[iceberg187]

New 20GB HDMI 360: 2/3 power consumption

Impress.jp disassemabled the new 360 SKU(20 GB w/HDMI,Forza 2& Viva Pinatas included) to find what is new on the HDMI 360.

http://pc.watch.impress.co.jp/docs/2...01/xbox360.htm
http://pc.watch.impress.co.jp/docs/2...02/xbox360.htm

90nm 65nm
Dashboard 153W 103W
Game Demo 155W 104W
Game Running 174W 122W
Running Temp. 118 109 (F)*

[AlStrong]

Nice find.

50W reduction all-around seems nice... But the thermal reduction just seems to give further impetus to get that GPU on the 65nm process.

What's the source of the temperature measurements?

edit: Ah I see, you converted the rightful Celsius to the inferior Fahrenheit.
->48C to 43C

edit2: FWIW, the console was manufactured on Oct. 11 and uses the Hitachi drive.

[Rangers]

Well, they include pics of the dies in there and once again, to the naked eye they dont look substantially smaller enough to be 65nm, but I guess they are.

Now why dont they do a 40GB PS3? (For the die shots, mainly)

Also, a xbox.com forumgoer had already discovered similar reduced power draw findings a few weeks ago with the Halo 3 360, but it's nice to have a more official source.

Old Llamma.com Xcpu pic:



New watch.ipmress Xcpu pic:



Looking at the leftmost row of five resistors, assuming they are the same size/spacing on both packages, you see the alledged 65nm CPU top edge lines up with about the bottom/mid 2nd resistor, and the bottom edge is just past the 4th resistor. On the non-65nm, the top edge is slightly past the top of the first resistor, and the bottom edge is down to the fifth resistor. There's definitly a change in size there, even if it's maybe not what you'd want from 65nm. Given the power consumption figures, I'd say it's 65nm. Of course this assumes the CPU package is exactly the same size, which I do not know about. Ideally somebody would measure the dies, to get truly "hard" data.

The GPU die, we've had another thread with this, is also somewhat different/possibly more compact, although seemingly not to the degree of the CPU. For certain, the shape of the EDRAM module has changed, for example. On the old die it is rectangular, while on the new it is pretty much a square.

[snakejoe]

another comparison pics, both from watchimpress.

original XCPU


New XCPU

[RobertR1]

Why is Xenos taking so long to get to 65nm. Much newer and more powerful hardware in PC land is coming out at 65, surely Xenos can't be that far behind? Major mobo and form factor revision maybe?

[Rangers]

Originally Posted by RobertR1

Why is Xenos taking so long to get to 65nm. Much newer and more powerful hardware in PC land is coming out at 65, surely Xenos can't be that far behind? Major mobo and form factor revision maybe?

Who knows..one rumor was MS was using all it's engineering resources on solving RROD so delayed it until spring.

I also wonder if EDRAM has anything to do with it. It's not just one piece to go to 65nm, like RSX. I would also like to know if RSX is 65nm. If it's not, then perhaps it makes Xenos 90nm look better by comparison.

It's really not too big a deal I guess. Holiday season is almost over anyway.

[Arwin]

RSX is 65nm, at least according to C't apparently (still have to read the magazine)

[Shifty Geezer]

Originally Posted by Rangers

Who knows..one rumor was MS was using all it's engineering resources on solving RROD so delayed it until spring.

65nm would have been the ideal solution!

[Arnold Beckenbauer]

Originally Posted by Arwin

RSX is 65nm, at least according to C't apparently (still have to read the magazine)

No, they thought, it is 65 nm, but:
http://www.heise.de/newsticker/meldung/98380
It's still 90 nm.

[Mintmaster]

Originally Posted by Arnold Beckenbauer

No, they thought, it is 65 nm, but:
http://www.heise.de/newsticker/meldung/98380
It's still 90 nm.

Interesting. So now we have both MS and Sony delaying 65nm migration compared to other chip makers.

Maybe 65 nm wafers are still priced more per transistor than 90nm at TSMC. For ATI and NVidia, their parts don't have the volume to make it worth taping out and debugging two designs, so they choose 65nm for long term price effectiveness rather than immediate. MS and Sony have both designs anyway so they will switch as soon as it's profitable to do so.

[Dr. Nick]

Originally Posted by RobertR1

Why is Xenos taking so long to get to 65nm. Much newer and more powerful hardware in PC land is coming out at 65, surely Xenos can't be that far behind? Major mobo and form factor revision maybe?

Isn't there a chance that they will skip 65nm and go for whatever ATI is going for with their next Directx 10.x part?

[Rangers]

Originally Posted by Mintmaster

Interesting. So now we have both MS and Sony delaying 65nm migration compared to other chip makers.

Maybe 65 nm wafers are still priced more per transistor than 90nm at TSMC. For ATI and NVidia, their parts don't have the volume to make it worth taping out and debugging two designs, so they choose 65nm for long term price effectiveness rather than immediate. MS and Sony have both designs anyway so they will switch as soon as it's profitable to do so.

Wait..aren't Cell (probably) and Xcpu 65nm?

RSX and Xenos are lagging, possibly?

Or do you think none of the chips in either console is 65nm yet?

