Xbox 360 Shortage Caused By Low Xenon Yields...Is Cell Affected?

typoEDR said:
They are making treehouses out of defective CPUs.

It's possible they could sell them as single cores. Defective athlon 64s (defective in whatever area makes the cpu capable of 64 bit) become semprons, and pretty much all cpus with defective caches are rebadged as a cheaper chip. (duron, sempron, celeron)
 
Fox5 said:
It's possible they could sell them as single cores. Defective athlon 64s (defective in whatever area makes the cpu capable of 64 bit) become semprons, and pretty much all cpus with defective caches are rebadged as a cheaper chip. (duron, sempron, celeron)
sell them for what though ?

Mabye in webtv boxes sometime in the future ?
 
Jawed said:
Xenos appears to have an extra shader array for redundancy purposes:

b3d34.jpg


and one of the whitepapers says, mysteriously, that Xenon has redundancy built into the core (along with heaps of debugging hardware).

Jawed

Didnt we go through this in another thread and came to some other conclusions if i remember correct?

It would make more sence to have a "couple"/one or two alus per unit instead of one whole array.
 
Last edited by a moderator:
jvd said:
sell them for what though ?

Mabye in webtv boxes sometime in the future ?

Could be. Maybe cheap Windows MCE terminals that share code with the 360 or maybe even HDTV PVRs with MCE connectivity. If they underclock the bad boys they might even run without a fans. MS can and should go the same way as Sony wants to with Cell, it's just that Cell is probably way better suited for these tasks.
 
jvd said:
sell them for what though ?

Mabye in webtv boxes sometime in the future ?

I was actually referring to the Athlon and Pentium dual cores. As far as I know, there's been no half alive dual cores sold as single cores, but it's not out of the realm of possibility.
 
overclocked said:
Didnt we go through this in another thread and came to some other conclusions if i remember correct?

It would make more sence to have a "couple"/one or two alus per unit instead of one whole array.
There were no conclusions.

There's the die photo. You explain the four large pairs of identical die area.

Jawed
 
Jawed,
I don't think looking at a die photo is going to reveal wether xenos has an entire duplicate of its shader array or not. Of all things, why would they duplicate the shaders, and insert a full redundant array? If MS is going to pay for twice the shaders, why not use ALL of them in that case since they're there anyway?

No, I think this is just a case of layout on the actual chip die. Some components appear in one column, others appear in the other. For whatever reason. After all, actual chip layout rarely matches block diagrams very closely or if at all.
 
Guden Oden said:
Come again?

No one is suggesting an entire duplication of all shader arrays. (Don't know where you got that idea.)

The die photo (as marked up) suggests 4 shader arrays...one of which is a redundant array for the 3 production / working arrays.
 
Jawed said:
There were no conclusions.

There's the die photo. You explain the four large pairs of identical die area.

Jawed


If we would to assume that each small block is a part of an ALU(and logic of it) i count 12 per square and theres 4. 13 if the ones sitting at the edges in the left column(not sure if they are exactly identical). But my guess would go towards rather 13 than a whole square because it would make more sense in a redundancy purpose. Why waste 16 alus with all their dedicated logic when you can "waste" just four of them and gain better endresult?

Im not saying that the above is true based on a dieshot but if you compare it that way i take a shot in the same manner. Ie just because there are three arrays they wouldnt nessecary "hang in a group" like three grapes on a branch all structured and fine like in a diagram or the description.

If you could pick up the older thread i think there was some other reasonable theories besides your own witch also is reasonable btw.
 
Thats a large proportion of die space wasted for redundancy purposes. Improve yields? Sure. However simply not having the chip so big (aka having 3 arrays instead of 4) would probably achieve the same effect and would be cheaper to manufacture. Or couldn't they have made provisions for a couple of redundant ALUs in an array instead of taking an whole array out?

Disappointingly, thats alot of wasted transistors (if true).
 
It is common knowledge that, for the first couple of months at least, Cell will be fabbed only at East Fishkill. The first slide at Sony's E3 Press Conference said Cell 90nm@ East Fishkill.

Nagasaki 90nm is still doing PS2 parts, Nagasaki2 is going 65nm for eventual Cell production in the next 1-2years. Toshiba at Oita will start Cell production at 90nm some time soon.
 
Back
Top