Xbox One (Durango) Technical hardware investigation

Status
Not open for further replies.
8 core "Low power" jaguar is not really that low power. It uses around the same power as the whole Wuu console.

Wishful thinking....

jeff_rigby is wayyy out there. Shouldnt use him as a source for anything.
 
Power regulation on the board looks normal to me.....

2 Phase RAM
2 Phase CPU ( Maybe 3 )
4 Phase for the GPU

So you could be looking at bare minimum of 8 phases for the whole machine which looking at the pictures is what it appears to have...
 
Last edited by a moderator:
Do those hardcore Wuu fans still believe even after the xrays of the dies?

I doubt it. But many have not looked.

But we are not talking about the same situation.

That chip must be the biggest (by far) and the most expensive ever put in a console. (By area, transistor count and likely price.)

So I do not think it is unreasonable to question the low performance CPU and GPU claims.

We do not know enough yet but the biggest chip does not seem to match with pulling a Nintendo or low performance. I think MS increased their team, talent and capability from 360 Xenon + Xenos SoC APU not the other way around. And by team I mean internal plus external involvements.

Could certainly be proven wrong but looking at the size of that chip I have confidence that the performance is of a similar magnitude.
 
What do you think is going to happen in 4 days? Specs? Really?

In four days I expect to get a closer look at the games performance achieved with what looks like the biggest chip ever put into a console. I think it is the biggest in terms of die size, transistor count and cost.

So I will try to infer from that. It would be nice to hear more specs but who knows. We might need to wait for chip delayering and photos from chipworks. But that least you know the die size is not the tiny thing inside the Wii U. You can at least see that now.

Does anyone know if any prior console had a chip anywhere near this size? (Die area/cost budget?)
 
MS is not going to release specs at E3. It clear they are pulling a nintendo.

Wuu 2.0 :cry:

As you see with the wuu is was a very smart move to not release specs. Hardcore wuu fans still believe is just some beast. It alway wait until e3... wait until whatever.

Once you release specs there is no hope....

No way it is a Wii U 2.0.

Take a look at the tiny die sizes at 40nm? (not sure):

http://www.anandtech.com/show/6465/nintendo-wii-u-teardown

Wii U
CPU 5.2mm x 6.3mm 32.76mm2 (tri PPC)
GPU 12.3mm x 12.7mm 156.21mm2 (I think the 32MB eDRAM is in here)

:arrow: That includes 32MB eDRAM. :!:

Xbox One is 32/28nm and massively bigger die size.

I just don't buy it at all.
 
I do not think so due to my arguments for the use of 1T-SRAM. So others might not agree but I consider 1T-SRAM more likely than 6T or 8T. 80MB 1T is is use at much higher freq in Power 7+ (and later Power 8). 32MB 1T is in Wii U die already, as someone pointed out.


1T-SRAM is a pseudo-static random-access memory (PSRAM) technology introduced by MoSys, Inc., which offers a high-density alternative to traditional static random access memory (SRAM) in embedded memory applications. Mosys uses a single-transistor storage cell (bit cell) like dynamic random access memory (DRAM), but surrounds the bit cell with control circuitry that makes the memory functionally equivalent to SRAM (the controller hides all DRAM-specific operations such as precharging and refresh). 1T-SRAM (and PSRAM in general) has a standard single-cycle SRAM interface and appears to the surrounding logic just as an SRAM would.

Due to its one-transistor bit cell, 1T-SRAM is smaller than conventional (six-transistor, or “6T”) SRAM, and closer in size and density to embedded DRAM (eDRAM). At the same time, 1T-SRAM has performance comparable to SRAM at multi-megabit densities, uses less power than eDRAM and is manufactured in a standard CMOS logic process like conventional SRAM.

MoSyS markets 1T-SRAM as physical IP for embedded (on-die) use in System-on-a-chip (SOC) applications. It is available on a variety of foundry processes, including Chartered, SMIC, TSMC, and UMC. :arrow: Some engineers use the terms 1T-SRAM and "embedded DRAM" interchangeably, as some foundries provide Mosys's 1T-SRAM as “eDRAM”. However, other foundries provide 1T-SRAM as a distinct offering.

The evidence doesn't support your claim.

Xbox engineers are on camera saying that their cooling system is designed to cool a chip with a 100W TDP. When you add up Jaguar TDP (25 to 35W) and 7790 TDP (85W), you have nothing left. Even after downclocking the 7790. Nothing.

Anand Shimpi says it's 6T:

https://twitter.com/anandshimpi/status/337296343208972288

And Major Nelson didn't correct the person on twitter with addressing the ESRAM rumor issue.
 
The evidence doesn't support your claim.

Xbox engineers are on camera saying that their cooling system is designed to cool a chip with a 100W TDP. When you add up Jaguar TDP (25 to 35W) and 7790 TDP (85W), you have nothing left. Even after downclocking the 7790. Nothing.

Anand Shimpi says it's 6T:

https://twitter.com/anandshimpi/status/337296343208972288

And Major Nelson didn't correct the person on twitter with addressing the ESRAM rumor issue.

I think your arguments against 1T are weak. The 100W is a big concern. But the brick looks kind of big for that, but perhaps Kinect 2 pulls power over USB3.

Licensing 1T is not an issue, chips contains licensed physical IP bits all the time. Also there are other 1T out there. Consider 80MB eDRAM in Power 7+ and the fact that 1T-SRAM is eDRAM at the core and that the two get called the same thing in the industry.

I do not see how Shimpi would have any idea what is inside that SoC.
 
So what's your theory for the extra 1.2B transistors then? Dual gpu's? SHAPE is huge?

6T (sram) makes sense, 1T (edram) does not.
 