And we still have yet to see 40GB PS3 die shots. And the new 360 die Xcpu die shots are inconclusive, but could be 65nm.

[djskribbles]

wasn't it confirmed that at least Cell is 65nm? the 40GB PS3 uses ~100w less than the 60GB, its lighter, it has fewer vent holes and reports said its quieter. they're also selling it for $100 less.

[Rangers]

Originally Posted by djskribbles

wasn't it confirmed that at least Cell is 65nm? the 40GB PS3 uses ~100w less than the 60GB, its lighter, it has fewer vent holes and reports said its quieter. they're also selling it for $100 less.

None of that is really confirmation though. Other factors could change the power usage, such as streamlining the motherboard.

Die shots are the best confirmation. Although again, in the case of new 360, they're almost inconclusive. Except to say that Xenos is surely not 65nm.

[rekator]

I have an question 50watts less for the 65nm Xenon than a 90nm Xenom (around 100watts) it's really plausible?
I don't remember we see a 50% less consumption for other 65nm processor…
Same for the CELL 100 watts less for it alone…

[dskneo]

the new 40gb does not consume 100w less..... the last measurement i read was ~150w against 200w. 50w difference

[Carl B]

Originally Posted by rekator

I have an question 50watts less for the 65nm Xenon than a 90nm Xenom (around 100watts) it's really plausible?
I don't remember we see a 50% less consumption for other 65nm processor…
Same for the CELL 100 watts less for it alone…

Neither of those CPUs was running at 100 watts burn at any stage of this game, that is almost certain. So the drops come from die revisions, yes, but also from the use of increasingly 'power-light' componentry throughout the system.

As for the 360, I'm still not locked on it having migrated to 65nm on the CPU, if only because it doesn't look massively shrunk. But then again, who knows - initial runs on a new node aren't always that much smaller, depending - but the 80nm theory is still out there as well. Anyway I just wish those people that took the time to open up that Falcon unit before, took the extra five minutes to actually measure the dies as well.

And as for C't and the 40GB PS3, if a big tech publication like that was going to run the tests they did, they should have popped those heat spreaders off to get some real info on those chips, rather than making an assumption as to the shrink. I appreciate the efforts by both teams, but it's so "half-way" in some respects.

[rekator]

Yes, Carl B Xenon it's not 100w, but I indicated this number for showing that the 50w less It's not only due to Xenon at 65 nm (65nm generally reduce watts by around 25%?), so like the PS3 40go the reduced watts come also from other composant…
If GS consume 20 watts may be the reduce watts for PS3 40GO come from 65nm CELL and lost of GS… but we need more precis mesure some more 80watts less are surfing…
But for the 360… 50 watts less for only the 65nm Xenon, I don't think so…
If yes, AMD have to got this process!
And the Dual-G5 up to 3Ghz are possible!

[djskribbles]

Originally Posted by dskneo

the new 40gb does not consume 100w less..... the last measurement i read was ~150w against 200w. 50w difference

someone at avsforum tested and it was definitely more than 50w difference on average. it would be impossible for me to find the thread again though...

[RobertR1]

Was trying to hunt out a Flacon Elite but there are not out yet. Still, sold my Premium to a freind and got a regular Elite. It does run cooler and quieter than my premium console. Got lucky and it came with the BenQ drive so even when the disc is spining the sounds stays good. Since my set processes better over HDMI, that was a nice plus also Over the 65nm hunting now.

The black look on the elites is very nice. Makes the console apprea quite slim and sleek. And I like that it's not a glossy black.

[AlStrong]

Anandtech busted out the ruler:

http://www.anandtech.com/gadgets/sho...spx?i=3152&p=5

CPU: 176mm^2 -> 133mm^2
GPU: 182mm^2 -> 156mm^2
eDRAM: 80mm^2 -> 68mm^2

[Rangers]

Now Do It For Ps3 Anand.

[Carl B]

Originally Posted by AlStrong

Anandtech busted out the ruler:

http://www.anandtech.com/gadgets/sho...spx?i=3152&p=5

CPU: 176mm^2 -> 133mm^2
GPU: 182mm^2 -> 156mm^2
eDRAM: 80mm^2 -> 68mm^2

Yeah the GPU and eDRAM seem to be 80nm (although even for 80nm the die shrink wouldn't be too large), and the CPU is probably 65nm... although again with a smaller shrink than one might otherwise expect. But the power reduction figures are certainly appreciable themselves though.

[AlStrong]

Originally Posted by Carl B

Yeah the GPU and eDRAM seem to be 80nm (although even for 80nm the die shrink wouldn't be too large), and the CPU is probably 65nm... although again with a smaller shrink than one might otherwise expect. But the power reduction figures are certainly appreciable themselves though.

The 80nm reduction seems pretty good, which should theoretically give up to 20% reduction. I wonder what sort of issues they had with the 65nm Xenon. An updated shot of the chip layouts would be neat.

I'm sort of kicking myself for getting that 360 last week (doesn't have the newer chips, it's saving grace is the BenQ drive), but I couldn't resist getting it in the US while some family was there. Saved myself ~$65 CAN. Oh well... I guess I'll settle with a room heater for the coming winter months.