[AlStrong]

Originally Posted by MrFox

I thought there was a roadmap for gddr5 to be fully differential, doubling the bandwidth per pin? Is this a reality yet?

Not much news about it really. No one seems to be jumping on board yet at least.

[Gubbi]

Originally Posted by AlStrong

Not much news about it really. No one seems to be jumping on board yet at least.

I used to be a firm supporter of differential signalling, but not so much anymore.

Single ended signalling require lots of pins on the DRAM package for ground and power planes to manage noise. However differential signalling requires two traces to be routed on the PCB per signal pin. As long as single ended signalling is within half the per pin bandwidth of differential signalling, the only advantage seems to be power usage.

Cheers

[AlStrong]

On the other hand, we've got issues with memory controllers supporting 6.0Gb/s GDDR5 and above, so what alternative is there but to go with D.S.?

[Gubbi]

Originally Posted by AlStrong

On the other hand, we've got issues with memory controllers supporting 6.0Gb/s GDDR5 and above, so what alternative is there but to go with D.S.?

More on-die RAM ?

The lower power consumption per bit transmitted might be important in mobile applications. As these devices grow in capability and capacity we might see a migration to diff. signalling, and then from mobile devices to high end solutions (after all, everything is power limited these days).

Cheers

[MrFox]

I don't think it's a zero sum. SE won't get much more than 6Gbps per line, probably ever. DS can do 20 Gbps per pair. There's also a much better signal integrity, less noise, simpler board layout, the PCB might cost less to design, test and mass produce.

Sony and IBM are working hard on optical interconnects between chips, but I would guess that's for PS5 or Xbox1440.

[Entropy]

Originally Posted by Gubbi

More on-die RAM ?

The lower power consumption per bit transmitted might be important in mobile applications. As these devices grow in capability and capacity we might see a migration to diff. signalling, and then from mobile devices to high end solutions (after all, everything is power limited these days).
Cheers

The lower power consumption per bit transmitted is important in mobile application, the question is how important it is versus other concerns. SPMT (link to consortium), (link to brief description) may be one of the contenders for upcoming mobile solutions. Or not. The right names are on the promoters list, but it has been awfully quiet.

[tunafish]

Originally Posted by hoho

Sure but going from 8 (closer to 6 actually usable) to 16 is significantly bigger deal than going from 16 to 32.

Absolutely. Although, with arm only 13 of those were usable (the PC was a GPR). ARMv8 uses r31 for SP and as a null register (depending on context), r30 for link register, and r29 for frame pointer. The rest of them are GPRs, with a really nice split in the calling convention they are promoting -- 9 for args to/from functions, 10 for callee-save, 9 temporaries, and a single one for persistent platform state (probably heap allocation pointer in most languages.)

Generally, I don't expect programs to get that much direct boost from the changes, but building faster processors should get easier.

x86 also doubled the amount of SIMD registers, how about ARMv8?

They used to have 32 64-bit registers, also addressable as 16 128-bit regs. ARMv8 has 32 128-bit registers, and when used as 64-bit ones, they are all packed into the lower-order bits of the respective regs.

There's one bit of braindamage left in the calling convention -- some of the floating point regs are marked as callee-save, but the callee is only expected to save and restore the lowest 64 bits. This means that if you are using SIMD, there effectively are no callee-save regs.

[steampoweredgod]

Aren't fabs moving to 450mm wafers reducing costs for chips with larger sizes? Wouldn't that affect the initial chip sizes of nextgen consoles as they're likely to benefit from such wafers down the road?

[AlphaWolf]

How would 450mm wafer production disproportionately benefit larger chips?

[steampoweredgod]

Originally Posted by AlphaWolf

How would 450mm wafer production disproportionately benefit larger chips?

It benefits all chip sizes but reduces cost/die-area vs 300mm, iirc, so that even bigger chips should become viable/cost-competitive.

[AlStrong]

Sure, the larger the wafer, the less wastage for larger chips, but at what cost? There's a simple reason why it's taking so long and that it's simply not a trivial thing to manufacture. The whole issue surrounding 450mm adoption is the trade-off in the high difficulty in producing that large an ingot.

