Fujitsu at 90 nm tells us something about 65 nm CPUs

[Panajev2001a]

Oh btw isn't this Fujitsu chip using 6.5 MB of SRAM??? How does standard SRAM compare to SONY/Toshiba's eDRAM with regards to size and transistor counts. Would 32MB of eDRAM take up the same amount of space as 6.5MB of standard SRAM??? More? Less?

CELL would be using both e-DRAM and SRAM: 128 KB of LS per APU multiplied by 32 APUs yelds 4 MB of SRAM.

I am currently not expecting anymore 32 MB of e-DRAM, but I set it in the realm of what is possible.

It might be 16+ MB of e-DRAM and 4 MB of SRAM in total for the CELL CPU in PlayStation 3.

IIRC, the 65 nm manufacturing process SCE and Toshiba presented earlier this year ( when they announced its completion ) has some of the smallest SRAM cells and DRAM cells in the industry ( even smaller than their direct competitor, in the Japanese 65n nm race, NEC ).

Toshiba is Japan's 1st semiconductor player after-all.

 

[V3]

Oh btw isn't this Fujitsu chip using 6.5 MB of SRAM??? How does standard SRAM compare to SONY/Toshiba's eDRAM with regards to size and transistor counts. Would 32MB of eDRAM take up the same amount of space as 6.5MB of standard SRAM??? More? Less?

32 MB of eDRAM should take less than 6.5 MB of SRAM.

 

[notAFanB]

systems keep pace and implement AA/shaders of the future?

one word 'speed', the shaders of the future which you are referring to are here now. just not within acceptable real time requirements (yet).

 

[Panajev2001a]

I want to know the size and wattage of that thing.

As the architect said: "concordantly, while your first question maybe the most pertinent you may or may not realize it is also the most irrelevant".



I will try to look for it

It's pretty relevant if you need to put it in a freezer to keep it from frying itself

Well, they could sell a mini freezer to go with it

Seriously these are two different chips, designed in different manufacturing technologies and clocked differently ( I already covered what I expect to be the situation regarding CELL's local clock domains ).

I am curious too and I will keep looking for data.

 

[Megadrive1988]

I am predicting

Cell-CPU: 500-900 million transistors

Cell-GPU: 500-900 million transistors


whole PS3: at least 1 Billion transistors ( two chips of at least 500m T each)



GC2 / N5 related:

I've been thinking a little bit about Fujitsu. with NEC and Hitachi, they might be able to come up with a decent Cell rival, if they had started say 1 year ago (2002-2005, is 3 years to get a CPU together)

It would not have to rival Cell in every area of sheer power, just be compeititive.... what do you think Panajev?


also,

CELL would be using both e-DRAM and SRAM: 128 KB of LS per APU multiplied by 32 APUs yelds 4 MB of SRAM.

I am currently not expecting anymore 32 MB of e-DRAM, but I set it in the realm of what is possible.

It might be 16+ MB of e-DRAM and 4 MB of SRAM in total for the CELL CPU in PlayStation 3.

At one time it was predicted that 4 PEs would share 64 MB eDRAM

I think that 16 or 32 MB eDRAM would be falling into DMGA's hands, as he is predicting only 1 or 2 PEs thus 8 or 16 APUs.

 

[chaphack]

At the end of day, i will still put faith in ATI shaders/AA/IQ/whatnot being more advance while PS3 will excel in other aspects. Tera/Giga flops sounds big and all, but the final output will not highlight THAT level of difference.

So pardon my skepticism Sonyfans, i have seen enough of SCEI systems and yes this is also ATI we are speaking about. Not only are they known for IQ/whatnot but i believe Mr Orton has a fair bit of insider info.

 

[rabidrabbit]

At the end of day, i will still put faith in ATI shaders/AA/IQ/whatnot being more advance while PS3 will excel in other aspects. Tera/Giga flops sounds big and all, but the final output will not highlight THAT level of difference.

So pardon my skepticism Sonyfans, i have seen enough of SCEI systems and yes this is also ATI we are speaking about. Not only are they known for IQ/whatnot but i believe Mr Orton has a fair bit of insider info.

well... there has only been two SCEI systems... so there's hardly yet any continuous pattern to prove your point.

 

[London Geezer]

Errrmmm... For those people worried about about how much Ps3 wil cost to manufacture:

1) Why should you care? Unless you hold quite a lot of Sony's shares, this aspect shouldn't even touch your minds. The thing will be released at a standard launch price (no less than 299 no more than 399), Sony will not release it at 500 quid a piece.

2) Have u ever heard of something called Marginal Costs? If not, then I advice u look into an Economics book and check it.

 

[DeadmeatGA]

...

2.4 GHz at 0.09µ. Weighing in at 690 million transistors

And 600 million of those are going into caches, I presume???? 690 million transistors and only dual core??? And now you were telling us how SCEI was going to fit in 4 PPCs and 36 VUs + 4 MB cache in a single cheap die?

