Cell Processor abilities may be luck of the draw
In a recent interview at DailyTech, they sat down with IBM vice president of Semiconductor and Technology Services Tom Reeves. He discussed the abilities of the cell processor and, more importantly, the inability to successfully create many working 8-core units. The cell processor is so complex that "IBM even accepts chips that have only four out of the eight cores working." Egads! Hopefully those aren't going into our PS3's, right? No worries. However...Sony won't allow cell processors with less than seven working cores. But they aren't clear when it comes to using seven or eight. Will it be a throw of the dice when we buy a PS3 as to whether we have a working 8-core processor or one with only 7 working bits? Would making the lower SKU accept the 7-core be any better? If that seventh one shorts out and you had the PS3 with only seven working cores, you have to send it back to Sony or IBM or Toshiba to get it fixed. That's a pain and it depends on your warranty. No matter what, they need to get these things working because November is pretty close.









Reader Comments (Page 1 of 1)
Fan @ Jul 14th 2006 2:30PM
No. 1 core is reserved for redundancy, that means there is allways only 7 cores. Another one is reserved for the OS so that leaves 6 cores total for game use.
Not to bitch but this is also old news.
SuicideNinja @ Jul 14th 2006 2:45PM
I personally think it's bad practice to call the SPE's cores, because SPE's are not equivalent to the actual core of the CELL (they are all dependent on the single "main" core, and cannot do anything on their own, unlike multiple cores in other processors). It's like thin-client vs full-client (a full-client has a local store and can be it's own server). I know the source article made the unfortunate "core" reference, so I'm not blaming you, Nick.
Anyway, it wouldn't matter if an 8-SPE-functional cell processor made it into a PS3. 1 SPE is disabled for all PS3's so the spec is to have 7. They have to keep the spec consistent as not to cause problems for software development (present and future).
I'm just hoping that a "cell fiasco" doesn't replace the "disc read error" fiasco. Or worse...we get both fiascos in the future. This is the problem of jumping on these technologies.
Early adopters may have some chip problems, but this will likely be ironed out a year or a year and a half after launch.
PS3 needs good press @ Jul 14th 2006 3:08PM
It's old news that IBM is allowing sub 8-core Cell chips to be used? It's old news that Sony isn't demanding all 8 cores working, just 7? Or maybe you are referring to the posted article which is dated from yesterday as old?
From the article--
"In the interview, Reeves also talks about failure rate as there is a possibility that one of the cores in the Cell will "blow" at any given time."
"When asked what would happen if a 7-core PlayStation 3 ends up losing another core during operation, Reeves stated that the user would simply send the unit back for replacement. Unfortunately, this only applies if the console is still under warranty -- if it isn't, the console is dead."
Statements like these strike fear into the heart of men. The cell sounds awfully delicate. Don't leave any store that you've bought your PS3 from without first getting a 2-3 year warranty. For Pete's sake, tell me something GOOD about the PS3 for once. These articles are painful.
Kid Muscle @ Jul 14th 2006 3:28PM
I personally hate to hear how things are being used, and everything that could go wrong being reported as if it WILL go wrong... >_>
If I am gonna buy a chicken dinner, just let me buy it and eat....don't tell me about how you killed it, and the possable disease's that chickens can transfer.
Just give me my PS3, and let me play it....IF anything goes wrong (Which applys to EVERYTHING in life) I will deal with it when it happens.
Paul @ Jul 14th 2006 3:42PM
Post 3 makes me think of an analogy:
The Ps3 is a hot dog.
If you take all the parts seperately they all seem like things that arnt very good and you really dont want. But whne you combine them, it creates something magical.
Eric @ Jul 14th 2006 6:06PM
#1 is right. This is old news not even worth posting. Accepting that 1 core will be "dead" (whether due to faulty production or deactivating a working SPE) equals higher production yields.
Though most of what Sony's been doing is a major cluster, manufacturing is one area Sony does have an advantage and I'm sure there's major differences in 7 SPE yields vs 8.
OkayItsMe @ Jul 14th 2006 6:16PM
Wait, I'm reading the possibility of 4 out of 8 cores will be working. That's old news? It's new to me! Also, I didn't know a core could fail during use. Are production yield problems going to affect supply?
Vince @ Jul 14th 2006 6:29PM
"If that seventh one shorts out..."
There's a lot of things that can go wrong that can leave a console useless, a lot of if's.
If one tyre a my car blows I have to stop and fix it, if the 0 button on my mobile breaks it becomes useless, if my heart stops beating the rest of me dies.
While I'm in "fanboy" mode I have to say what an amazing sight the cell is, it's amazing how far we've come in so short a time.
I'll never understand the gaming luddites who don't embrace this sort of innovation.
Tom @ Jul 14th 2006 6:56PM
Mmm, magical hotdog...
Kid Muscle @ Jul 14th 2006 7:21PM
I agree with #7...
Same chances of: "If I drop my wii remote controller will the impact break, or injure the gyroscope sensor inside?
If I buy an Xbox 360 from the first batch, will I get one of the ones that overheated or broke like many others got.
There's an "IF" situation for EVERYTHING in gaming and life!
sputnik @ Jul 14th 2006 8:01PM
But theres no "If" in yields lower than 20% being the norm when they are usually 95%. With yields that low the chip will cost them a fortune. Which means the Cell probably won't be adopted by anyone else and major shortages this fall for PS3...which means cost will be slow to come down...which means price of the PS3 will be much slower to come down.
No wander they are rushing to get the chips made in 65nm process. It's a last ditch effort to improve the horrible chip yields.
Wow...I think the world may be watching the great business blunder of the decade. Financially floundering behemoth throws all it's biscuits in one basket and rolls the dice.
