[EuroGamer] AMD has developed a 20nm version of Xbox One APU

[ShintaiDK]

well, I was also thinking about the other power related costs, like the power supply and cooling

The PSU for example is 20-25$. Cooling is around 5-10$.

Its very small savings at best. The biggest saving may simply be logistics and increased appeal due to lower power consumption and smaller cases.

 

[ShintaiDK]

That is only in the beginning, APU cost will go down as 20nm will get depreciated over time and wafer price will decrease.

But when will it beat 28nm in cost? The savings/benefits are simply to be found elsewhere, until EUV.

 

[3DVagabond]

But when will it beat 28nm in cost? The savings/benefits are simply to be found elsewhere, until EUV.

Well, there was obviously a worthwhile benefit, or they wouldn't have done it.

 

[AtenRa]

But when will it beat 28nm in cost? The savings/benefits are simply to be found elsewhere, until EUV.

Roughly when wafer cost at 20nm will be lower than double that of 28nm(assuming same yields etc).

 

[CakeMonster]

sony bumped up the psp ram and cpu speed with the slim. It isn't impossible and if they can get a few extra MHz then maybe they go for it.

We know they can. 20nm, no matter how immature, will give that headroom within a still lower power limit. And since its mature enough for mass production on this scale, they could bump the MHz or GPU cores.

History suggest that they won't even if it would have been extremely easy.

 

[naukkis]

For the same die size you get 11,4B transistors (1.9x) and keeping the same speed your power stays at 250W as well.

No it won't because power per transistor reduces only 25%. So 1.9x transistors at same speed results 1.9x*0.75*250W=356.25W

 

[jpiniero]

Well, there was obviously a worthwhile benefit, or they wouldn't have done it.

As I mentioned, it was probably part of the contract. Cost (to AMD) should be about breakeven or maybe slightly cheaper if it comes out sometime in the middle of 2015.

 

[NTMBK]

A refreshed Xbox One Slim at E3 2015 in a bundle with Halo 5 sounds like a solid strategy to me. Hell, I might buy one.

 

[AtenRa]

Ok i just remembered we have a comparison of direct shrink with NVIDIAs 9800GTX+.

http://www.nu-vista.com:8080/downlo...High Performance & Low Power Applications.pdf

55nm Technology
TSMC's 55nm process is a 90% linear-shrink of the 65nm process. It provides cost savings while maintaining the same
speed at similar or lower power. The 55nm process logic family includes general purpose (GP) and low power (LP) options.
Different from other half node process, 80GC or 0.11G, the 55nm process has no non-shrinkable rule. It means customer
doesn't need to pay effort to modify layout to meet layout rule after shrink if the simulation results shows performance can
meet original target. As with other half-nodes, the 55nm process provides a silicon-based SPICE model. The 55nm process
also provides standard cell, compiler and critical IPs for customer direct design-in. The 55nm technology provides compatible
speed but better power than comparable 65nm processes. The most attractive point is the directly shrink that provides more
than 20% gross die.

The 9800GTX was at 65nm
Die size = 324mm2
Transistor = 745M
Speed = 675MHz
Power = 140W TDP

9800GTX+ at 55nm
Die size = 260mm2
Transistor = 745M
Speed = 738MHz
Power = 141W TDP

So by keeping the same transistor count at 55nm (Half node) gave them almost 20% higher density, increased Speed by 9%, keeping the same power as 65nm.

So at 20nm you get 1.9x density AND 30% higher speed at the same power of the previous node(28nm).

So actually if you directly shrink a Hawaii to 20nm you will get both a smaller die (1.9x density) and a speed boost (30%) by keeping the same power.

 

[Haserath]

I can tell you what a 20nm direct shrink of 28nm Hawaii could be,

28nm R9 290X
Die size = 438mm2
Transistor Count = ~6B
Speed = 1000MHz
Power = 250W

20nm R9 290X
Die size = ~230mm2
Transistor Count = ~6B
Speed = 1000MHz
Power = 187W

Or you could raise speed to 1300MHz at 250W.

At 20nm, you can also have a ~438mm2 die with ~11,4B transistors at 1300MHz 1000MHz and 250W. That is almost double the tranistor count at same power with higher speed as 28nm Hawaii.

For a 20nm 550mm2 die, it could have ~14,3B transistors at ~1000MHz ~250W

But all that by simple using the same architecture of Hawaii.

Edit: 1000MHz with double the transistor count

11.4B transistors@1000Mhz would be ~350W. And 14.3B ~450W.

 

[cyclohexane]

only benefit is power consumption. Will not be clocked higher - due to (legitimate) bitching from people who bought the launch consoles.

 

[3DVagabond]

As I mentioned, it was probably part of the contract. Cost (to AMD) should be about breakeven or maybe slightly cheaper if it comes out sometime in the middle of 2015.

You have nothing to base that assumption on.

 

[tviceman]

In order to understand it better,

28nm Kabini
Die size = 105mm2
Speed = 1.6GHz
Power = 15W

20nm Kabini (direct shrink only) Two options

Option 1 (speed +30%)
Die size = ~55mm2
Speed = 1.6GHz + 30% = 2.08GHz
Power = 15W

Option 2 (Power _25%)
Die size = ~55mm2
Speed = 1.6GHz
Power = 15W - 25% = 11,25W

Hope that helps

Except those are all best case scenarios, and typically don't get anywhere near perfect scaling in perf/w and perf/mm2 when shrinking an existing product.

