Sony 65nm SOI Cell Broadband Engine Details

[D_o_S]

ISSCC (International Solid-State Circuits Conference) seems to be packed with interesting stuff this year. Earlier today, we reported how Samsung demonstrated 4 GHz GDDR4 memory modules at the conference, and now, VR-Zone reports that the Cell Broadband Engine is in for a die-shrink. The 'new' 65nm chip will replace the current 90nm units that can be found in PS3s. It is smaller as compared to the 90nm version, thus it is cheaper to produce and emits less heat. Also, the SRAM on the 65nm Cell B.E is improved. Sony engineers managed to run the 65nm Cell B.E. chip at 6GHz with just 1.3V. Sony is expected to continue using Cell B.E. architecture into their next generation PlayStation 4 and there is a possibility that IBM's eDRAM technology will be employed.



View at TechPowerUp Main Site

 

[C.Ash]

Cell overclocked better than the Core 2 Duo's.

 

[Completely Bonkers]

Dont want to be pedantic, but

It is smaller as compared to the 90nm version, thus it is cheaper to produce and emits less heat

is incorrect. Size doesnt change power or heat output.

It emits less heat due to

Sony engineers managed to run the 65nm Cell B.E. chip at 6GHz with just 1.3V

A decrease in Watts does change heat output! And we presume that with a lower V there is a lower W and hence less heat.

 

[WarEagleAU]

This is cool. If they could put these in the PS3's they are making now, price should go down 200 bucks.

 

[KennyT772]

your kidding right? if anything the price to produce will drop about $25. until bluray players go down and the console is put into larger production numbers there will be no price drops. hell the ps2 stayed at $300 until almost a year later when the price dropped to $250 then $200.

 

[newtekie1]

I hope this means a price drop is coming soon from Sony, we all know the PS3 desperately needs one.

Dont want to be pedantic, but is incorrect. Size doesnt change power or heat output.

It emits less heat due to

A decrease in Watts does change heat output! And we presume that with a lower V there is a lower W and hence less heat.

Well the side effect of a die shrink is that less power is needed to maintain clock speeds, and hence less heat is produced at the same clock speeds. So a die shrink=less heat at the same performance level.

 

[Completely Bonkers]

@-1,

Incorrect. A die shrink does not decrease power UNLESS it operates at lower volts. Typically a die shrink also reduces resistance, so if there is the SAME voltage, power and heat will actually be HIGHER. Go back to your physics formulas, and think Pentium Prescott if you need an example.

The new cell is lower power ONLY because they are running at a lower voltage. The die shrink ALLOWS them to run at the lower voltage (within design/manufacture/silicon specs).

 

[newtekie1]

A die shrink allows them to run the same speeds and same performance at a lower voltage and hence allows the same performance with less heat. A die shrink allows them to lower the heat output, not every processor manufacture chooses to use that die shrink in that way, most use it to up clock speeds and keep the same heat output.

However, with consoles they don't have to increase the performance, so the heat output decreases.

If you want an example of that, look at all the AMD chips that went through a die shrink and came out at the same clock speeds, using less voltage to stay stable, and putting out less heat.

My point was not incorrect, it was 100% correct. A die shrink means less voltage is needed to maintain performance, which means less power, which means less heat. You are arguing about nothing. A die shrink CAN mean less heat, it just depends on how that die shrink is use. In this case it will certainly mean the Cell processors in the PS3s will put out less heat.

 

[Completely Bonkers]

blah blah blah blah blah. See my OP. That's what I stated. Less heat ONLY with less watts. Less watts from lower volts. It is incorrect to say, de facto, shrink=less heat.

This forum is about sharing knowledge or correcting misunderstandings. Stop trying to defend your position. I accept you speed read my OP and misunderstood. Thats OK.

 

[KennyT772]

Bonkers....Die shrinks are for 2 reasons. Increase clock speeds on same voltage and heat output, or lower voltage and heat output and maintain clock speeds.

Die shrinks are generally understood to either increase performance or lower heat output.

 

[Completely Bonkers]

@kenny

You are also falling into this futile discussion. Nobody is arguing that USUALLY die shrinks result in less power. The scientific point is this: the shrink itself DOESNT decrease power. In fact, at the SAME VOLTS a die shrink would INCREASE power. The FACT is that decreasing voltage allows for LESS POWER... and LESS POWER = LESS HEAT.

