Xbox One (Durango) Technical hardware investigation

Status
Not open for further replies.
Thread Reopened!

We didn't get specs, so the official line is that they are as per the leaks. 1.2TF SOC, 12 CUs, 32MBs eSRAM, yadayada. As this is a technical thread, anyone holding out for 'secret sauce' will have to present serious evidence, otherwise it'll just be noise and they'll be evicted. I'll move some posts over from other threads.

We got SOMETHING...

GIMMIE GIMMIE GAMES said:
It has been rumored for months now that the OS for the new Xbox would use 3GB RAM and Whitten confirmed in this interview that it will require “around” 3GB. So it could be a bit more, could be a little bit less. This leaves about 5GB RAM for game developers.

http://gimmegimmegames.com/2013/05/microsoft-confirms-xbox-one-os-uses-3gb-ram-5gb-ram-for-games/
 
anyone holding out for 'secret sauce' will have to present serious evidence

there is already the thread for the special sauce :) the cloud computing, the major evidence is that it's actually official by the official source, microsoft

let's talk about 6 thread per core, this is not jaguar as we know it, if it's jaguar at all
 
I haven't heard of the 6 thread per core. Still a haven't found a good summary! Can someone provide something here of the actual info revealed?
 
It's regular Jaguar... and it's not six threads per core, but six operations per cycle, which is what Jaguar does.

edit: or maybe it was instructions per cycle, but the point is, it's a regular Jaguar.
 
Last edited by a moderator:
the internals and mainboard of xbox one
http://www.wired.com/gadgetlab/2013/05/xbox-one-development-photos/?viewall=true

Xbox-one-mainboard-exposed.jpg


iths high res, the markings readable

sorry if its been posted
 
yes, seems like
you don't need an high end gpu if it will be used only in low interactions latency areas near the players

Shall I kindly ask MS to change the GPU to a 3Dfx Voodoo then ?
The more power you have the more you can do, the fact you *may* be able to move around *some* processing on a server doesn't change that at all, at best it just means you'll be able to spend that power all in a smaller area, but you still want as much power as possible for photo-realism...
 
yes, seems like
you don't need an high end gpu if it will be used only in low interactions latency areas near the players
Nonsense! You imply cloud servers would actually render part (or rather, the majority!) of the frame for each player? Good luck with making that either practical OR economical... Also, it would mean slaving the console to the internet permanently, no offline play possible.

So, NO. This is NOT what they're doing!
 
The cynic in me says they really, REALLY mis-spent those 5B transistors. Sony may have a SMALLER APU which has more compute power AND more bandwidth.

In the long term, the MS system will cost-reduce much, much better. SRAM gets perfect scaling, and with all that die area, they will never be pad-limited. Every full node shrink will cut the cost of the main chip in half. I can see the XBox One selling for $99 in the trail end of it's life, I can't see PS4 ever doing that.

So, while it's a much worse system right now, it will remain to be seen.
 
So we're looking at 17GB/s bandwidth?

multiply by four...it's not connected thru a 64bit bus, but a 256bit one. so 68GB/sec.

in post conference they talked about a 200 GB/sec bandwidth, i guess an aggregate one, so the ESRAM would be 132 GB/sec
 
I think you need to understand the technology better before you try telling people that they are wrong.
Remember that you're talking with the guy who has been dead convinced that XB1 would have GDDR5 RAM because some Rare guy said so (whom obviously knows much better than some clueless guy at vgleaks...), and who has viciously attacked anyone disputing his ideas.

So, I think you're wasting your time really. ;)
 
so the ESRAM would be 132 GB/sec
SRAM is likely at the rumored 102GB/s, with the remainder 30GB/s coming from somewhere else in the system (speculated in another thread/post as the CPU-GPU on-chip interconnect being a likely candidate.)
 
multiply by four...it's not connected thru a 64bit bus, but a 256bit one. so 68GB/sec.

in post conference they talked about a 200 GB/sec bandwidth, i guess an aggregate one, so the ESRAM would be 132 GB/sec


If (a big if) they are not rounding up aggregate bandwich, did they upped GPU clocks?

Wasn´t the suppoused Esram bandwich to be just 102 GB/s?
 
SRAM is likely at the rumored 102GB/s, with the remainder 30GB/s coming from somewhere else in the system (speculated in another thread/post as the CPU-GPU on-chip interconnect being a likely candidate.)

If (a big if) they are not rounding up aggregate bandwich, did they upped GPU clocks?

Wasn´t the suppoused Esram bandwich to be just 102 GB/s?

of course i dont know, but i head in the post conference 200 GB/sec for sure ;)
 
Status
Not open for further replies.
Back
Top