Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]
- Thread starter Shifty Geezer
- Start date
- Status
TR is a pretty bad basis of comparison though given how heavily modified it was (the hairworks implementation in particular). The PC version is severely hampered by the last gen console legacy and poorly optimised first gen hairworks tech. If they were to release the DE on PC today I'm sure a 970 would have no problem whatsoever hitting a solid 60fps and well above that at the same graphics settings..
But that's not what i am saying. It's obvious that the exact Ps4 version will run better on my PC, all i am saying is that the drawbacks of the Ps4 version are not as obvious as 30 v 60 so they were correct in ditching out tessellation and environment quality and possibly LOD too, to get to that "somewhat" 60 experience. Also, the first iteration of TressFX still has LOD for the strands.
upnorthsox
Veteran
Any info/links on v3? I haven't seen anything on it yet. Thanks.
D
Deleted member 11852
Guest
Nose hair?
I have finished the game on PS4 and I can assure you that the game runs at 60fps more often than not even in some big levels. Rarely does not correctly quality it.
Megadrive1988
Veteran
Assuming Sony and Microsoft's successors to PS4 and XB1 are going to, for starters, have APUs with Zen CPU cores (regardless of specifics like Zen+ Zen 2, core count, clockspeed, etc) would that not inherently be a huge boost in CPU capability over Jaguar in the current consoles? Based on what we already know about Zen.
I can only imagine, as Mark Cerny's team conducts a postmortom of PS4 and starts to poll 3rd parties asking questions what they'd like to see in a next generation system, they're almost certainly going to want better/more powerful/faster CPU.
I can only imagine, as Mark Cerny's team conducts a postmortom of PS4 and starts to poll 3rd parties asking questions what they'd like to see in a next generation system, they're almost certainly going to want better/more powerful/faster CPU.
We know little about Zen, but the data we do have and an assumption it can at least reach a level comparable to desktop Sandy Bridge or later chips provides a lot of room for improvement given how vastly more powerful even pedestrian i5 or i7 chips are over the consoles.
Zen should bring SMT, extensions like FMA, a more performant FPU, and the promise of a revamped cache hierarchy.
That Zen is expected to scale to server loads, since that and the server-derived FX tier is the initial target, is a hint that one of the pain points in the console CPUs should be helped significantly. Jaguar's module sharing penalties show it was not particularly built to scale to higher counts, and the memory latencies for the consoles are terrible.
The promise of an improved cache hierarchy and the design's targets are indications that Zen better not have trouble doing much better.
Zen should bring SMT, extensions like FMA, a more performant FPU, and the promise of a revamped cache hierarchy.
That Zen is expected to scale to server loads, since that and the server-derived FX tier is the initial target, is a hint that one of the pain points in the console CPUs should be helped significantly. Jaguar's module sharing penalties show it was not particularly built to scale to higher counts, and the memory latencies for the consoles are terrible.
The promise of an improved cache hierarchy and the design's targets are indications that Zen better not have trouble doing much better.
That's not what @psorcerer's questionnaire response will be.
It's likely too early to say what devs will want, because they'll only just be exploring compute this gen. But better CPU will be a given, I'm sure.That's not what @psorcerer's questionnaire response will be.
I am sure unified memory with HBM 2 or 3 will be a must.
Megadrive1988
Veteran
Any predictions on what kind of bandwidth we might be looking at for eventual HBM3 ? Given that HBM2 is meant to get 1 ~ 1.2 TB/sec.
D
Deleted member 13524
Guest
Is such a thing as HBM 3 in the pipeline at all?
I figure HBM2 should be able to serve the consumer market perfectly fine for many years to come.
I figure HBM2 should be able to serve the consumer market perfectly fine for many years to come.
Nobody has publicly mentioned it. I expect HBM2 to remain on market for a long time. At first it will be used only for very expensive cards, and in 2-3 years go down to midrange.
Even if there is a HBM3 in those timescales it's unlikely consoles will utilise the highest spec option of that memory tech I would imagine something equivalent to a 2 stack version of HBM2. So with a hypothetical HBM3 we're probably looking at something like 32GB @ 1TB/s with the full spec of that memory tech being twice that.
Megadrive1988
Veteran
Not specifically, or at least publicly, but SK Hynix did publish a roadmap with plans to scale bandwidth per stack through the early 2020s. Again, nothing specific.
I'm sure it will. Probably next-gen consoles too. But who's to say, by 2019-2020 HBM2 might not be enough for high-end / enthusiast graphics cards.
It's like asking Facebook users questions about features Facebook should have. Hint: nobody, who implemented exactly what users want, ever succeeded.
- Status