 

[Phynaz]

Why is everyone arguing over node shrinks? It's been nearly a decade since the "simple" shrink. Any new node today requires a redesign to the new nodes design rules.

You can't just say the same chip will be half the size, as it can't be the same chip.

 

[Silverforce11]

Except those are all best case scenarios, and typically don't get anywhere near perfect scaling in perf/w and perf/mm2 when shrinking an existing product.

Also it's TSMC, what they claim and what they can deliver reliably are two different scenarios.

 

[nenforcer]

hopefully it will give them enough power savings to build a much smaller and simpler console, and help to cut costs and allow them to target a significantly lower price than the PS4, because the sales figures for the Xb1 are disappointing compared to the PS4, and I think making it significantly cheaper than the PS4 is something they have to do.

as for performance/clock, they would never target higher performance with a die shrink like this, it would only cause problems, it's all about lowering the cost

I think this is exactly the reason why we are hearing about this first out of Microsoft and nothing from Sony on die shrinking the PS4 right now.

Sony, as the sales and market leader, can afford to stick at the 28nm node where they are ahead and are breaking even or even slightly profiting with each PS4 sold.

Microsoft, however, has to play catch up being in second place and cutting production cost quickly is something they can pass on to the consumer right away.

Sony can afford to wait until the 20nm process is very mature (and cost effective) before they shift any production over to it.

Microsoft's XBOX One Slim is also rumored to have a 1TB hard drive in it just like the new Call of Duty Advanced Warfare bundle does.

http://www.dailytech.com/Xbox+One+S...er+Cooler+20+nm+APU+From+AMD/article36813.htm

This is something Sony can easily match however without a manufacturing die shrink.

 

[NTMBK]

Why is everyone arguing over node shrinks? It's been nearly a decade since the "simple" shrink. Any new node today requires a redesign to the new nodes design rules.

You can't just say the same chip will be half the size, as it can't be the same chip.

To be fair, Jaguar cores were developed with ease of portability in mind.

 

[Bateluer]

I think this is exactly the reason why we are hearing about this first out of Microsoft and nothing from Sony on die shrinking the PS4 right now.

Sony, as the sales and market leader, can afford to stick at the 28nm node where they are ahead and are breaking even or even slightly profiting with each PS4 sold.

Microsoft, however, has to play catch up being in second place and cutting production cost quickly is something they can pass on to the consumer right away.

Sony can afford to wait until the 20nm process is very mature (and cost effective) before they shift any production over to it.

Microsoft's XBOX One Slim is also rumored to have a 1TB hard drive in it just like the new Call of Duty Advanced Warfare bundle does.

http://www.dailytech.com/Xbox+One+S...er+Cooler+20+nm+APU+From+AMD/article36813.htm

This is something Sony can easily match however without a manufacturing die shrink.

Keep in mind, the Xbox One's APU at 20nm will perform identically to the 28nm XB1 APU. The Xbox One will still be weaker than the PS4 is, always will be.

 

[f1sherman]

Except those are all best case scenarios, and typically don't get anywhere near perfect scaling in perf/w and perf/mm2 when shrinking an existing product.

I would think TSMC is basing their numbers on some kind of a test vehicle, a simplest possible(diode?) semiconductor device (if that at all) ,
which was then shrinked keeping the density fixed at 1.9x 28nm, and measuring two options - one with same power(30% clocks up), and the other one with same freq. (-25% power)


It's not like they are going to waste time designing several shrinks of existing and commercially available chips and then averaging these results.

So yes, definitely best-case scenario.

 

[AtenRa]

I would think TSMC is basing their numbers on some kind of a test vehicle, a simplest possible(diode?) semiconductor device (if that at all) ,
which was then shrinked keeping the density fixed at 1.9x 28nm, and measuring two options - one with same power(30% clocks up), and the other one with same freq. (-25% power)


It's not like they are going to waste time designing several shrinks of existing and commercially available chips and then averaging these results.

So yes, definitely best-case scenario.

TSMC (and every Fab) numbers are for transistors, not for ICs.

From Intels 14nm,

Switching Energy = power
Gate Delay = Speed

What that graph tell us is that you get the same speed at lower power OR higher speed at same power. Same aplies for TSMC 20nm over 28nm as well.

 

[Rakehellion]

The Xbox One Slim is coming early.

 

[Headfoot]

Keep in mind, the Xbox One's APU at 20nm will perform identically to the 28nm XB1 APU. The Xbox One will still be weaker than the PS4 is, always will be.

Most likely true.

I could see one other scenario though. Given that the Kinect takes 10% of the GPGPU processing time slices, and given that they now offer a Kinect-less Xbone: they could upclock an Xbox One Slim by 10% so that with the Kinect it has equal processing to a Kinectless Xbox One. That way you could play the games that normally take the Kinect's processing share, but still use your Kinect. Wouldn't do anything for day-to-day processing power as it would ultimately be the same top-performance