is hardly a "rehaul" let alone a complete one.

I'd have to go with Jawed that the g71 is running into "problems" as well and that there is a plan "B" on the way now..

Like increasing ROPs amount which doesn't make any particular sense in my mind right now? I don't see any plan "B" yet; the amount of functional units seems pretty clear to me right now from the rumours that have been circulating lately, what I'm uncertain about is the quite high clock frequency that is being attached to those rumours.

If there's something severely "wrong" with it, another possible scenario would be no G71 at all in the end for which there are no indications yet.

[_xxx_]

Originally Posted by Ailuros

I doubt it'll be a new architecture in the sense you see to be putting it in...
...A complete "rehaul" would be something like G80, a G7x refresh on a smaller manufacturing process w/ or w/o additional units is hardly a "rehaul" let alone a complete one.

Well to me, rehaul != new architecture. They will certainly tweak some stuff, I believe we'll see some sort of non-optimized aniso and HDR+AA (even if it's just a dirty hack), maybe even more.

[_xxx_]

Originally Posted by Joe DeFuria

It would also make the G71 a curious part....it would increase the ROPS and Z and leave the shading power untouched...almost directly contrary to the ATI approach.

Not if they implement a new scheduler kinda like ATI's...

[neliz]

Originally Posted by _xxx_

Not if they implement a new scheduler kinda like ATI's...

Then there would be a LOT of stuff happening at once for a step which was originally listed as a die shrink..

[Jawed]

Originally Posted by Ailuros

Like increasing ROPs amount which doesn't make any particular sense in my mind right now?

Might solve G70's inability to keep up with intense AA.

Jawed

[_xxx_]

Originally Posted by neliz

Then there would be a LOT of stuff happening at once for a step which was originally listed as a die shrink..

I was surely just kidding, but nothing's impossible. I mean, all we know for _sure_ is that there's a new chip coming and nothing more than that