New ATI Console interview

[LunchBox]

how many gigapixels a sec for xenos?

supposedly 4 Gigapixels/sec...

or 16 Gigapixels/sec if you want to do the "coz of 4x anti-aliasing"

 

[Shifty Geezer]

I understood his comment to mean 256 GB/s between the logic and the memory on the Daughter die itself.

If you (or anyone else) already knew the hardware (from Dave's article for example) than seeing the 256 GB/s figures you know what it is, so would read it as such. If you take him word for word (and I did only because I finicky ), he definitely says there's a 256 GB/s input into the 'intelligent video memory' which only means the one thing, but I'm gonna stop flogging the issue now - I'm just repeating myself

And as for hidden, unspecified hardware Let's drag out the old 'PPU in XB360' rumours and 'Reyes rendering module in PS3' rumours . If theres more high-performance hardware there, why is the marketting not telling everyone how great it is? Has there ever been an ocassion when a games hardware has been unveiled, it's chips and performance laid down with schematics and metrics, and then when released there's been another processor involved that no-one knew about? :? If the 'intelligent video memory' isn't the eDRAM daughter die, you can colour me pink and serve me up as hors d'oeuvres on the captain's galley!

 

[DopeyFish]

And as for hidden, unspecified hardware Let's drag out the old 'PPU in XB360' rumours and 'Reyes rendering module in PS3' rumours . If theres more high-performance hardware there, why is the marketting not telling everyone how great it is?

Announcement schedule? Pending patents? Or not wanting some certain competition jump on a great idea? Who knows.

But regardless, it's not ATI who is control of Xenos, it's Microsoft. ATI, naturally would want to absolutely pimp out their GPUs at first chance. Microsoft having Xenos as part as an upcoming product probably see it as "having it's own time". That said, most if not all of the details will be known within the next month or so (most likely)

 

[Lord_MiL]

I realize that I am a newbie to this site (FIRST POST!!!) but it seems to me that there is a very simple answer that everyone has overlooked.

In the interview he said 256 GigaBytes / sec into the daughter die. However, elsewhere this has been stated as 32 GigaBytes / sec between parent and daughter die. It seems pretty obvious to me that this is the result of him misspeaking or whoever typed up the interview making a typo. Simply: 32GB/s = 256 Gb/s. The bandwidth is 256 GigaBits not GigaBytes. Seems like a simple mistake to me. Cheers


EDIT: Well after posting this I went back and actually read the interview in its entirety and realized that he in fact does say 2 TeraBits / 256 GigaBytes. So I guess my idea is totally wrong and now I'm just as confuzzled as everyone else

EDIT2: I've tried to scrounge up any other ATi interviews I could find and it seems that Huddy has consistently stated 2 Tb / 256 GB as the bandwidth between parent and daughter die. Is the article here on B3D possible in error?

 

[Megadrive1988]

you have a reason to be confused. there are two different bandwidths that are getting mixed up.

parent die (GPU shader core) to daughter die (eDRAM module) is 32 GigaBytes per second which is 256 Gigabits per second.


the inner or intra daughter die bandwidth, that is, the bandwidth between the eDRAM on the daughter die and the logic on the daughter die, is 256 GigaBytes per second which is 2 Terabits per second.

[Mod: Corrected Mega => Giga]

 

[DopeyFish]

you have a reason to be confused. there are two different bandwidths that are getting mixed up.

parent die (GPU shader core) to daughter die (eDRAM module) is 32 MegaBytes per second which is 256 megabits per second.


the inner or intra daughter die bandwidth, that is, the bandwidth between the eDRAM on the daughter die and the logic on the daughter die, is 256 MegaBytes per second which is 2 terabits per second.

I think you're both confused!

Megabytes?! What is this? The early 90s?

 

[Sonic]

Hey megadrive, change the mega in your post to giga. You had me confused with that last post!

 

[Alpha_Spartan]

....what about Nintendo?

ATi's only competition is itself when it comes to console. So they can say stuff like that.

 

[Megadrive1988]

you have a reason to be confused. there are two different bandwidths that are getting mixed up.

parent die (GPU shader core) to daughter die (eDRAM module) is 32 MegaBytes per second which is 256 megabits per second.


the inner or intra daughter die bandwidth, that is, the bandwidth between the eDRAM on the daughter die and the logic on the daughter die, is 256 MegaBytes per second which is 2 terabits per second.

I think you're both confused!

Megabytes?! What is this? The early 90s?

I was trying to avoid the confusion often generated with MB vs Mb.

one more time:

'parent' main GPU shader core ---><---'daughter' eDRAM module: 32 MB /sec

eDRAM ---><---logic on daughter die: 256 MB/sec

 

[aaaaa00]

you have a reason to be confused. there are two different bandwidths that are getting mixed up.

parent die (GPU shader core) to daughter die (eDRAM module) is 32 MegaBytes per second which is 256 megabits per second.


the inner or intra daughter die bandwidth, that is, the bandwidth between the eDRAM on the daughter die and the logic on the daughter die, is 256 MegaBytes per second which is 2 terabits per second.

I think you're both confused!

Megabytes?! What is this? The early 90s?

I was trying to avoid the confusion often generated with MB vs Mb.

one more time:

'parent' main GPU shader core ---><---'daughter' eDRAM module: 32 MB /sec

eDRAM ---><---logic on daughter die: 256 MB/sec

MB?!

 

[Inane_Dork]

...256 MegaBytes per second which is 2 terabits per second.

I think you're off by, say, 3 orders of magnitude.

 

[randycat99]

you have a reason to be confused. there are two different bandwidths that are getting mixed up.

parent die (GPU shader core) to daughter die (eDRAM module) is 32 MegaBytes per second which is 256 megabits per second.


the inner or intra daughter die bandwidth, that is, the bandwidth between the eDRAM on the daughter die and the logic on the daughter die, is 256 MegaBytes per second which is 2 terabits per second.

I think you're both confused!

Megabytes?! What is this? The early 90s?

I was trying to avoid the confusion often generated with MB vs Mb.

one more time:

'parent' main GPU shader core ---><---'daughter' eDRAM module: 32 MB /sec

eDRAM ---><---logic on daughter die: 256 MB/sec

MB?!

Uh, yeah, don't you know retro-technology is en vogue now? It's all the rage to have bus speed as fast as harddrive speeds. ATA-33, baby!

 

[Shifty Geezer]

one more time:

'parent' main GPU shader core ---><---'daughter' eDRAM module: 32 MB /sec

eDRAM ---><---logic on daughter die: 256 MB/sec

In case you've missed the hints, try replacing 'Mega' with 'Giga'.