Not all cores of the cell processor work

[PINKO]

http://www.dailytech.com/article.aspx?newsid=3295

maybe some tech guy can enlightin me more on this

[thetshell]

Short version: the one chip will have 8 "cell" processors...of which the PS3 needs 7 functioning to work properly. Assuming the article is accurate, here's the bottomline:

"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."

[mykevermin]

"Buy the PRP" is the lesson to be learned.

[MusicNoteLess]

[ :hot: :hot: :hot: :hot: :hot: :hot: :hot: :cold: ] = :joystick:

[:hot: :hot: :hot: :cold: :hot: :hot: :hot: :cold: ] = :oldman: ->:wall: if not under warranty

could this be an example of a core processor blowing a fuse?
http://www.youtube.com/watch?v=c_4InHXPHvs

[Lice]

could this be an example of a core processor blowing a fuse?


OR the game freezing.


If it blew a cell it would not be on anymore... the same if you blow a cpu out... all this blowing talk makes me feel unsafe.

[Kaijufan]

I get the feeling that a core in the Cell dying will be the PS3's DRE.

[Vinny]

The Cell has 8 "synergistic processing elements" (basically a core but you know how Sony likes to sound cool:-P). Basically, if one fails another one could pick up the slack, assuming most games will only need one initially, and later make use of the rest as developers learn the hardware.

Or as musicnoteless put it:
:hot::):):):):):) = :joystick:
:headache::hot::):):):):) = :joystick:
:headache::headache::headache::headache::headache: :headache::headache::) = :joystick:

We don't really need all those SPEs but you know, incase 7 of them fail together at once, we'll be safe. Thanks Sony.:roll:

[D4rkN1ght]

. all this blowing talk makes me feel unsafe.

haha

[Ugamer_X]

This was officially announced by Sony back at E3 2005 (yes, 2005).

It's not like they've been trying to hide it.

[TimPV3]

Yeah, 7/8 working has been known since E3 2005. I'm going on memory, but I believe the reason was when you produce a processor with 8 cores, it's very likely one of the 8 cores won't work. If the PS3 only needs to use 7, it doesn't matter if one of the 8 cores doesn't work, and therefore makes it cheaper to produce the cell since they won't have to throw all the chips with only 7 of the 8 processors working.

I guess it can be compared to a carton of eggs. They sell them 12 in a carton, but a lot of the time one is broken. If you only need to use 11 eggs, it's a lot easier to find cartons with 11 perfect eggs than all 12.

Of course, I haven't read up on cell production since last year so my memory could be totally wrong on this.

[epobirs]

The Cell has 8 "synergistic processing elements" (basically a core but you know how Sony likes to sound cool:-P). Basically, if one fails another one could pick up the slack, assuming most games will only need one initially, and later make use of the rest as developers learn the hardware.

Or as musicnoteless put it:
:hot::):):):):):) = :joystick:
:headache::hot::):):):):) = :joystick:
:headache::headache::headache::headache::headache: :headache::headache::) = :joystick:

We don't really need all those SPEs but you know, incase 7 of them fail together at once, we'll be safe. Thanks Sony.:roll:

No, no, no, no, no.

The SPEs are intended to be used. All of them. Just like the three dual-thread PowerPC cores comprising the Xbox 360 CPU. Or like the vector units in the PS2's Emotin Engine processor. No decent PS2 game foregoes using all of those unit just because one might be bad.

While early PS3 games may not fully utilize the entire array of SPEs, none of them is considered a spare at the developers end. It was a major biting of the bullet for Sony and IBM to give up on having enough full spec Cells, which is 8 SPEs, for the launch. If you read the full interview linked within the article, Reeves indicate that yields for the full spec units were only 10-20%, which is disastrously bad for a mass market product. The Cells would cost almost as much as the retail price for the whole PS3.

The highly purified silicn wafer upon which the chips are built have a substantial cost. Imagine you have a chip design that needs to be delivered at $15 per unit. The wafer start costs $1,000 and would produce 100 chips if every one was good. So the silicon cost of those chips is $10. If the assorted other costs don't add more than three or four dollars the product has a decent one buck left for profit.

But if only 20 of those chips can be used after testing, the wafer cost is up to $50 a piece and the product swimming in red ink. Sony had a situation like that with the initial production of the PS2 chipset. The intended .18 micron process wasn't ready in time, so Sony had to use the existing .25 micron process they'd used for engineering samples and developer's kits and that had a much higher cost per chip even if it gave excellent yields. Which it didn't. The EE is a monster of a chip at .25 and yields sucked. Those first several hundred thousand PS2 units in Japan were big money losers but keeping the schedule and putting off proifitability for a few added months was preferable to missing the launch schedule by up to six months.

So those 8 SPE Cells are fiendishly expensive and the quantities are far short of what Sony needs to get the PS3 launched. But IBM found that the number of Cells that were good except for a single failed SPE was nearly double the number of full spec units. Standard design allows for failed units in a processor to be disabled during the test phase if the product is still valuable in that form. If the defect rate is understood in advance the chip can be design with extra function units on the assumption some will fail but no more than are needed to meet the spec.

So for the PS3 they lowered the spec. By requiring only 7 SPEs a lot more chips could be delivered at a much lower price. The price is still much higher than intended but they expect to turn things around when the Cell is produced at 65nm with expectedly much higher yields and overall cost reduction since more chips, good or bad, will be produced per wafer.

