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

Status
Not open for further replies.
These historical norms are becoming very bad to use as a way to guess/hope what new consoles will be.

I'd be shocked if a PS5 had 32GB of RAM. I wouldn't be surprised if was only double what PS4 has...but at much higher bandwidth.

Could happen if there's another density improvement... Basically 32Gb per chip.
 
Maybe they can use the same tech as the SSG and make use of NAND as ram while everything that is being used per scene can fit inside the HBM cache. That way, we might be able to have 512GB of "RAM" in a console.
 
I think there's a good proportion of gamers who love awesome visuals who don't choose PC for numerous reasons. These people would love monster consoles with many bling.

You bet. Isn't the bar really high with even original PS4? So, graphics "up to eleven" on high end / superfast console might be enough for most.
PC gaming is doing very well, maybe due to the ever increasing amount of global players. [as in people who play games]

But some who would have played on PC in older times don't : it used to be that any new PC in the early 2000s, good for Internet and pirated movies etc. ran Unreal Tournament and Half-Life at 100fps (unless it had a really terrible GPU but these were still playable). Or every 486 ran Doom basically. Now, get an Atom quad core laptop with 2GB / 32GB and you can't even run Crysis or if it does run it might be hellish (and that's 10-year-old now!)

So, I assume there are many "unsophisticated" users who don't know game_title_X on a PS4 Pro or Xbox X is supposed to look like crap because on web forums there are "downgrading scandals" or the draw distance is a kilometer less than the PC version or they should count the specular highlights etc.
But in screenshots, videos and game play the damn thing looks better than Crysis and runs in real high res.
 
I think there's a possibility that Sony could throw a curveball by switching to an NVIDIA solution. Sure that would break compatibility, but if they are not interested in the rolling generations, then it's a possibility.

That might allow them to use the 12nm NVIDIA node from TSMC to produce a very large chip earlier and cheaper than whatever solution AMD can come up with.

A Volta based GPU with 1080ti performance coupled with a based ARM CPU.
 
I think there's a possibility that Sony could throw a curveball by switching to an NVIDIA solution. Sure that would break compatibility, but if they are not interested in the rolling generations, then it's a possibility.

That might allow them to use the 12nm NVIDIA node from TSMC to produce a very large chip earlier and cheaper than whatever solution AMD can come up with.

A Volta based GPU with 1080ti performance coupled with a based ARM CPU.
That's totally unnecessary.
Sony could use any node it wants.
Like they did with PS4 Slim/Pro APU which uses TSMC 16nm unlike AMD cards on GlobalFoundries 14nm.
Ryzen is better than any ARM now.
TSMC 12nm is very short lived and used for a very expensive stuff.

I think it will be Ryzen (+)/ Custom Navi APU with low cost HBM without interposer.
 
That's totally unnecessary.
Sony could use any node it wants.
Like they did with PS4 Slim/Pro APU which uses TSMC 16nm unlike AMD cards on GlobalFoundries 14nm.
Ryzen is better than any ARM now.
TSMC 12nm is very short lived and used for a very expensive stuff.

I think it will be Ryzen (+)/ Custom Navi APU with low cost HBM without interposer.
Won't low cost HBM only be available in 2020 though?
 
From what I understand... Samsung's proposed low cost HBM still requires an interposer, but it can potentially use a lower cost organic interposer instead of silicon. The pitch have to be wider, but the number of vias is lower so I guess it balances out?
 
From what I understand... Samsung's proposed low cost HBM still requires an interposer, but it can potentially use a lower cost organic interposer instead of silicon. The pitch have to be wider, but the number of vias is lower so I guess it balances out?

What capacity is projected by then??
 
I'm sure there are far more lucrative markets than consoles that would suck up what little volume there will be.
 
From what I understand... Samsung's proposed low cost HBM still requires an interposer, but it can potentially use a lower cost organic interposer instead of silicon. The pitch have to be wider, but the number of vias is lower so I guess it balances out?

AFAIK, HBM doesn't really need an interposer. The HBM stacks could be stacked on top of an APU/dGPU and connect directly with TSV. The reason why it hasn't been done is because a) memory amount per-stack has been rather small and b) HBM is made to pair directly with high-performance chips which in turn need to be in direct contact in the heatsinks or they'll start heating up and throttling down very fast.

Also, I think working on interposer-less HBM would make it too close to Wide I/O?

I think I saw an article about IHVs trying to come up with something like vertical copper tubes to solve the cooling problem on stacked solutions, but I don't know how fast such a solution would come to the market (or if it's even effective).
 
32GB 1TB/s HBM3 would be pretty nice and seems reasonable especially if PS5 hits the market towards the end of 2020 in NA and Europe, early 2021 everywhere else.

Yeah in a complete vacuum ignoring performance balancing and costing that go into designing a console ...32GB 1TB/s HBM3 would be pretty nice in 2020....
 
Also, I think working on interposer-less HBM would make it too close to Wide I/O?
Sony used Wide-IO directly attached to the Vita's SOC. I haven't heard much about Wide I/O for some time. I'm not sure being too similar to it would worry HBM designers.

I think I saw an article about IHVs trying to come up with something like vertical copper tubes to solve the cooling problem on stacked solutions, but I don't know how fast such a solution would come to the market (or if it's even effective).
Some initial steps have been taken, given that HBM includes thermal vias and dummy bumps that help conduct heat through the stack. It's still a significant barrier.
 
Sony used Wide-IO directly attached to the Vita's SOC. I haven't heard much about Wide I/O for some time. I'm not sure being too similar to it would worry HBM designers.

Yes, I meant that pursuing interposer-less HBM would have made it too close to Wide I/O during its inception period (I assume ~2008-2010?).
The latest article I can find about Wide I/O2 dates back to 2015, and the only Wide I/O adoption I know of is Vita's SoC from 2011, so I don't think HBM designers are worried with technology/market overlapping nowadays.
 
Status
Not open for further replies.
Back
Top