Interesting interview with IBM (Cell, backwards compatability...)

[Titanio]

There's some pretty interesting comments from Tom Reeves, VP of semiconductor and technology services at IBM, regarding Cell and PS3. Some of them are, in fact, quite confusing (particularly re. PS2 - see the end)...

Turn Down the Heat … Please

On Cells with fewer operational cores:

Electronic News: Let’s look at design for manufacturability from a different standpoint. IBM has said it needs seven of the eight cores on the Cell processor to work for Sony’s Playstation. Will there be an aftermarket for chips with fewer operational cores?
Reeves: There are a lot of chips with six cores operational, and we’ve been thinking about whether we should really throw all of those away. We also have a separate part number for chips with all eight cores good. The stuff that’s going to be for medical imaging, aerospace and defense and data uses eight cores.

Electronic News: But might it be the less-expensive version of Playstation 3?
Reeves: It could, but I don’t think Sony has thought about offering that. That doesn’t mean there aren’t good uses for a chip with four SPEs [synergistic processing elements].

(I'm pretty sure we can agree there'll never be a Playstation3 using a Cell with fewer than 7 SPEs)

On (apparent) Cell yields:

It becomes a bigger problem the bigger the chip is. With chips that are one-by-one and silicon germanium, we can get yields of 95 percent. With a chip like the Cell processor, you’re lucky to get 10 or 20 percent. If you put logic redundancy on it, you can double that. It’s a great strategy, and I’m not sure anyone other than IBM is doing that with logic.

So - using logic redundancy, 20-40% yields on Cell with 8 cores then..? Or..?

And finally, the really confusing bit - or perhaps the really insightful bit - some comments on backwards compatability:

Electronic News: With the price Sony is going to charge, it can easily add that into the cost.
Reeves: Sony is very concerned about quality and backward compatibility. They want to get this right. They tested game after game after game. When there were about 40 Playstation 1 games that didn’t work properly, that didn’t pass their criteria for quality.

Electronic News: So does that mean the current Playstation 2 systems have a Cell processor?
Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.

Errr...? Did I miss something?

Also, is that "40 Playstation 1 games that didn't work properly" a reference to PS3 backwards compatibility? I can't see it being anything but in that context..and if that didn't meet Sony's standards, then perhaps that bodes very well for the level of compatability we'll see..

There's more at the link, though not much more of specific interest to PS3 - although he does also mention that Sony "burns in" its processors to test their reliability over time.

 

[TheChefO]

Electronic News: So does that mean the current Playstation 2 systems have a Cell processor?
Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.

What is he talking about?? When did ps2 get a power 440?

 

[Carl B]

Both some of the questions and some of the answers seem totally out in space.

 

[Titanio]

I think we're missing some of the flow of conversation between the interviewer and Reeves. The article starts by mentioning these are excerpts from a conversation. Though the "With the price Sony is going to charge, it can easily add that into the cost." comment that led into the BC-related answer does directly refer to the previous answer, but I don't know why Reeves suddenly started talking about BC. Maybe he wanted to hammer home a general point about quality.

 

[Jawed]

It becomes a bigger problem the bigger the chip is. With chips that are one-by-one and silicon germanium, we can get yields of 95 percent. With a chip like the Cell processor, you’re lucky to get 10 or 20 percent. If you put logic redundancy on it, you can double that. It’s a great strategy, and I’m not sure anyone other than IBM is doing that with logic.

So - using logic redundancy, 20-40% yields on Cell with 8 cores then..? Or..?

95% redundancy is typical for memory. So maybe that's what he's referring to as one-by-one

20-40% sounds to me like the yield for a 7-SPE Cell. Off the cuff, A64s seem to yield at about 50%, and then you've got speed grades.

Cell is problematic because the vast majority (99%+ ) are going into PS3 where there's no wriggling room. It sounds like it would have cost way way less if Sony needed 6 SPEs active.

But it's worth noting that Cells have presumably been rolling off a production line for around a year now, and they're prolly only 18 months away from halving in cost (assuming that 65nm kicks in) and the cost now is prolly lower than it was a year ago, simply due to tweaking.

I expect IBM can charge pretty much what it likes for the 8-SPE Cells, so the low yield/low-margin on those is prolly not a big deal. There's also plenty of value-add when they're sold to scientific customers.

Jawed

 

[kimg]

I had no idea even power 440 existed, so, i was looking on net now, and found this:
http://www.synopsys.com/news/announce/press2005/ibm_powerpc_pr.html

Is power 440 a customisable cpu,it could maybe be a modified one who works as cpu in PS2?

It give meaning, even. It was a japanese PS2 series who didnt work with some few PS2 games, if i remember correct.