The downside of this is that the PS3 is forever after a 7 SPE Cell system. Even if the 65nm production reliably delivers a high number of functional 8 SPE Cells, PS3 developers will be forced to ignore that extra SPE because it would break compatibility with several million PS3s that will already been sold by then.

Programmers could do things that were conditional based on whether the Cell reported 7 or 8 SPEs but this would cause a lot of bad feelings among the early adopters. Not a good things, so most developers will avoid it.

[epobirs]

[ :hot: :hot: :hot: :hot: :hot: :hot: :hot: :cold: ] = :joystick:

[:hot: :hot: :hot: :cold: :hot: :hot: :hot: :cold: ] = :oldman: ->:wall: if not under warranty

could this be an example of a core processor blowing a fuse?
http://www.youtube.com/watch?v=c_4InHXPHvs

More likely just a run of the mill software crash of the sort that can happen on a perfectly good system.

The PS3's equivalent of a PC's POST will likely test for the functionality of each SPE on startup. Having a failed SPE will likely bring the system to a halt with an error message before it even looks at the disc drive.

[epobirs]

This was officially announced by Sony back at E3 2005 (yes, 2005).

It's not like they've been trying to hide it.

That isn't how I recall it. Back in 2005 Sony was still pretending the yield problems could be fixed. I don't believe they confirmed the 7 SPE configuration until the 2006 E3.

In 2005 they were also maintaining the pretense that Nvidia had created an entirely new GPU for the PS3 when in fact it is a close derivative of a PC product.

[Vinny]

No, no, no, no, no.

The SPEs are intended to be used. All of them. Just like the three dual-thread PowerPC cores comprising the Xbox 360 CPU. Or like the vector units in the PS2's Emotin Engine processor. No decent PS2 game foregoes using all of those unit just because one might be bad.

While early PS3 games may not fully utilize the entire array of SPEs, none of them is considered a spare at the developers end. It was a major biting of the bullet for Sony and IBM to give up on having enough full spec Cells, which is 8 SPEs, for the launch. If you read the full interview linked within the article, Reeves indicate that yields for the full spec units were only 10-20%, which is disastrously bad for a mass market product. The Cells would cost almost as much as the retail price for the whole PS3.

The highly purified silicn wafer upon which the chips are built have a substantial cost. Imagine you have a chip design that needs to be delivered at $15 per unit. The wafer start costs $1,000 and would produce 100 chips if every one was good. So the silicon cost of those chips is $10. If the assorted other costs don't add more than three or four dollars the product has a decent one buck left for profit.

But if only 20 of those chips can be used after testing, the wafer cost is up to $50 a piece and the product swimming in red ink. Sony had a situation like that with the initial production of the PS2 chipset. The intended .18 micron process wasn't ready in time, so Sony had to use the existing .25 micron process they'd used for engineering samples and developer's kits and that had a much higher cost per chip even if it gave excellent yields. Which it didn't. The EE is a monster of a chip at .25 and yields sucked. Those first several hundred thousand PS2 units in Japan were big money losers but keeping the schedule and putting off proifitability for a few added months was preferable to missing the launch schedule by up to six months.

So those 8 SPE Cells are fiendishly expensive and the quantities are far short of what Sony needs to get the PS3 launched. But IBM found that the number of Cells that were good except for a single failed SPE was nearly double the number of full spec units. Standard design allows for failed units in a processor to be disabled during the test phase if the product is still valuable in that form. If the defect rate is understood in advance the chip can be design with extra function units on the assumption some will fail but no more than are needed to meet the spec.

So for the PS3 they lowered the spec. By requiring only 7 SPEs a lot more chips could be delivered at a much lower price. The price is still much higher than intended but they expect to turn things around when the Cell is produced at 65nm with expectedly much higher yields and overall cost reduction since more chips, good or bad, will be produced per wafer.

The downside of this is that the PS3 is forever after a 7 SPE Cell system. Even if the 65nm production reliably delivers a high number of functional 8 SPE Cells, PS3 developers will be forced to ignore that extra SPE because it would break compatibility with several million PS3s that will already been sold by then.

Programmers could do things that were conditional based on whether the Cell reported 7 or 8 SPEs but this would cause a lot of bad feelings among the early adopters. Not a good things, so most developers will avoid it.

I wasn't being serious.:-P I could care less how the Cell works... as long as it works.

[Ugamer_X]

That isn't how I recall it. Back in 2005 Sony was still pretending the yield problems could be fixed. I don't believe they confirmed the 7 SPE configuration until the 2006 E3.

In 2005 they were also maintaining the pretense that Nvidia had created an entirely new GPU for the PS3 when in fact it is a close derivative of a PC product.

http://www.gamespot.com/news/2005/05/16/news_6124681.html

[depascal22]

Looks like I'm not buying a PS3 until at least 2008. I'll wait for all these bugs to work out and I'll be playing this humongous backlog I've accumulated in the last year.

[MadFlava]

Looks like I'm not buying a PS3 until at least 2008. I'll wait for all these bugs to work out and I'll be playing this humongous backlog I've accumulated in the last year.

Yeah, that's a good plan. The one thing that can sway me to buy a launch PS3 is if the DVD player will upconvert DVD's to 720p or 1080i. Also, if PS2 backward compatibility will also display all games in progressive scan.