Next Generation Hardware Speculation with a Technical Spin [2018]

[Jupiter]

64GB seems like a lot to ask. GDDR6 is expected to go upto 32Gb per chip, so we may see 32GB (8 chips / 256bit bus) or 48GB (12 chips / 384bit bus). With HBM, HBM3 will have densities that could support 64GB with 4 stacks, I don't think HBM3 will be ready in 2 years for consumer deployment nor would they put 4 stacks on the interposer for a console.

Yes.

XBX1 can do 4K native for some games with 6Tflops. 12 Tflops would probably be sufficient for native 4k with minimal improvements overall in rendering. I don't think that will sell. There needs to be a compelling visual reason to upgrade.

In 2020 HBM should be cheaper in mass production than GDDR and 64GB should be possible too. Before they use 32GB they will probably wait another year.

I see no problem with other native resolutions like 1620p or 1800p. Especially since interlacing introduces ugly artefacts in motion. More important is the used AA where I liked the TAA of AC: Origins very much.

 

[Scott_Arm]

That's been proven super unpopular before, no? XBox arcade and 20GB PS3s weren't at all popular. Early adopters want the full-fat experience, with gimped consoles only making sense down the line selling to the bottom end. A normal console and expensive elite makes more sense to me, at maybe $400 and $600. You're bound to get your biggest fans paying the upper price.

A large drive is pretty much a necessity now too. It'll be tough to scale based on drive size. Games are already 50GB. Next gen will be insane. The base unit will need a big drive, or people will be upset.

 

[Scott_Arm]

I think there are a lot of arguments for keeping the core SOC + DRAM constant; economies of scale, simpler validation etc. Then differentiate on accessories.

Maybe a flash-only base SKU (Arcade, yay!) and a premium model with additional HDD+optical storage + possibly extra/elite controller. Launch at $300 for the base unit (loss leading) and $400 for the premium unit (at a profit). Drive the price down to $200 / $300 for base/premium models in three years time, then launch a PRO/X upgrade at $400.

Cheers

Playstation 4 is still $299 USD. I could easily see a base launch price of $400 again, and the price bottoming out around $300. No need to sell it at a loss at launch.

 

[MrFox]

In 2020 HBM should be cheaper in mass production than GDDR and 64GB should be possible too. Before they use 32GB they will probably wait another year.

Hbm needs a breakthrough right now. It's a high risk choice considering hbm repeatedly failing it's targets.

Hbm have additional costs associated, stacking 8 dies is still 8 separate chips from the wafer, and yield drops because of the stacking. Then there is the interposer affecting yield.

The "low cost" hbm, and organic interposers, are going to be very interesting. But If there was any indication of hbm being anywhere near or even lower cost than gddr6 per GB in the next year or two, I think we wouldn't have gddr6 appearing at all. Now all three major memory manufacturers are ramping up volume gddr6 for gpus from both nvidia and amd. (and gddr5x will go the way of gddr4)

Gddr is one chip in a plastic case, tested individually, and the PCB have near 100% yield. Sony tried high risk tech in the past and got burned. The low yield of cell and blue lazers at launch was bad luck, but it was caused by the choice to take that risk. OTOH, there is nothing high-risk in the PS4.

 

[HBRU]

I think the right move for Sony is a 16 GB GDDR6 console first coming at 2018 november with improved 4K CB and maybe stable 60 fps on all the PS4 games where standard ps4 is 720@30 fps or 900@30 fps... Bus 256 bits of course CPU jaguars... Build at 7nm dont know how much they can be maximal clocked... Cooling similar to xbox x... 4k blue ray... Price 600

 

[HBRU]

then really next gen 2025

 

[Deleted member 11852]

The speculation threads are always good for a laugh.

 

[Grall]

The speculation threads are always good for a laugh.

You weren't lurking around when Deadmeat swore that the xbox 360 would come with a 10GHz Intel "Tejas" CPU, were you? This was probably around 2003 or so, I don't really recall.

 

[Michellstar]

You weren't lurking around when Deadmeat swore that the xbox 360 would come with a 10GHz Intel "Tejas" CPU, were you? This was probably around 2003 or so, I don't really recall.

Well, Actually it was Intel´s fault

 

[Gubbi]

That's been proven super unpopular before, no? XBox arcade and 20GB PS3s weren't at all popular.

The Arcade SKU hit a price point where it competed with the Wii on price. Games ran without a HDD on the 360, although not well. Later games basically required a HDD making the Arcade+HDD a lousy value proposition due to the $90 price tag of the HDD, but the Arcade SKU sold well enough.

I'm suggesting a less gimped base system, with $50-60 worth of FLASH storage in it (should be around 200GB in 3 years time). Then allow any USB-C drive to connect to the console for mass storage. Yeah, a lost opportunity to cash in on peripherals, but also a great way to reduce your BOM.

