Cell Hits 45nm, PS3 Price Drop Likely to Follow

Septimus writes "At this weeks ISSCC, IBM announced that the Cell CPU used in the PlayStation 3 will soon make the transition to IBM's next-gen 45nm high-k process. 'The 45nm Cell will use about 40 percent less power than its 65nm predecessor, and its die area will be reduced by 34 percent. The greatly reduced power budget will cut down on the amount of active cooling required by the console, which in turn will make it cheaper to produce and more reliable (this means fewer warrantied returns). Also affecting Sony's per-unit cost is the reduction in overall die size. A smaller die means a smaller, cheaper package; it also means that yields will be better and that each chip will cost less overall.'"

Re:Effect on cost

[scubamage]

For sony, yes. For end buyers? Nope. To sony this just means their profit margin got bigger.

Re:Since when?

[jd3nn1s]

I believe it's because the chip is smaller therefore more fit on the same size wafer.

Re:Since when?

[ThreeGigs]

Since when does going to a smaller process increase yields?

Always has.

Assume there will be 20 defects on a wafer that will render 19 large chips (out of 100) unusable. Your yield is 81%.
Same 20 defects, but affecting 20 small chips (out of 170). Now your yield is 88%, or 150 chips versus 81 chips per wafer.

The number of defect sites per wafer is generally rather constant, thus the more chips you can fit on a wafer, the better the yield.

Re:Effect on cost

[hansamurai]

Of course it will reduce the price of the Playstation 3. Why do you think when consoles are first released they're $200-$300 (last generation for example) and then five years later they're floating around $100 retail? Some of it has to do with the bottom line, but most of it has to do with the falling price of components over time due to exactly what was listed in the summary, exactly what is happening here. This one event might not directly lead to a price drop, but enough of these do.

Re:Since when?

[timster]

Power supplies also generate a lot of heat -- notice that those bricks tend to be warm under load, even through that insulation. Put them inside the laptop and you're adding a bunch of heat to a place that you want to be removing heat from. So you need bigger fans and it takes even more space. It's just unworkable.

Often can

[Sycraft-fu]

The reason is that wafer size doesn't change. I don't remember what is current, 8 inch I believe (that's the largest I've seen) but regardless. So when you reduce the size of an individual chip, you get more chips per wafer. Now unless the percentage of chips that fail increases, that means you get a better yield/wafer.

Well cost is based per wafer. It doesn't cost any more to make a wafer with 1000 small chips than it does to make one with 4 big chips. In either case it is the same size wafer, same mask, same process, etc.

Now yield could go down if a company has problems with a new process. Suppose that the old process yields 10% non-working chips per wafer. You get a new process that yields 20% more chips per wafer than the old one, however now 50% of them are non-working. That would equal a lower yield, despite the more chips per wafer.

However assuming a roughly equal failure rate, shrinking the die size will increase the yield.

Why would Sony drop the price?

[DarkTitan_X]

Sony is already losing money on the cost of production vs. the sale price of each Playstation 3 (sale of a PS3 averages around a 35% loss of profit per unit).

Simply put, they reduce the cost of production, they lose less money on each one they sell. Considering the Playstation 3 is slowly gaining market share at it's current price, they have no need to drop the price right away.

Re:Isn't the blue laser the biggest cost?

[ivan256]

That's ancient news. Since then, the blue laser shortage has ended, and Sony has gotten the costs of PS3 manufacture down to under $400 [businessweek.com].

Re:Since when?

[mikael]

The size of a defect is of a fixed size. Usually it is a particle of dust that got in the way of the optical etching process. The distribution of such defects is even across the surface of the silicon wafer, so the distribution can be modelled mathematically.
Suppose there are 20 defects across the wafer. If your chip were the size of the entire wafer, it would be guaranteed to be defective.
Try half the size of the wafer, and there would be on average 10 defects. A quarter of the wafer, 5 defects. If you have a chip that is one hundredth the size of a single wafer, then the odds are now in your favour; on average 20/100 that you will have a defect, 80/100 that you will not.

The Cell processor is etched with eight processors anyway. If one is defective, they can ignore it, otherwise if all eight are working, then they will just deactivate one.

I wonder how long it will be before they start adding more processors to the chip.

