So there's no way PS3 CPU is getting more than one PE~Cell?

[ysoya]

"4. SCE will use shrunken version of cells for BE. (half-sized SRAM, single precision, optimized logic, ... )"

no, SPE use only double precision floating multiply for FDIV and SQRT with newton-raphson metod
for parallell with single floating multiply
double precision circuit only 64 bit simple

Thanks for good info.
Hmm.. could I remove 'single precision' from my assumption then?

 

[Mythos]

They are talking about intel. Does son'ys fabs have this second generation strained silicon process for 65nm ?

Does sony's fabs have low k ?

Buoya...

"In a release issued late Monday, the companies described some of the details of the process, including the development of a high-performance transistor with a 30-nm gate length.

Fabricated with 193-nm lithography tools and phase-shift photomasks, the transistor is said to have switching speeds of 0.72-ps for NMOSFET and 1.41-ps for PMOSFET at 0.85-Volt (Ioff=100nA/um).

The transistor makes use of a nitrogen concentration plasma nitrided, oxide-gate dielectrics to suppress gate leakage current. This optimization reduces leakage current approximately 50 times more efficiently than conventional silicon dioxide film and allows formation of an oxide with an effective thickness of only 1-nm.

To reduce wiring propagation delay and power dissipation, a low-k dielectric material is adopted. The target effective dielectric constant of the interlayer dielectric is around 2.7."

Source: eetimes.com Dec. 2, 2002.

 

[rabidrabbit]

Cell is only 1 Gigaflop (or is it 9 GF???) ... anyway, the ISSCC is wrong claiming it's 256GF!!!

 

[zidane1strife]

Btw, Megadrive, do you have a link to that 500 mil CPU claim in 1999? Thanks in advance...

Well, I remember seeing something about that in an old sciam magazine. Back there it was claimed to be a .1micron 500m transistor cpu(for 2005).

If you want to be depressed, consider that ISSCC Cell is only as complex as a high end GPU. Array processors (GPUs) will be the first single chip to hit a programmable TFLOP. The first 200+ GFLOP GPU come off the production line recently, we are still slightly ahead of Moores Law in GPU land, so mid 2006 should hit 500 GFLOP, with a TFLOP in 2007

That's strange, so gpus can effectively scale their performance in a consumer level product when using 65nm and below but the cell architecture can't?

 

[Titanio]

An 8-SPU 1PE CPU in PS3 will be fine...

 

[DeanoC]

That's strange, so gpus can effectively scale their performance of a consumer level product when using 65nm and below but the cell architecture can't?

Yes! Because they are array processors, with all the limitations that come with that.

GPU trade general processing capability for the ability to scale speed linearly with processing units. As long as the incoming data set is greater than the width of the array of processors than you can just add a couple more to execute more data elements faster.

You can't do that in architectures that imply scalar execution like CPUs.

Where this idea that a GPU is somehow just a normal CPU is beyond me, they may be able to trade FLOPS counts etc. but they are as different as chalk and cheese. Of course GPUs suck at lots of normal jobs, branching on a GPU can be much slower than on a CPU for example.

 

[zidane1strife]

An 8-SPU 1PE CPU in PS3 will be fine...

It's very good, but more is often better (Besides if we got BE, it'd cause such a ruckus, can you imagine say for example... the teamxbox boards? slashdot? the F@nb0ys? the guys over at MS? That'd be priceless)

ed

 

[Titanio]

An 8-SPU 1PE CPU in PS3 will be fine...

It's very good, but more is often better (Besides if we got BE, it'd cause such a ruckus, can you imagine say for example... the teamxbox boards? )

I'm sure this is causing a ruckus as is Of course, more would be nice, but 1PE is your safest bet, and still a beast.

 

[Megadrive1988]

Megadrive

You're mistake was reading something written by Dean Takahashi

It's interesting you mention this...the first time I came across CELL was some article on another site by an ex Naughty Dog dude (Eagle guy, sorry can't remember his name) and he did a summary of the patents and claimed the PS3 was 1.5 TFLOPS! ...I think that set the tone for PS3!