 

[Titanio]

I mailed the article's author, and he confirms that the comment re. PS1 games is referring to PS3 backwards compatibility testing, and he says "In addition, Sony is now using the 440 Power processor in the PS2 systems." to clarify the last answer.

 

[Urian]

I believe that PowerPC 440 is for CED (Cell for Electronic Devices) and is combined with 2 SPE and a DDR interface, clocked around 667-700Mhz.

 

[Npl]

I mailed the article's author, and he confirms that the comment re. PS1 games is referring to PS3 backwards compatibility testing, and he says "In addition, Sony is now using the 440 Power processor in the PS2 systems." to clarify the last answer.

Faf mentioned Sony replaced the IOP with something else in the newest PStwo models. Bet the 440 is that else.

 

[Asher]

The 440s are the processors in BlueGene supercomputers.

 

[SPM]

Faf mentioned Sony replaced the IOP with something else in the newest PStwo models. Bet the 440 is that else.

Is he sure that is what the IBM guy meant?

I think it might have been talk about Cell yields, switching to Sony's burn in testing of the PS3 components in which he mentions Sony burn-in testing of all components including the PS2 components required for BC in the PS3, plus the 440 in the Brava TV, which might have got a little muddled in the reporting.

 

[Npl]

Electronic News: So does that mean the current Playstation 2 systems have a Cell processor?
Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.

Seems pretty straight information for me. Recent PStwo Models have the IOP replaced (MIPS R3000A CPU).

 

[Crossbar]

Faf mentioned Sony replaced the IOP with something else in the newest PStwo models. Bet the 440 is that else.

Pretty smart move of Sony if this is the case, it would have given them the opportunity to move all PS1 bc into sw-emulation on the PowerPC architecture years ahead of Cell.

Reeves: No, they have a 440 Power processor. It’s a 130-nanometer, single-core ASIC chip. It’s the same technology as if you buy a Sony DVD or a Sony Bravia TV. Sony is replacing all the Mips design points with Power design points.

You could interpret this as if they actually replaced the MIPS core in the EE with the 440 or some other PowerPc design. If that's the case, they pretty much also had the sw bc of the PS2 solved in advance. Then it's probably more or less is the GS emulation that is the last piece in the bc jigsaw, that Sony had to solve for the Playstation 3. If this is true, it seems they have taken care of quite a lot of work in advance already on the PS2 platform.

But, it sounds strange that the 440 is said to be on a 130 nm process as the EE+GS@90 is(was?) at a 90 nm process. Where is the economy in this?

Does anyone have a pic of the PCB of the latest PSTwo?

 

[Guden Oden]

Of course the MIPS core in the EE is not replaced by some kind of a PPC chip, these two architectures aren't even remotely binary compatible. You also can't eliminate the IOP CPU from PS2, since PS2 games can (and do) run program code on it. It has to be there.

The guy was either drunk and speaking about things he doesn't understand, or he was misquoted.

 

[Crossbar]

Of course the MIPS core in the EE is not replaced by some kind of a PPC chip, these two architectures aren't even remotely binary compatible. You also can't eliminate the IOP CPU from PS2, since PS2 games can (and do) run program code on it. It has to be there.

It can be replaced if you add emulation sw, but you are probably right that the reporter missunderstood a few things. If there is a 440 in the PSTwo, it would be interesting to know where it is used.

 

[Npl]

Its only about the IOP - the cutdown PS1 Processor sitting in most PS2`s. http://www.beyond3d.com/forum/showthread.php?t=31200 (The last post on the first page and following). Ripping apart the EE would be no good.
MIPS is easy enough to emulate, but from a hardware-perspective its surely more expensive to replace it with another chip which sole purpose is to emulate it (Still seems strange to me, but seems to be true). If Sony puts 440s everywhere the economy of scale could kick in, and aparently Sony had to pay licenses for the IOP...

 

[Mefisutoferesu]

OH, do you think the 440 could be the chip Sony's using to do the scaling in the PS3? That would explain it's relation to the PS2, they might have been talking about having the 440 upscale PS2 games and the interviewer mistook what was said?

 

[PiNkY]

IBM claims 8spe cell yields only at 20%

With a chip like the Cell processor, you’re lucky to get 10 or 20 percent. If you put logic redundancy on it, you can double that...
...There are a lot of chips with six cores operational, and we’ve been thinking about whether we should really throw all of those away

Not to go all doom & gloom, but along with the alleged blue ray manufacturing issues, it makes me wonder whether my SECD preorder certificate will actually net me a console come November

6SPE PS3, perhaps?

link

 

[Pugger]

So 20-40% yield for 7 functioning SPE's?

 

[PiNkY]

Sorry just saw that a link to the interview has already been posted. But perhaps this still warrants its own thread