They can be big and run pretty fast

And this is supposed to be a multi-thousand dollar workstation/server CPU, not a $399 console CPU...

 

[DeadmeatGA]

...

Why should you care?

Sony's board of directors does care. Since they are planning a massive company wide restructuring and 20K lay-offs, they might put Kutaragi Ken's name on the lay-off list as well if the cost gets out of control.

 

[London Geezer]

Re: ...

Why should you care?

Sony's board of directors does care. Since they are planning a massive company wide restructuring and 20K lay-offs, they might put Kutaragi Ken's name on the lay-off list as well if the cost gets out of control.

HA! Funny...

the question was:

why should YOU (DEADMEAT and everyone else) care?

Regarding the Ken Kutaragi comment:

 

[Panajev2001a]

Re: ...

Why should you care?

Sony's board of directors does care. Since they are planning a massive company wide restructuring and 20K lay-offs, they might put Kutaragi Ken's name on the lay-off list as well if the cost gets out of control.

The layoffs and restructuring is to reduce business ventures which are not only loosing money, but they are considered flawed and without much hope to turn profitable in the not too distant future.

The PlayStation business is not that kind of business, but more like the cash cow and one of the biggest name brands Sony has.

If to save money you only buy half the seeds you need even if there is a big chance that weather will be good for your land and what you cultivate, you will end up with half the crop you could have had.

One thing is saving money froom stopping failing business operations, one other thing is cutting investments to one of your cash cows in the next 5-7 years and the platform you plan to use to further estabilish your domincance in the Home Entertainment market.

I know that you would love to see Sony cut all the funding for the CELL project and see Ken Kutaragi fail and that is why you bring that scenario up.

The reality is that Ken Kutaragi's popularity inside Sony has risen and that Sony is indeed restructuing, but it is restructuring to be once again more technology oriented and more advanced: SCE and the technology SCE co-developed with partners like Toshiba and IBM will be one of the key pillars of the new Sony structure.

Why do you think SCE is trying to make as much cash from PlayStation 2 atr the current MSRP ? They are making the most money they can so that they can soften the initial blow they will take when PlayStation 3 is launched.

Do not worry, they have things under control.

 

[Vince]

32 MB of eDRAM should take less than 6.5 MB of SRAM.

Exactly, eDRAM is inheriently dense as. Didn't you do the area calculations once apon a time after that Toshiba/Sony announcement? Off hand, it [64MB] is well under 20mm^2, perhaps 10mm^2. But, I'm not sure because I don't know what the actual numbers are... you're good at this stuff...

At the end of day, i will still put faith in ATI shaders/AA/IQ/whatnot being more advance while PS3 will excel in other aspects. Tera/Giga flops sounds big and all, but the final output will not highlight THAT level of difference.

Lets put it this way, if Sony actually hits "teh Tera flops" in the MPU - then ATI can bend over and kiss their ass goodbye. Because nothing they do will overcome that big of a system-level differential.

PS. Megadrive, when are you going to stop talking in transistors? It really is a useless metric with all the embedded RAM they're thought to have.

 

[nondescript]

32 MB of eDRAM should take less than 6.5 MB of SRAM.

SRAM cells require 6 transistors. eDRAM (1-T DRAM, same thing) cells require 1 transistor, and no deep trench capacitor. So 39 MB of eDRAM has about as many transistors as 6.5 MB of SRAM (not counting I/O logic required to read/write to cells). In terms of technical difficulty, 32 MB of eDRAM is not a problem. Samsung already makes 1 GB memory chips right now.


DMGA, your concern for the Sony and Ken Kutaragi is very touching, but I don't think they need you to worry for them. But I'm glad you show such concern about CELL's future.

 

[nonamer]

Heat shouldn't be that big of a problem, for reasons discussed in the Clearspeed thread. Besides, no one is better suited to solve that problem than Sony, which has plenty of experience in cooling from their studio audio equipment, and their electromechanical engineering is second to none.

Heat will be the problem, not even remotely close to not being "that big of a problem." Please remember that static heat(leakage) is ~ 50% of the heat in 90nm and a lot more in 65nm, with or without SOI or whatever. This isn't some small co-processor here, and even with the most advanced processes (see Intel), it will still translate into massive heat/cooling problems for Cell/PS3.

At the end of day, i will still put faith in ATI shaders/AA/IQ/whatnot being more advance while PS3 will excel in other aspects. Tera/Giga flops sounds big and all, but the final output will not highlight THAT level of difference.

No offense but I have little faith that ATI can deliever either. The R500 (the supposed GPU for the XB2) is not even a .09u part, looks to be like a .11u part or even a .13u if Dave's theory on ATI process upgrading is correct. Even with major upgrades it's not to be all that impressive given the very lowly origins. Hopefully the R500 thing is more of a rumor than a fact. Probably I would guess that the PS3 will win the performance wars by default from simply having so much money spent on it, no matter how inefficient it is or how much it falls short.