YAWN. Will people please start reading at the BEGINNING of the thread rather than wasting everyones time by jumping in on the last comment and getting the wrong end of the stick.

 

[newtekie1]

blah blah blah blah blah. See my OP. That's what I stated. Less heat ONLY with less watts. Less watts from lower volts. It is incorrect to say, de facto, shrink=less heat.

This forum is about sharing knowledge or correcting misunderstandings. Stop trying to defend your position. I accept you speed read my OP and misunderstood. Thats OK.

Except I didn't speed read your post, don't try to insult me when things aren't going your way, it doesn't fly too well around here.

Die shrink=Less heat. Let me explain it again:

Die shrink=Less Voltage Required to maintain stablity

Less Voltage Required to maintain stablity=Same clock speeds with lower voltage

Same clock speed with lower voltage=Less power for the same performance

Less power for the same performance=Less heat for the same performance

Less heat for the same performance=Less heat

Therefore we can use the transitive property of logic to say that Die shrink=Less heat.

 

[Completely Bonkers]

I did not insult you. If you think I did, quote the expression. You'll be looking a long time.

Transistor gate leakage associated with the ever-thinning silicon dioxide gate dielectric is recognized by the industry as one of the most formidable technical challenges facing Moore's Law. To solve this critical issue, Intel replaced the silicon dioxide with a thicker hafnium-based high-k material in the gate dielectric, reducing leakage by more than 10 times compared to the silicon dioxide used for more than four decades

Reducing die size increases gate leakage which increases power consumption and heat output. See Pentium Prescott for a well known case of this effect.

There needs to be a corresponding reduction in voltage to offset current leakage. And while this can sometimes be done with the same "technology" usually a new fabrication process is required, and sometimes new dielectric.

The cell uses less power BECAUSE they reduced voltage NOT BECAUSE they shrunk it.

Power is a function of Watts is a function of V x A. That is the transitive relationship.

Power is not a function of die size.

There is a CORRELATION between chips we see in the market... their die size... and the power. But there is not "a functional relationship".

 

[Completely Bonkers]

... unless you think that the weather is a function of the number of short skirts worn.

 

[newtekie1]

I did not insult you. If you think I did, quote the expression. You'll be looking a long time.

Yes you did.

Stop trying to defend your position. I accept you speed read my OP and misunderstood. Thats OK.

That is down right insulting. Don't speak to me like a child, and definitely don't claim victory then try to rub it in my face, especially when you aren't actually right. The worst insult of all is when you insult someone then lie right to their face and say you didn't. I am done with you. Enjoy my ignore list.

 

[magibeg]

hey now, no fighting you 2, it REALLY doesn't matter. Just browse through wikipedia and gain knowledge (although wikipedia might be banned in schools for incorrect knowledge).

What it all comes down to is smaller gates = less insulation = more leakage

However because everything is smaller it uses less voltage so things even out a little nicer. In any event as far as i'm concerned die shrinks are a good thing usually

 

[Completely Bonkers]

There's an interesting psychological profile to bellicose people that try to escalate an issue or misunderstanding. It's related to their need to find an avenue to vent their frustration.

However, it's pretty unusual to find that character type here, on an enthusiasts tech forum, frequented by the inquisitive, experimental, knowledge searching/sharing people. While people OFTEN get the wrong idea... it shouldn't downtown like it did in the last couple of posts.

I hope newtekie calms down. The phrases I read are typical of teenage insults probably he had directed at him in the past... which hurt... Expressions like "don't speak to me like a child" and "lie right in their face" are women's tongue. He probably got a roasting once from his sister or mum... (I know I have) and the expressions are being reused with the same emotional intent.

I've seen newtekie in other threads. He's usually pretty sensible. I think he's just having a bad day. I've not taken offence. He shouldn't either.

newtekie: if something I wrote offended you, then please understand it was not mean to be an insult. Perhaps just a poor choice of words. Sorry.

 

[erocker]

Does it occur to you that there IS heat loss due to a smaller heat source.

 

[Completely Bonkers]

LOL

 

[Benpi]

PS3's toast.

 

[Zubasa]

Does it occur to you that there IS heat loss due to a smaller heat source.

Yes if the heat source is a star