Ultimately, it just means it won't be cheap, and it'll be a very long time before the fabs pass on such cost savings, just like any new fab process (hell, TSMC won't have it available until 2013-2014, and that's not even close to replacing 300mm wafer usage). I wouldn't count on it being a "saviour" nor cost-friendly for the time frames we are currently interested in.

As for it facilitating larger chips, I'd say you have other problems to deal with rather than worrying about more efficient use of silicon wafers. Big chips are still costly to make just by nature, and typically very power hungry for the incredible number of transistors you'd be packing in there. And who knows how much of it goes to waste anyway since you'll be wanting to design a lot of redundancy so that your huge chip doesn't fail. It just doesn't make sense for a console design.

450mm again, is not some magical saviour.

[steampoweredgod]

Originally Posted by AlStrong

Sure, the larger the wafer, the less wastage for larger chips, but at what cost? There's a simple reason why it's taking so long and that it's simply not a trivial thing to manufacture. The whole issue surrounding 450mm adoption is the trade-off in the high difficulty in producing that large an ingot.

Ultimately, it just means it won't be cheap, and it'll be a very long time before the fabs pass on such cost savings, just like any new fab process (hell, TSMC won't have it available until 2013-2014, and that's not even close to replacing 300mm wafer usage). I wouldn't count on it being a "saviour" nor cost-friendly for the time frames we are currently interested in.

As for it facilitating larger chips, I'd say you have other problems to deal with rather than worrying about more efficient use of silicon wafers. Big chips are still costly to make just by nature, and typically very power hungry for the incredible number of transistors you'd be packing in there. And who knows how much of it goes to waste anyway since you'll be wanting to design a lot of redundancy so that your huge chip doesn't fail. It just doesn't make sense for a console design.

450mm again, is not some magical saviour.

Wasn't a formula given a while back, that indicated a small drop in clocks would yield a drastic drop in energy consumption?

Recall

Originally Posted by Gubbi

Power consumption scales with frequency cubed, a 20% reduction should result in around half the power consumption. I could see both MS and Sony go for largish dies but clocked lower than their PC counterparts. The cost of die area falls continuously, the cost for a given cooling solution does not.

An initial size of 250 mm² for the GPU shouldn't be a problem, over a consoles lifetime (~7 years) it should see two shrinks. Even if new process nodes fail to materialize, the cost of producing at the same node will continue to fall as capital costs are amortized.

I expect around 500mm² for CPU/GPU/eDRAM; The PS2, the PS3 and 360 were all around that at launch.

Cheers

Data from the past...

Playstation 2
Emotion Engine size 225 mm2 .25micron/250nm process (wikipedia entry suggests EE may be slightly bigger, iirc)
Graphics Synthesizer 279 mm2 .25micron/250nm process
Power consumption

if the GH-003 based SCPH-18000 is the same hardware-wise as the 10k and 15k (other than that small extra board), why did Sony downgrade the rated wattage on the case? It went from 50 watts in the prior two versions to only 48 watts in the 18k.-Elijah

Playstation 3
Cell B.E __________ 235.48mm^2 90nm process
Reality Synthesizer_ 258mm2 90nm process
Power supply 380 W
Consumption while gaming 195-209 Watts

While the cost benefits of 450mm wafers may take some time to pass, it is my understanding the industry has decided to move foward with such, and with 2013-2014 console launches it may affect them later during their lifespan transitions to smaller processes.

500mm^2 total combined seems fair game at least.

EDIT:
An 800W power supply can carry the latest and greatest single card pc hardware, if 20% clock drop halves power consumption, a 400W power supply should suffice for 20~% drop in performance.

[Ninjaprime]

Originally Posted by steampoweredgod

Wasn't a formula given a while back, that indicated a small drop in clocks would yield a drastic drop in energy consumption?

Recall


Data from the past...

Playstation 2
Emotion Engine size 225 mm2 .25micron/250nm process (wikipedia entry suggests EE may be slightly bigger, iirc)
Graphics Synthesizer 279 mm2 .25micron/250nm process
Power consumption


Playstation 3
Cell B.E __________ 235.48mm^2 90nm process
Reality Synthesizer_ 258mm2 90nm process
Power supply 380 W
Consumption while gaming 195-209 Watts

While the cost benefits of 450mm wafers may take some time to pass, it is my understanding the industry has decided to move foward with such, and with 2013-2014 console launches it may affect them later during their lifespan transitions to smaller processes.