CBE Performance

[shadowofwind]

Relevance of CBE beyond PS3 of course depends in large degree on its computing performance. For the applications I've looked at, I haven't been very impressed. They say it does 204GFLOPS, but approaching that requires being able to use all multiply-add instructions, which count as two operations. (Some sources say the two operations per clock cycle per SPU is due to there being two pipelines, however, only one of the pipelines handles arithmentic operations and the other is exclusively for load, store, control, and a few shift operations.) Also, it seems to take a lot of select, shift, and shuffle instructions to make efficient use of the quadword (SIMD) instructions. With Xeon and Opteron, use of the quadword instructions seems to require far fewer other additional cycles. And this is with floats, with instruction related stalls completely eliminated on CBE through careful loop unrolling and other methods. (The quadword instructions have 6 cycle latencies.) I can only get performance comparable to 2 quad-core Xeons, which doesn't seem that good considering what is advertized, and considering the 4x difference in the peak performance specs. And CBE does much worse where double precision is necessary, with 6 cycle stalls being unaviodable on every instruction. It seems overblown. Comments?

Re:Effect on cost

[Amouth]

Imagine a macbook powered by something like this, 45nm, 8 cores, low power usage, cheap...

And nothing to run on it...

Re:Does that make for a slimmer ps3?

[Adhemar82]

No, that's just a mock-up made by some creative people at T3: http://www.t3.com/news/sony-playstation3-slim-and-lite?=35165/ [t3.com]

Re:Often can

[solarium_rider]

Actually they have been moving to larger wafers with 90nm and below. They are using 300mm wafers (about 12 inches.) I think non-submicron wafers are about 180mm in diameter for most fabs.

Re:It would be really great, IF

[mikearthur]

You can get them in IBM Blades or from a company called Mercury that will sell you a Cell BE on a PCI-E accelerator board.

Re:Effect on cost

[Fozzyuw]

To sony this just means their profit margin got bigger.

You mean their loss margin just got smaller. They're still looking forward to making a profit. [reuters.com]

Re:Absoluely not.

[Pulzar]

Not so long as the consoles continue to sell at the current price. Sony charges what they think people are willing to pay, no more and no less.

That's a very simplistic view. First, "people" is a collection of persons all willing to pay different prices. So, there's no one price at which "people" will buy, and another at which "people" won't buy.

A company selling a product will try to maximize the profits. Once the cost of production goes down, the "maximum profit" formula changes -- you will either get more profit per unit, or you will sell at the same profit but sell more... or do something in between. The new magic "maximum profit" price will almost certainly be different than before.

Die Yield Not as Important for Cell

[Doc Ruby]

A smaller die means a smaller, cheaper package; it also means that yields will be better and that each chip will cost less overall.

The redundancy of the Cell's 8 SPUs (DSP coprocessors) is the main point of the Cell's design. Defective SPUs (nearly always from dust particles in the nearly - but not quite - perfect "clean rooms" in which they're manufactured) can be tested and turned off as they roll off the assembly line. The shut down SPUs are even physically disconnected from power by hard fuses, so they don't cost any performance in operation. The perfect Cells with 8 SPUs cost the most, in high-end IBM RS/6000 workstations (and some blade servers). 7 SPUs go into PS3s. The rest of the yield, supposedly down to a single SPU (but even 0 SPUs still have a 3.2GHz PPC and superfast IO), go into HDTVs and other consumer electronics. All of the yield gets sold, instead of a fraction in older manufacturing processes.

So smaller dies don't really affect Cell yields. Smaller dies just mean smaller parts of the wafer that would get spoiled by a single defect, which is already taken care of with the redundant SPUs.

In fact, smaller dies mean multiple defects are less likely to land on a single die. Which means that more Cells would turn into low-SPU, cheaper Cells. While larger dies would concentrate multiple defects into a single dies, by landing on a single die more often, leaving more perfect Cells getting the highest prices.

45nm does mean more Cells, at any defect rate, per wafer. Which means, for the same number of defects per wafer, more dies per wafer. So there is a yield increase, but not for the same reasons as traditional ones. And of course 45nm has so many other valuable benefits, like speed, and more transistors if they keep the same die size, that the move is very valuable overall.

Re:More SPUs?

