Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]

Status
Not open for further replies.
PS5:
-32GB HBM2
-4 core Cell 2.0 with 16 satellite processors each: so 64 SPU's. Each Cell has 12MB of L1 cache, no L2 cache
-24.78TF (DP) GPU, built in 4 clusters, each side of the Cell has one for "0 latency processing"
64GB of SSD (6400MB/sec read) attached to 4TB mechanical HDD. it takes a few minutes to 'change' games, but after that, almost zero loading

TDP 315W
The cooling consists of at least 2KG of high quality heatsinks/pipe system. Noise level comparable to the first generation PS4

Xbox4:
-option on windows PC's enabled by having a pc with certain requirements PCI 4-spec and such
-widely scalable; being able to receive PS5 ports, as well as downscale to different levels of PC, also above.
-MS has a "OEM" Xbox4 (PC) every year, built inside a console casing

Nintendo:
-handheld only
-lincensed games for apple tv, and android / iOS
 
Egmon83 said:
PS5:
-32GB HBM2
-4 core Cell 2.0 with 16 satellite processors each: so 64 SPU's. Each Cell has 12MB of L1 cache, no L2 cache
-24.78TF (DP) GPU, built in 4 clusters, each side of the Cell has one for "0 latency processing"
64GB of SSD (6400MB/sec read) attached to 4TB mechanical HDD. it takes a few minutes to 'change' games, but after that, almost zero loading
Click to expand...

413.gif


It's been a while since someone wanted Cell back in a console. :D
 
DieH@rd said:
It's been a while since someone wanted Cell back in a console. :D
Click to expand...

Cell would probably do better now than it did in 2006/07 when it was a brutal introduction to mass-parallelization of code to extract performance. Assuming the architecture had kept pace with modern developments and the tools were of the calibre of PS4. I'm not advocating a return to Cell but it launched at a time when many game developers only had experience of building games around a sequential game loop with little concurrent activity except between the CPU and the GPU.

Remember that the Cell project began before 80x86 had HyperThreading. I'd been in servers a few years by then so it was all interesting to watch those game devs transition to new ways of writing code, which I'd just been through myself. Yeah bitches, welcome to coding for men 101. :yep2:
 
Shifty Geezer said:
Would be nice ot see what a Cell 2 would be like. As a first outing it wasn't a bad design. Really need a chance to learn from real use and adapt. But I guess GPU compute has taken over everything SPU was really designed for.
Click to expand...

Well not completely, since with current designs there's still an appreciable latency and thus scheduling impact when switching between CPU tasks to GPU compute.

There was certainly something to be said for having some beefy vector processing units local to the main CPU core(s). I suppose the AVX(2) units have really taken over that role though.
 
You're not curious? I'm not suggesting Cell in a console insteadof x86! Just curious as to what the new processor would be like, what changes would be made, and how it'd compare.
 
Well of course they aren't because they have to fit the bill! Me not paying, I'm very curious for IBM to make a Cell2 (even if no-one'll use it!)...
 
Shifty Geezer said:
Would be nice ot see what a Cell 2 would be like. As a first outing it wasn't a bad design. Really need a chance to learn from real use and adapt. But I guess GPU compute has taken over everything SPU was really designed for.
Click to expand...

The Cell architecture did scale incredibly well without the performance bleed you see with a lot of other microarchitectures. IBM's 'Roadrunner' supercomputer was the first non-distributed supercomputer to break the petaflop barrier and IBM did that in 2008 with a system design for the Los Alamos National Laboratory that wasn't even finished.

It reminds of other flash-in-the-pan technologies which appeared in a blaze of excitement then vanished. Do you remember transputers and later DSP? Like Cell, both were also designed to crack very specific problems but disapeared when conventional technology either subsumed the technology (DSPs) or an alternative solution was found by repurposing existing hardware (Compute). Not to mention physics co-processors.

Shifty Geezer said:
Well of course they aren't because they have to fit the bill! Me not paying, I'm very curious for IBM to make a Cell2 (even if no-one'll use it!)...
Click to expand...

Talk to IBM's Aerospace and Defence Division. :nope:
 
DSoup said:
If that happens, the UK won't just leave the EU, we'll leave Earth. And take Rockstar North with us! :yep2: We will, however, leave Peter Molyneux behind.
Click to expand...

Oh God that is cruel. I really, on an entirely personal level, don't much care about Rockstar North, but you cannot leave us with both Trump and Molyneux! Mass suicides, dogs and cats living together!
 
cheapchips said:
So TMCS are saying to shareholders that 7nm will be ready in H1 2018 for mobile and performance chips. They must be pretty confident then?
Click to expand...

lol

When TSMC says "1H 2018", translated back into regular English that means "99% coming out before 2020".

Their timetables are always laughably inaccurate.
 
Status
Not open for further replies.

Similar threads

A
Full transcript of John Carmack's QuakeCon 2005 Keynote
3 4 5
Replies
98
Views
32K
nemisis
N
Back
Top