500mm^2 total combined seems fair game at least.

EDIT:
An 800W power supply can carry the latest and greatest single card pc hardware, if 20% clock drop halves power consumption, a 400W power supply should suffice for 20~% drop in performance.

Theres an interesting major factor in those stats. Power consumption from PS2-->PS3 increased by ~8x to achieve its new performance. First gen PS3 was about as big and as hot as a console can really get, theres no 8x power difference this time. All your new performance has to be in the same or less power envelope.

[steampoweredgod]

Originally Posted by Ninjaprime

Theres an interesting major factor in those stats. Power consumption from PS2-->PS3 increased by ~8x to achieve its new performance. First gen PS3 was about as big and as hot as a console can really get, theres no 8x power difference this time. All your new performance has to be in the same or less power envelope.

80% performance of a geforce 580, how would that compare to current console performance?

And that is worse case as AMD tends to be more energy efficiency, and the powervr thread hinted 8-10x is viable, with advances in mobile gpus, within these energy constraints.

[TheChefO]

Originally Posted by steampoweredgod

Wasn't a formula given a while back, that indicated a small drop in clocks would yield a drastic drop in energy consumption?

Recall


Data from the past...

Playstation 2
Emotion Engine size 225 mm2 .25micron/250nm process (wikipedia entry suggests EE may be slightly bigger, iirc)
Graphics Synthesizer 279 mm2 .25micron/250nm process
Power consumption


Playstation 3
Cell B.E __________ 235.48mm^2 90nm process
Reality Synthesizer_ 258mm2 90nm process
Power supply 380 W
Consumption while gaming 195-209 Watts

While the cost benefits of 450mm wafers may take some time to pass, it is my understanding the industry has decided to move foward with such, and with 2013-2014 console launches it may affect them later during their lifespan transitions to smaller processes.

500mm^2 total combined seems fair game at least.

EDIT:
An 800W power supply can carry the latest and greatest single card pc hardware, if 20% clock drop halves power consumption, a 400W power supply should suffice for 20~% drop in performance.

Thank you.

I've been saying this 500mm2 die size for a while now.

Though I disagree on the need for a 400watt power supply.

Current high end GPU HD 7970 draws 250w @ 925MHz.

Dropping the frequency by 20% (740MHz) gives us 125w.

Matching the Tahiti Core with a Cell x2 or Xenon x2 @ 28nm (no increase in frequency aside from a "turbo core" mode should do the trick. This chip at 28nm should draw <40watts.

Another 10 watts for optical drive, ram, & misc brings the total draw to 175watts.

125w gpu
40w cpu
10w misc

Factor in power inefficiencies of the PSU, and this puts the need at around 220watts. 250watts to be safe.

Of course, the speed of these components may be adjusted to account for a specific power envelope, but the above is reasonable.

Component size:

352mm2 CPU
120mm2 CPU

472mm2 total

[hoho]

I'm fairly certain you can safely increase the "misc" power usage by at least twice. BD reader alone should take at least 10W, probably more. Also I don't think that fast interconnection between CPU and GPU is all that nice for power usage.

[AlStrong]

Originally Posted by steampoweredgod

While the cost benefits of 450mm wafers may take some time to pass, it is my understanding the industry has decided to move foward with such, and with 2013-2014 console launches it may affect them later during their lifespan transitions to smaller processes.

That makes it completely irrelevant for the design decisions. The upfront costs would be quite enormous, and there'd be huge uncertainty and risk as to when the cost savings would be passed down. Hoping on it for it to happen years later makes no business sense.

[TheChefO]

Originally Posted by hoho

I'm fairly certain you can safely increase the "misc" power usage by at least twice. BD reader alone should take at least 10W, probably more. Also I don't think that fast interconnection between CPU and GPU is all that nice for power usage.

As I pointed out pages back:

The ENTIRE power envelope for a netbook is 30watts.
(Screen, CPU, GPU, RAM, north & south bridge, HDD, USB power for external optical drives) power draw (19V x 1.58A = 30 watts)

For the discussion of a console, the TDP of "other" is fairly irrelevant.