[TheRaven64]

Unless their yields have gone up a lot recently, they put all of the ones with 8 SPUs into (very expensive) blades and put the ones with only 7 working in PS3s. If they had more with 8 working, they might sell quite a few more to the scientific computing community.

Re:The Little and the Big

[Anonymous Coward]

Just FYI, the cell processor is made in Nagasaki, Japan in a fabrication plant formerly owned by Sony, recently acquired by Toshiba.

Although IBM had their hands in the R&D with Sony and Toshiba, Toshiba is responsible for all manufacturing of the Cell processor.

(I work in the semiconductor industry in Japan...)

Re:The last couple of paragraphs are the best

[GringoCroco]

I don't know the full extent of the "cycle compatibility" but IBM had previously stated some *facts* about Cell's SPEs: ALL memory accesses on the Local Store take *exactly* 6 cycles. As all applications must be throughly modified to take advantage of the SPEs (things may include vectorizing your computations, distribution of data between cores, etc.) you were inclined to optimize your app based on this memory latency.

For example, you have to process Y GB of data. You split the data in chunks of size X bytes so that while you process X bytes, in the background X bytes are transferred from main memory of wherever (transfers would be done through DMA). You switch buffers and you don't see the latency of the DMA transfers. But if the *6 cycles per local store access* rule were to be invalidated, your program may behave differently.

Re:Effect on cost

[ToasterMonkey]

A normal multithreaded application isn't going to magically run parallelized on a Cell BE. Currently they only have one general purpose CPU with a number of specialized coprocessors.

So, yes, many applications use SMP to do parallel work, but few of those do it in a way that makes sense on a Cell BE. IOW, merely running the audio subsystem of a game in a separate thread won't scale to a system with specialized coprocessors.

Do you think there are many threaded applications out there that use a model where the main logic is in one thread that farms data out to threads on other processors to crunch data in bulk? Standard OS's do not use this model, they only make use of multiple identical processors. Well, unless you count an accelerated graphics system, because a GPU is used this way. How many applications today could take advantage of having several coprocessors without a significant amount of work? Probably just a handful of experimental ones that try to offload work onto the GPU.

Re:"Price drop unlikely" does not follow.

[feepness]

Last you heard, you heard wrong. [businessweek.com]

Re:Since when?

[wannasleep]

Defectivity (i.e. the "dust problem") is just one of the yield detractors. There are many more and they get worse and worse. For instance, there are litho problems, etching problems, CMP problems, not to mention gate leakage, and a bunch of other parametric issues. So, you can not just look at defectivity. Even if you did, with a smaller feature size, small particles that could be tolerated in an older generation will now cause yield loss.

PS: the distribution you are talking about is a poisson distribution

Re:Still Can't use it for anything other than gami

[rbanffy]

Here I am, replying to AC... Slow day indeed.

Sorry, but you are wrong (as your moderation points out).

You can buy a PS3 to do numerics and the Cell inside it is an average performer. Not bad at all for under $1000.

But, if you need to upgrade (and consider your workload is heavily parallelized and optimized for the SPUs because it already runs on your PS3's SPUs) you can buy one or more IBM QS21 blades and a suitable chassis. It's obvious these new Cells will be in these blades as soon as they become available. In the blades, the Cell is not limited as it is in the PS3, there is plenty of memory for the PPUs and you can run all your SPUs at full throttle if your data and programs demand it. And, while you are at it, you can add POWER or x86 blades to the chassis as well as Linux does not run particularly fast on the Cell PPUs and you may want a fast machine to feed the Cell node.

Sorry if you wanted current supercomputer power on the cheap. The PS3 is good enough for a lot of stuff and a lot cheaper than anything that approaches its numeric performance.

Re:Effect on cost

[Moonpie Madness]

laserdisc? Ha, I bet that thing is actually pretty cool, man.

All I mean is that, unlike many blu-ray players, the PS3 does not send out multichannel sound. You need a device that can decode the optical sound. If you have a audio system that will taket the optical outpout and give you surround sound, you're good to go.

And yeah, the cheap blu-ray players are similar in this respect, but it's still a fair point for those wanting the PS3 solely as a blu-ray player. It's not as good in the audio department at similarly priced standalone players, and you need a modern audio device.

Not a huge point, perhaps, since if you really care about audio you might as well get a seperate audio system, but it's still something to note.