I would expect a lot of slides of some very informative information that will match up with this blog and the hardwork that mrx mrc and others are putting forward. Plus demos and hardware talks.
And now to address that big elephant in the room... nobody on this blog is stupid. On the conjecture the ones that ask questions or have dug for answers. When things are not correct, that is using common sense. The ones that are fools are the people who cant see both ways and believe that 3 billion on rnd was all for nothing.. there is so much going on in regards to transitional shift moving an industry forward in technology and software is no easy feat.
The next week is going to be fun.. enjoy
Insider: The official word should be " two heads are better then one " :)
Insider: The biggest clue for you is this X!! Check out the logo for dx12 and xbox one.. dig that.. right in front of your eyes.
When cd project got there new sdk they new they had to delay witcher 3 .. cos pc and x1 demand the dx12 ... watch what is happening at e3.. boom :)
Mistercteam: (mrX blog) already gone viral, they still ask where is 3D stacked?,we partly already uncover it
We already discuss it several weeks ago
Maybe I have to put or reposted that again
this is a summary from Side Xray that we got from XboxTurnon from chipwork (sometimes ago), then I enhanced it a bit
All explanation on Slide
It is clear it is stacked !!!
Mistercteam: DX12, new architecture, new execution model, multiple kernel
Mistercteam: I noticed that so many forum considered tech forum, not talk much about DX12-X1
It is different when before the console wars happen (after x1 & PS4 launch)
On those days they speculated and give a very good analysis
on upcoming DX or DX11.5 or later on
But after PS4 vs X1
it is change:
i think i know why
1. Most of that site are lately bit more like N4g, neogaf, i witnessed several times, when i joined several of that forum, post some tech related to x1, and someone try to downplaying it, even said it is FUD, when MS still not stated the spec,
2. So instead doing what they usually do, they feed up into the trap that probably based on what we see on N4g or neogaf that downplaying X1 , not even try to check like they usually do, what is op/cycle, why low yield ?, why MS as software company seems unprepared ?
fast forward today
it is clear that today is something on X1, the proof is started to get inline with MrX blog, it seems they feel it , as they are the one
that ridiculated X1 as 2 ACE as DX 11.1 , then also eSRAM etc,
Then BAM MS show DX12 with X1 logo
And we know , most of the forum dont want to discuss it, as it is like slap into their faces, as they try not to discuss it
Misterx: Agree. there is still time for thous experts to admit they were wrong and don't look like low grade students after so many info available on the internet they ignored.
Mistercteam: People will still thinking that can we fit that so many ALU?
this question is understandable
we live in old paradigma, when most of us only comparing
7970 is 32 CU need > 340 mm^2
are people forgot 2 thing
1. the ALU is not the one that takes space, it is the register,SRAM
2. Intel show they can fit 80 cores of Simple in order ALU, the precursor of Intel cloud chip capable to do 1.2TF at 2009 only
in area of ~ 200mm^2 at 65nm, at 28nm it is < 32 mm^2
so why not takes point 2 if we want to do many TF
remember we need register, we need something that can be programmed
it is why Intel 80 cores is not going forward at that time
as people getting difficult to program it, plus not much to
do as the limited SRAM etc
so to overcome this
SUre MS will take every step to make sure
remember this is not off the self part like Sony
this is super optimized.
1. They will use new node process --> hinted by Glofo
2. will use stacking
3. will super optimize the need of SRAM, reduce local SRAM
but make bigger 1T SRAM or eDRAM
with point 1-3, you can almost fit 2x7970 in the same area as 1x7970
not counting the cpu block, but X1 2nd layer is also multi layer
this is 10 year console remember.
There is no metric relate to closed box
first we must use standard metric like CU number ALU number etc
so better to said
that we have XX op/cycle resulted in ~ x TF
but with closed box performance we will have more power than that.
the TFlop metric will always relate to internal hardware
we must know the op/cycle first, and suprisingly
AMD slide showed 32CU = 512 op or 512 Ex unit
when PS4 known since 2012 as 192 Ex unit ---> 20 CU
X1 is 768 Ex unit active(from max 1024 Ex unit)
remember new node process, less wattage less heat
also they hinted power island --> Gated off per block
Plus insider said the Flop is resulted from the low clock they use
on 2nd layer, as the headroom is still there.
he said, 0.45 from 100% (as usual hinted word) (misterx: 65% as he said reduced by 45%)
smash_mouth: Is this real??...
The Crapgamer @clanMOH Follow
http://gyazo.com/28f5946a6ef96d302f5bcbf93ad0c5f9 … Do you guys see this as a legit Facebook post about PS4 and Sony groups? Looks pretty legit.