Increase it from 10w to 20w. Doesn't change the thermal engineering of the box. Bump up the PSU a bit and case closed.

[function]

Originally Posted by TheChefO

Current high end GPU HD 7970 draws 250w @ 925MHz.

Dropping the frequency by 20% (740MHz) gives us 125w.

Just wondering how you came to that conclusion.

The Kinect transformer output ~13W btw. Not that it'd put much heat inside a console that supplied power to the Kinect sensor itself, but it'll go on the cost of the main PSU.

I really hope the next Xbox doesn't have an actively cooled power brick. That's a bit lame on the 360 tbh.

[TheChefO]

Originally Posted by function

Just wondering how you came to that conclusion.

Gubbi Quote above.

I don't care one way or the other about the power brick having a fan or not.

What's lame is gimping the hardware in the guts of the machine which limits the onscreen experience.

If a fan in the PSU means a Tahiti instead of a Turks, mark me down in favor of the fan in the PSU.

[hoho]

Originally Posted by TheChefO

Current high end GPU HD 7970 draws 250w @ 925MHz.

Dropping the frequency by 20% (740MHz) gives us 125w.

How about instead using HD7950? It's smaller (or can be if you actually produce a chip with less units instead of locking some out), roughly 30% slower and using only a little bit more than half of bigger brother's power.

Basically I'd say in the end it's cheaper to make a smaller but faster clocked chip than big and lower clocked one.

[3dilettante]

Originally Posted by hoho

How about instead using HD7950? It's smaller (or can be if you actually produce a chip with less units instead of locking some out), roughly 30% slower and using only a little bit more than half of bigger brother's power.

It's the same chip, just clocked lower, undervolted, and with parts fused off.

[function]

Originally Posted by TheChefO

Gubbi Quote above.

He says power consumption scales with frequency cubed, which fits pretty well with what I see on my laptop, and that would put a 250W chip at 80% speed at around 160W unless I dun goofed with my maths. If it's a full board you're looking at then power might not scale as it would for a processor alone.

If you're feeding the CPU from the same memory as the GPU then you might not be able to scale that GDDR5 speed back by 20%.

I don't care one way or the other about the power brick having a fan or not.

What's lame is gimping the hardware in the guts of the machine which limits the onscreen experience.

If a fan in the PSU means a Tahiti instead of a Turks, mark me down in favor of the fan in the PSU.

Consoles are always gimped in a number of ways. Not being able to leave the power supply where it might get dusty is super lame and somewhat gimpy. Put the PSU in the system (and add the cooling requirements to the rest of the system) or use a passively cooled brick. And I don't want a console as big as the launch PS3.

A "gimped" console that you buy is infinitely more performant than the console you don't buy.

[hoho]

Originally Posted by 3dilettante

It's the same chip, just clocked lower, undervolted, and with parts fused off.

I know but my point was that instead of taking a high-end huge chip and downclock it just make it a little bit smaller and run it somewhat faster. Yes, power usage won't drop as drastically but you'll make up the cost difference from increased chip count per wafer.

[AlStrong]

It's also better to not gimp clocks too much when taking triangle setup rate into account.

Originally Posted by function

I really hope the next Xbox doesn't have an actively cooled power brick. That's a bit lame on the 360 tbh.

PSUs have to be cooled somehow. It won't get any better if people get their wish for 300W on day 1.

Put the PSU in the system (and add the cooling requirements to the rest of the system) or use a passively cooled brick. And I don't want a console as big as the launch PS3.

Seems rather contradictory to me. The whole point of the external brick is so that you can reduce the size of the console (despite the internal PSU being smaller than external), or at least not factor it into the cooling design as well.

Making the external brick passive would mean a pretty big jump in heatsink size too, no? i.e. just adds to the shipping weight. At any rate, it's no small concern when there are all sorts of costs to consider!

edit:

lol http://k8.dk/Projekter/XBOX360%20PSU%20Project.htm

This mod seems a bit extreme, and I'm sure with that much metal it ought to be cool to touch after awhile, but I'd be curious about what the TDP of the PSU actually is so that you could find a more suitable heatsink. Ultimately, it's going to be enclosed within a non-conductive shell, so passive cooling may simply just not be feasible.