Will R580 have 48 shader 'pipes?'

[ERK]

I thought we might like to get our predictions in on this before all the R520 reviews hit.

I think it is quite likely to have the 48 shader processors, just because the leaked ATI slides all seem to hold up to the X-X-X-X rumors, putting R580 at 16-1-3-1, and thus 48 "pipes" in the same way as X1600 has 12.

Enjoy!

ERK

 

[trinibwoy]

Yep, solely based on Dave's assertion that R580 = xRV530

 

[tEd]

To make any kind of prediction i first want to know how such shader processor looks like and what it exactly it's capable of.

 

[ERK]

Yeah, I expect opinions could change after seeing the details. But will we know too much by then (take the fun out of the poll)?

 

[LVSeminole]

Does anyone have any "rumors" of what the R580 will have under the hood? Just a refresh of the R520, or something different, architecture changes as well as more pipelines, speed?

LVS

 

[Razor1]

Does anyone have any "rumors" of what the R580 will have under the hood? Just a refresh of the R520, or something different, architecture changes as well as more pipelines, speed?

LVS

Most likely no more "pipes" just 16

...waits for Rys to Pounce on me....

 

[AlphaWolf]

Does anyone have any "rumors" of what the R580 will have under the hood? Just a refresh of the R520, or something different, architecture changes as well as more pipelines, speed?

LVS

Most of the rumors I have seen suggest more of the same, and 580/590 would seem to be a refresh based on nomenclature.

 

[epicstruggle]

Most of the rumors I have seen suggest more of the same, and 580/590 would seem to be a refresh based on nomenclature.

I could be imagining it, but i thought Dave was suggesting otherwise.

epic

 

[ZoinKs!]

Well, there was that tidbit about rv515 = r520 and rv530 = r580 which implies there's an architecture difference worth talking about.

Also, it seems the term "pipeline" is losing it's relevance. Or at least modifying it's definition.

 

[Moloch]

I voted no.
If the R520s 16 pipes are all powerfull then I think getting 48 of them would be quite the challenge considering the trouble ati had with the R520.

 

[Tim]

I voted no.
If the R520s 16 pipes are all powerfull then I think getting 48 of them would be quite the challenge considering the trouble ati had with the R520.

The pipes are not all powerfull, Ati gets high performance by running the pipes at a very high effiency not because the pipes has very high raw power. Also the troubles with the R520 was not related to the transistor count (they had problems with the R520 and RV530 but not with the R500/C1 and R515).

 

[Jawed]

Nor R580, apparently.

Jawed

 

[Ailuros]

IMHO the R580 will most likely have 48 ALUs. I wouldn't really count them as fragment shaders, because I've no idea how many FP32 operations each ALU can handle per clock.

 

[ERK]

But how can the R580 have a similar architecture to the RV530 (as mentioned above, etc.) and yet have a different pixel pipeline with more ALUs? This would be a major difference, no?

 

[Rys]

The pixel 'pipeline' isn't different. R580 is, pretty much, just wider. Easiest to think about it that way, I find. Which in turn leads to much more ALU ops per pixel thread because of that, which is RV530 and R580's main ethos. Get ALU:TEX up really high.

 

[ninelven]

I think the more interesting question is will all the alus be the same... or rather the one that follows from that...

 

[Jawed]

Well, as I keep pointing out, how come Xenos fits 64 shader pipelines into only 232m transistors?

http://www.beyond3d.com/forum/showpost.php?p=574627&postcount=64

http://www.beyond3d.com/forum/showpost.php?p=574679&postcount=66

I'm not going to repeat the arguments I posted in that thread. Should have put them in this thread, really Oh well.

You can safely ignore Chalnoth's comments on this subject in that thread, he admitted that he wasn't actually paying attention.

Jawed

 

[Jawed]

The pixel 'pipeline' isn't different. R580 is, pretty much, just wider. Easiest to think about it that way, I find. Which in turn leads to much more ALU ops per pixel thread because of that, which is RV530 and R580's main ethos. Get ALU:TEX up really high.

As I keep saying, ATI's credo is 3:1 ALU:TEX.

Jawed

 

[Moloch]

The pipes are not all powerfull, Ati gets high performance by running the pipes at a very high effiency not because the pipes has very high raw power. Also the troubles with the R520 was not related to the transistor count (they had problems with the R520 and RV530 but not with the R500/C1 and R515).

Well ok.
What were the problems with the R520 related to then if not an extremely complex chip?

 

[Rys]

What were the problems with the R520 related to then if not an extremely complex chip?

The main problems were electrical. That the chip is big, or complex, really isn't the issue, although it is related. It's essentially the endeavour to get good electrical signal quality, a solid ground plane and predictable current flow into and out of the chip. That's always the case no matter what the size or transistor count (which is such an asshat metric for me).