Exactly, eDRAM is inheriently dense as shit. Didn't you do the area calculations once apon a time after that Toshiba/Sony announcement? Off hand, it [64MB] is well under 20mm^2, perhaps 10mm^2. But, I'm not sure because I don't know what the actual numbers are... you're good at this stuff...

No way is eDRAM that dense. You need stacked capacitors DRAM for that kind of density, but eDRAM needs trenched capacitance AFAIK. I'm guessing 60mm^2 at least, probably more. That may or may not be a problem depending on how crowded the die is.

SRAM cells require 6 transistors. eDRAM (1-T DRAM, same thing) cells require 1 transistor, and no deep trench capacitor. So 39 MB of eDRAM has about as many transistors as 6.5 MB of SRAM (not counting I/O logic required to read/write to cells). In terms of technical difficulty, 32 MB of eDRAM is not a probe blem. Samsung already makes 1 GB memory chips right now.

This is not entirely accurate. First of all SRAM does not require 6 transistors, 4 can do in some cases. Also SRAM seems to denser on a per transistor basis than eDRAM. It may have 6 transistors, but I think it's more like only 4 times bigger in most cases. You can also have extra dense SRAM like in Intel's case, now that's some dense SRAM (like 2/3s the density of eDRAM in some cases, possibly even within half of Toshiba's extra dense form of eDRAM).

 

[V3]

This is not entirely accurate. First of all SRAM does not require 6 transistors, 4 can do in some cases. Also SRAM seems to denser on a per transistor basis than eDRAM. It may have 6 transistors, but I think it's more like only 4 times bigger in most cases. You can also have extra dense SRAM like in Intel's case, now that's some dense SRAM (like 2/3s the density of eDRAM in some cases, possibly even within half of Toshiba's extra dense form of eDRAM).

eDRAM should be more dense than SRAM. Come on we got 4 MB of eDRAM on a $300 mass market consumer product few years ago.

Also PSP CPU is planned to have 12 MB of eDRAM, next year, on a single chip solution too.

 

[jvd]

No offense but I have little faith that ATI can deliever either. The R500 (the supposed GPU for the XB2) is not even a .09u part, looks to be like a .11u part or even a .13u if Dave's theory on ATI process upgrading is correct. Even with major upgrades it's not to be all that impressive given the very lowly origins. Hopefully the R500 thing is more of a rumor than a fact. Probably I would guess that the PS3 will win the performance wars by default from simply having so much money spent on it, no matter how inefficient it is or how much it falls short.

Well lets see . Hyper z , thier smooth vision. Thier aa . Thier pixel shaders. Their extensive knowledge in building gpus. Compare to sony that made the gs . I would put my money on sony being a bit faster but ati having the image quality .

Not only that but an r500 should be more than on par with what the cell chip can actually put out. Not only that but they can just put two r500s in the xbox 2 if need be . Even 2 on one die. They also have experiance with on die ram. So that too can bring up the speed of the r500. Lots of things they can do. Sony isn't the only one that can spout off big numbers .

 

[olivier]

sparc processor are not releted to console. They are used in high end server. Fujitsu and Sun microsystem are using this architecture since a long long time. They are not made for gaming but for stability and io speed. They are very fast in float calculation (compare to a p4) but slower on integer calculation. We have some sparc server @work. Some old sparc2 (damn slow) and some dual 1ghz sparcIII. The dual rig is fast but cost a lot ( over 20000$). Anyway you cant compare server cpu versus console cpu ... it's like comparing xbox cpu (or gamecube) versus the itanium. You wont run game on the itanium and you wont run large database or network server on the xbox.

 

[olivier]

also im pretty sure they can affort lower yield in server market since their server cost a lot .. a lot and a lot. For the console they cant affort low yield since they sell the console with a price lower than what it cost to produce.

 

[Josiah]

No offense but I have little faith that ATI can deliever either. The R500 (the supposed GPU for the XB2) is not even a .09u part, looks to be like a .11u part or even a .13u if Dave's theory on ATI process upgrading is correct. Even with major upgrades it's not to be all that impressive given the very lowly origins.

you remind me of the people who judge hardware based on how many "bits" it has

consider this:
R350 (Radeon 9800): .15u
NV35 (GeforceFX 5900): .13u
now which one is better? NV35? nope.

more examples: the Radeon 9600 has about 1/2 the transistor count of the 5900, yet in many circumstances it is faster. Geforce 4 has about 1/3 the transistor count, same deal.

I'm not making any claims that XB2 will be better than PS3 or anything like that. but to say R500 is "not to be all that impressive" is nonsense. it will be ridiculously amazing compared to any PC or console we currently have. will it be better than PS3? who knows, only a fool would compare two things that don't exist.