Btw, Megadrive, do you have a link to that 500 mil CPU claim in 1999? Thanks in advance...

yeah,

http://groups-beta.google.com/group...laystation2/msg/770ef1370f493b4b?dmode=source

Emotion Engine 2 is slated to appear in 2002 and will contain some 50
million transistors. Its successor, known by the ambiguous moniker Emotion Engine 3, will sport half a billion transistors shoehorned onto the die by a 0.1 micron process.

originally (i think) from here:
http://www.theregister.co.uk/2000/03/10/playstation_3_to_ship/


however, i did not find what i thought i remember being out there on the web. I thought EEtimes reported 500 million transistors for EE3 as well. but apparently not. they did say Emotion Engine 2 would have 40 million transisors. the best I could find, transistor-figure wise for *EE3* from *EEtimes*, is 100 million transistors (DOH!)

http://www.eet.com/story/OEG20000602S0064

Emotion Engine 2, with 40 million transistors, is scheduled to be developed in 2002. A third version of the chip will feature 100 million transistors, with a launch date of 2005.

 

[one]

Btw, Megadrive, do you have a link to that 500 mil CPU claim in 1999? Thanks in advance...

Well, I remember seeing something about that in an old sciam magazine. Back there it was claimed to be a .1micron 500m transistor cpu(for 2005).

From Microprocessor Forum, Oct 1999. Probably you can find other pictures from this presentation which includes the process technology.

 

[PC-Engine]

It's very good, but more is often better

From a business POV?

 

[V3]

It's very good, but more is often better

From a business POV?

More profit is better, from business POV

 

[PC-Engine]

It's very good, but more is often better

From a business POV?

More profit is better, from business POV

Yes as opposed to more losses...

 

[pc999]

Anyways they brooke Moores Law, this thing is like (for what they say) a 25 cores P4.

No they didn't break Moores Law, its spot on where Moore predicted.

First, thanks for the correction and info, I thought that P4 had a lot less transistors.

Second, can anyone tell in which kind of things will Cell be a (at least relatilvely) bad performer.

 

[Megadrive1988]

thanks, One


I either have not seen this slide, or haven't seen it in a long, long time.

 

[zidane1strife]

It's very good, but more is often better

From a business POV?

It'd depend on what'd be achievable with it, and what effect it'd have. Is it substantial in terms of subjective appreciation by the masses? Is it not?

If subjectively appreciable, is it best for them to take a loss and launch such a product or is it better to launch a less-ambitious and more profitable product but with less appreciable difference with regards to competing products?

It depends on how big a loss they'd get, and what the consumer effect would be.

I would tend to agree with the more profitable product as being more apt, but come on now that wouldn't be very bold now would it? and history often favors the bold...

Besides the same philosophy applies to the psp, why take a loss and launch such an ambitious product? But so it was.

In any case it's understandable if they choose a safer approach, if they don't put their neck on the line.

 

[PC-Engine]

Whether the BE contains 1 PE or 2 PEs, PS3 would still be losing money. With 2 PEs it would be losing orders of magnitude more than a single PE.

You say more is often better, but later your explaination doesn't support that hypothesis.

 

[iapetus]

Whether the BE contains 1 PE or 2 PEs, PS3 would still be losing money. With 2 PEs it would be losing orders of magnitude more than a single PE.

Exaggerate much? Or is there a genuine reason that a second PE would increase the price of PS3 by a factor of over 100?

Of course it would bump up the price, but not by that much.

 

[London Geezer]

Whether the BE contains 1 PE or 2 PEs, PS3 would still be losing money. With 2 PEs it would be losing orders of magnitude more than a single PE.

You say more is often better, but later your explaination doesn't support that hypothesis.

Not an order of magnitude. Only the cost of the BE will be doubled. Which is a lot, but not an order of magnitude.

 

[iapetus]

Not an order of magnitude. Only the cost of the BE will be doubled. Which is a lot, but not an order of magnitude.

Assuming the multiple units would be on a single chip, the price would more than double. Should still be within an order of magnitude for the chip itself, though, let alone the full system.