Early adopters want the full-fat experience, with gimped consoles only making sense down the line selling to the bottom end. A normal console and expensive elite makes more sense to me, at maybe $400 and $600. You're bound to get your biggest fans paying the upper price.

Early adopters are early adopters because of the high initial price. If you can hit the low price point, you'll see mass adoption from the get go. That might not be such a good idea wrt. initial supply, but then stagger your launch with the cheap SKU launching 1-2 quarters after the premium SKU.

Cheers

 

[Michellstar]

The Arcade SKU hit a price point where it competed with the Wii on price. Games ran without a HDD on the 360, although not well. Later games basically required a HDD making the Arcade+HDD a lousy value proposition due to the $90 price tag of the HDD, but the Arcade SKU sold well enough.

I'm suggesting a less gimped base system, with $50-60 worth of FLASH storage in it (should be around 200GB in 3 years time). Then allow any USB-C drive to connect to the console for mass storage. Yeah, a lost opportunity to cash in on peripherals, but also a great way to reduce your BOM.



Early adopters are early adopters because of the high initial price. If you can hit the low price point, you'll see mass adoption from the get go. That might not be such a good idea wrt. initial supply, but then stagger your launch with the cheap SKU launching 1-2 quarters after the premium SKU.

Cheers

If the new console its BC, It would be the ideal SKU for those who currently have large external storage attached to their systems

 

[Silenti]

I am sure to display my ignorance with this one but...

What about a return to split pools of memory but instead of DDR and GDDR/HBM, could you go with an SSD/XPoint solution? Specifically, the high bandwidth expensive stuff is really only needed for graphics work and some extra for a loaded app/ OS. How much GDDR6 or HBM would you really need for that?

 

[Tkumpathenurpahl]

Don't worry, we're all displaying our ignorance here.

 

[Theeoo]

PS4 Super Pro 60 FPS edition.

4x Ryzen 2 cores @ 3Ghz
8TF Gpu
8Gb DDR5 + 4GB HBM2 Ram

Manufactured on 12nm to be released Christmas 2018 or H1 2019.

 

[Shifty Geezer]

I am sure to display my ignorance with this one but...

What about a return to split pools of memory but instead of DDR and GDDR/HBM, could you go with an SSD/XPoint solution? Specifically, the high bandwidth expensive stuff is really only needed for graphics work and some extra for a loaded app/ OS. How much GDDR6 or HBM would you really need for that?

Flash storage is far too slow to replace RAM. You need enough RAM for the graphics work and everything else in the program ;-), with your non-DRAM storage being good for streamed assets. So either you have one larger pool of graphics-speed RAM and share that with other workloads, PS4 style, or you have split pools of RAM and VRAM. With either, a lump of high speed flash would be highly beneficial, but it can't replace the multiple GBs needed for GPU and CPU workspace.

 

[Tkumpathenurpahl]

If we assume that the PS5 or XboxTwo utilise a large pool of NAND, what would be the best pairing in terms of memory? 24GB of GDDR6 or 16GB of HBM?

Both fast, but, when it comes to affordability, which is the best tradeoff? Fast and huge, or blistering and large?

 

[HBRU]

It depends on the bus byte-wide...

 

[McHuj]

PS4 Super Pro 60 FPS edition.

4x Ryzen 2 cores @ 3Ghz
8TF Gpu
8Gb DDR5 + 4GB HBM2 Ram

Manufactured on 12nm to be released Christmas 2018 or H1 2019.

Who would this console be for? And what compelling reason would there be to buy it?

Games and their design would still be constrained by the original PS4.

 

[MrFox]

If we assume that the PS5 or XboxTwo utilise a large pool of NAND, what would be the best pairing in terms of memory? 24GB of GDDR6 or 16GB of HBM?

Both fast, but, when it comes to affordability, which is the best tradeoff? Fast and huge, or blistering and large?

If it needs 4 stacks to get something faster than gddr6 then I doubt the price will reach console level any time soon.

The gpu would need to be sized to use that bandwidth. If we look at current mid-gen consoles, the balance is around 50GB/s per TF. Unless there are more features yet to be unveiled helping save bandwidth like compression etc... Is there some rumors about that?

256bit gddr6 can feed a 10TF console
384bit gddr6 is 15TF
2 stacks hbm2 is 10TF
4 stacks hbm2 is 20TF
That's assuming 16gbps gddr6 and 2gbps hbm2. Neither are the fastest expected but probably the affordable highest volume bins in a year or two.

Gpu cards are a bit less bandwidth per TF but it's logical since they don't have a cpu to feed, just a pcie bus doing memory transactions both ways.

Hbm3 would be in another league entirely.

 

[HBRU]

thank you Mr. Fox... Clear.... Of course a 8~10 TF console will still be into the PS4 / Xbox One family... IMHO... Also, please, do the same math related to CPU power as is shares bandwidth with the GPU...