Next year will be very interesting. If Blu-Ray goes the way of the UMD...(all fanboyism aside this is looking more likely every day) Sony may find itself completely sunk. It's Playstation division is the only profitable division. Without it the are in some deep crap.
ozymandias @ Jul 14th 2006 8:02PM
actually Fan, there are now 7 spes used on the ps3 cell. if you read the article, you would know that the 8 spe cell chips have a seperate part number.
"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."
So that means that the cell has 1 general cpu and 7 SPEs, with 1 reserved for fault insurance. that's still 8 "cores" just not the way sony spins it.
Flux @ Jul 14th 2006 11:11PM
First, why do bloggers who obviously have very little technical knowledge decide to post speculative articles like this constantly? I enjoy reading the news here, but it has been common knowledge for a long time that cell chips with less than 7 working SPEs will be used in other products...I would speculate that they will end up in items like camcorders, TVs, DVD players, etc. These are all products that need a processor that crunches numbers quickly, so the cell is ideal.
Which leads into my second point, in that the low yields from the manufacturing of the cell processor does not necessarily equate to an earth shattering loss for Sony. Where other chip fabs produce garbage, Sony has a process that just creates chips for its other products when it fails. And on top of that, people are constantly misquoting that IBM article: The article stated that they are getting 10-20% yields on the cell processor, but having one SPE redundancy can double that number. That means that they are probably getting 20-40% yeilds for the PS3. Not great, but it's a hell of a lot better than what all of these blogs are saying.
Eyecandy @ Jul 15th 2006 12:00AM
I want a set of rose colored glasses like the rest of the Sony Fanboys. How can ANY problem with production of the heart of the PS3 be considered anything but a problem. Much less a yeild of 10%-20%(or 20-40% depending on the color of your glasses).
Fan @ Jul 15th 2006 12:02AM
@ ozymandias
Yes that is what I am saying. Im not going to use the word "core" anymore because, as stated above, its rather misleading. They try to produce chips with 8 SPU's but they are hard to produce flawlessly. So instead, to save on costs, they accept chips with 8 or 7 working SPU's. That way they can save on costs and ensure a more steady production flow.
Of course there is one PPE controlling the whole show but you would expect IBM to know how to make them.
For a better understanding of the processor go here:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2379&p=3
Mutsch @ Jul 15th 2006 12:56AM
You guys really know what you are talking about ;-(
What Reeves talks about is pretty much standard in the chip industry. Yields of such highly integrated chips are usually very low, especially in the beginning. The same is most probably true for the Xenon chip of the XBOX 360. With new processes, e.g. 65 nm, this will improve over time.
Regarding the number of SPEs: the PS3 uses Cells with 7 active SPEs only. There is no redundant SPE! The Cell can not replace a defect SPE after the chip is build.
When chips are tested at IBM, defect SPEs will be deactivated via laser activated eFuses. Then you will get Cells with 8 SPEs, 7 SPEs and so on. The PS3 uses the ones with the 7 active SPEs. Again, this is industry standard. It's done with processors, memory and other complex chips.
While the original news was absolutly correct, all followup news is bascially twisting the facts a bit, intentally or not. I guess that's the price we have to pay having blogs these days ;-(
Highlander @ Jul 16th 2006 3:02AM
Sputnik, stop spounting crap/ Where in God's name did you get the idea that production yeilds are 95% for any modern microprocessor? ROTFLMAO at the very idea.
The 10-20% Reeves is talking about is for 100% perfect processors. PS3 requires only that 7 of the SPEs work, not all 8. So if any one of the 8 has one or more faults, it can be disabled during burn-in testing. If someone wants to do the statistics they can. What it boils down to is that regardless of the number of defects on the specific chip, as long as the PPE, and 7 out of 8 SPEs are working, the chip is 100% for PS3 use since the spec only requires 7 SPEs. There is no requirement that specific SPEs are functional, only that 7 out of 8 of them are. ANY one of the SPEs can be problematic and the chip can still work at 100% for PS3 use.
The very fact that it doesn't matter which SPE's work as long as 7 out of 8 of them do, makes the impact of a defect much, much less. So, if 20% of all Cells manufactured are perfect with all 8 SPEs working; a far greater number of cells will be produced with 7 out of 8 SPEs working. The really, really interesting yeild number is not the 100% defect free yield. the interesting number is the yield of cells that have 7 out of 8 SPEs working. That number will be much higher.
As usual, the various folks at the tech sites ought to be ashamed of their ability to analyze the news, and the authors of comments on this site an others need to review their motives.
It's also kinda curious to me that PS3Fanboy (supposedly a PS3 fan blog, is asuch an anti-PS3 spin. Perhaps a little too much playing with their joystiq's?
FullArmageddon @ Jul 16th 2006 9:55PM
So, how do you know it one of the SPE's burn out on your PS3? Will it just not work, or will it just perform really slow?
J. McNair @ Jul 17th 2006 3:03PM
@FullArmageddon
It will just not work. The PS3 BIOS will detect a dead core and games simply won't boot. Let's hope Sony's curtomer service has improved enough that customers won't have to go through heck getting PS3's repaired or replaced.
bub @ Jul 17th 2006 3:47PM
Not sure about the writer of this article. It was determined long ago that the ps3 would only use 7 of the eight SPE's (the writer of this article actually uses the word 'bits'-- incredible level of ignorance). IBM also explained that with logic checking, the yields are about 40-50%, and that Sony would be using that technique. He also explains that Sony and IBM went into this, knowing that, that they may have uses (not ps3, obviously, unless of course you wrote this article) outside of Sony, and that the ones with 8 working cores may be set aside for medical stuff. All you have to do is google-news ps3.