I think your arguments against 1T are weak. The 100W is a big concern. But the brick looks kind of big for that, but perhaps Kinect 2 pulls power over USB3.

Licensing 1T is not an issue, chips contains licensed physical IP bits all the time. Also there are other 1T out there. Consider 80MB eDRAM in Power 7+ and the fact that 1T-SRAM is eDRAM at the core and that the two get called the same thing in the industry.

I do not see how Shimpi would have any idea what is inside that SoC.

Any reason why you completely ignored my specific CPU and GPU numbers?
 
So what's your theory for the extra 1.2B transistors then? Dual gpu's? SHAPE is huge?

6T (sram) makes sense, 1T (edram) does not.

Why does it not make sense?

If IBM can use 80MB of it why does it not make sense?

If there is 32MB in Wii U why does it not make sense?
 
Any reason why you completely ignored my specific CPU and GPU numbers?

Not ignoring it. I am checking out of a hotel in Seattle and running out of time.

I will need to wait for more info until I have more ideas on how the power and transistor count was used.

I might be proven wrong but the die size and transistor count does not make sense to me with just jag and a smallish GPU.
 
Not ignoring it. I am checking out of a hotel in Seattle and running out of time.

I will need to wait for more info until I have more ideas on how the power and transistor count was used.

I might be proven wrong but the die size and transistor count does not make sense to me with just jag and a smallish GPU.

:rolleyes:

People are trying to tell you and yet you are just plugging your ears, covering your eyes and muttering 'I refuse to believe'. This is a forum for talking about actual tech, confirming and debunking rumors, and sussing out exactly what parts are and what they do.
 
Last edited by a moderator:
Why does it not make sense?

If IBM can use 80MB of it why does it not make sense?

If there is 32MB in Wii U why does it not make sense?

Apples and oranges. Different process, not on die.

If they had to maintain another die for edram the long term cost may well have favored using gddr5. By having it on die they can reduce cost with shrinks.

We know the 5B transistor number. We know the power budget. We know the gpu is 12CUs. ESRAM 6T fits, nothing else really does with what we know.
 
If they had to maintain another die for edram the long term cost may well have favored using gddr5. By having it on die they can reduce cost with shrinks.

I'll probably sound stupid, but if you don't have it on the same die, why can't you reduce costs with shrinks, you could still shrink separate elements? I thought having it on die had something to do with making it ultra-low latency and for lower power requirements.
 
I'll probably sound stupid, but if you don't have it on the same die, why can't you reduce costs with shrinks, you could still shrink separate elements? I thought having it on die had something to do with making it ultra-low latency and for lower power requirements.

Bus size has a flat minimum size regardless of process. They might be able to shrink a 400mm die 3x but a 100mm die might only benefit from one shrink.
 
Over at Semi Accurate Forums someone posted what I think could be a pretty good technical catch from looking at the Xbox One mainboard (the wired photos).

So jeff_rigby pointed out the size of the GPU and CPU power planes and specifically that the size of the CPU power plain copper plus the associated buck converter components look *way* too large for the power consumption of jaguar.

So take a look at both this photo (zoomed out) and then also go look at the zoomed in photo which you can see over at the Semi Accurate Forums:

http://www.wired.com/gadgetlab/2013/05/xbox-one-development-photos/#slideid-138498

In short the CPU power plane and buck converter looks like it can power some much larger bigger cores or other blocks than the lower power jaguar cores.



So based upon die size, transistor counts, power consumption (as suggest by the CPU power plane copper cross section and the buck converter components) and MS's trends in the original and 360 I doubt it is jaguar. My bet and hope is that it is far removed from jaguar at the other end of the power spectrum.

Or I could have some fun and suggest that there are "additional powerful blocks" attached to the CPU power supply. Back to the dual APU rumors or the ray tracing block? (I am mostly just kidding but I would like to see revolution and serious power instead of evolution and jaguar cores.)

Maybe W8-lite runs on Jaguar and games runs on Power8 custom or K10.5+ enhanced/customized husky cores or steamroller-like customized cores.

So it might have jaguar but for the W8-lite and the constant on. Not for the games at all. The two get stitched together through the display planes which might involve the eSRAM.

Not sure which power plane the eSRAM is attached to. I would guess GPU.
Don't get your hopes up.
 
Over at Semi Accurate Forums someone posted what I think could be a pretty good technical catch from looking at the Xbox One mainboard (the wired photos).

So jeff_rigby pointed out the size of the GPU and CPU power planes and specifically that the size of the CPU power plain copper plus the associated buck converter components look *way* too large for the power consumption of jaguar.

So take a look at both this photo (zoomed out) and then also go look at the zoomed in photo which you can see over at the Semi Accurate Forums:

http://www.wired.com/gadgetlab/2013/05/xbox-one-development-photos/#slideid-138498

In short the CPU power plane and buck converter looks like it can power some much larger bigger cores or other blocks than the lower power jaguar cores.



So based upon die size, transistor counts, power consumption (as suggest by the CPU power plane copper cross section and the buck converter components) and MS's trends in the original and 360 I doubt it is jaguar. My bet and hope is that it is far removed from jaguar at the other end of the power spectrum.

Or I could have some fun and suggest that there are "additional powerful blocks" attached to the CPU power supply. Back to the dual APU rumors or the ray tracing block? (I am mostly just kidding but I would like to see revolution and serious power instead of evolution and jaguar cores.)

Maybe W8-lite runs on Jaguar and games runs on Power8 custom or K10.5+ enhanced/customized husky cores or steamroller-like customized cores.

So it might have jaguar but for the W8-lite and the constant on. Not for the games at all. The two get stitched together through the display planes which might involve the eSRAM.

Not sure which power plane the eSRAM is attached to. I would guess GPU.

Jaguar + SHAPE?
 
Status
Not open for further replies.
Back
Top