If you're still wondering at Cell wafer's "mystery chip

[JF_Aidan_Pryde]

It's not :
- EE + GS
- Super large 4 way switch
- NV5x
- IOP

Email response from IBM:

These are process and circuit characterization structures. It is typical to devote significant amounts of area to these structures in early runs of a new processor.

Bah!

 

[PC-Engine]

Edited by moderator:


Do not troll on the boards. Last warning.

 

[Sonic]

Did you actually email an engineer at IBM? If so are there any other details that would be helpful from those emails?

Anywho, does htis mean the PS2 on-a-chip and GPU are on the same chip?

 

[one]

Anywho, does htis mean the PS2 on-a-chip and GPU are on the same chip?

I think "PS2 on-a-chip and GPU are on the same chip" etc. is from Version's speculation (my answer to it) and its continuation. The Email response from IBM to JF_Aidan_Pryde confirmed it's indeed for the test purpose.

 

[JF_Aidan_Pryde]

Did you actually email an engineer at IBM? If so are there any other details that would be helpful from those emails?

Anywho, does htis mean the PS2 on-a-chip and GPU are on the same chip?

It's from an IBM engineer but unfortunately he said nothing more on the testing circuit.

Thanks for the good moderation Sonic.

 

[Shifty Geezer]

Why would anyone think otherwise? Suggestions of EE+GS seemed nonsensical as these are Cell samples and not components for the PS3.

 

[PC-Engine]

Some people just need to hang on to the hope of the CELL chip being more than what it really is...

 

[Farid]

Why would anyone think otherwise? Suggestions of EE+GS seemed nonsensical as these are Cell samples and not components for the PS3.

It could have make sense, since the PS3 was the first (and main) target for the Cell CPU.
The first (final) mask could be used to produce other PS3 chips if they saw fit. It wasn't a crazy idea since creating/testing a mask is not trivial at all.

 

[j^aws]

JF_Aidan_Pryde, thanks for clearing that up with the IBM engineer...

At any rate, expecting more than 1 CELL is highly optimistic even if that wafer was suggesting otherwise...but hey who knows!

 

[Lumine]

Could the mystery chip be the switch required for an array of 4 or more cells? Seems logical to me.

 

[Guden Oden]

Lumine,
Maybe you should re-read the first post in this thread, it seems you didn't quite grasp its meaning the first time 'round.

 

[Lumine]

heh...heh... uhh... Oh, he meant "switch" like a processing device.... I uh.. I, uh, thought he meant a stick that you get beat with when you misbehave.. uh.. yeah.

My bad, completely didn't see that line

 

[Guden Oden]

No man, as the first post states, that "mystery thing" "chip" isn't a chip at all, much less a switch for four other cells (it'd be extraordinarily LARGE for a switch, on the order of a typical x86 processor, or close to it). It's an area used to test the manufacturing process, it serves no other function. Okay?

 

[randycat99]

OK, fair enough. However, I am still unclear as to does this mean that the layout shown is a bigger die surface than will be used in production then? Will they trim the die down for production (removing the areas of testing circuitry), or will the space remain the same, and something else useful be put there?

 

[a688]

OK, fair enough. However, I am still unclear as to does this mean that the layout shown is a bigger die surface than will be used in production then? Will they trim the die down for production (removing the areas of testing circuitry), or will the space remain the same, and something else useful be put there?

Lets put on our common sense thinking caps. The circuitry is there for testing. Looking at the wafers of mass produced chips, the chips are closer together on the wafer. Honestly, if you are trying to produce the most of a chip that you can, why fill the wafer with extra components if you can use that same space to put more of the chip you really need? In short they will be able to fit more cores on the same size wafer when they go to into full production.

 

[randycat99]

No one is suggesting they would just remove the testing circuits but leave that space unused when sizing the dies for production. What I was asking is if it is equally plausible that there was something else that may occupy that space to better effect, rather than simply pushing all the cores closer together after the testing circuits are removed from the picture. Granted, you could fit more units on a wafer by doing one thing, but this does not eliminate the possibility that something else might be planned to go there, as well. You never know... I don't think you can say one way or the other for sure, at this early juncture.