Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

[iroboto]

You need to understand who was saying it and their perspective. Cerny wasn't lieing but it was from their perceptive and benchmarking.
Also it was about what they felt was the best way to get to 10TF, not that a higher raw TF value isn't better.

Agreed, He wasn't lying, and he did explicitly also state that by increasing clocks there were limitations because memory wasn't going any faster. Most people choose to ignore his caveat. So while his statement is that a rising tide lifts all boats; the pipeline goes faster everywhere, this places additional pressure on memory being able to delivery on both it's bandwidth and latency.

This wasn't the case with XBO however, as much as some people try to make equivalence, XBO main form of bandwidth was ESRAM. And esram being embedded into the GPU meant that it was sync'd with the clock rate increases. So when they moved the GPU an additional 50+ some odd Mhz, that increased the bandwidth of esram to feed the increased pipeline. If they ever managed somehow to get XBO to PS5 clockspeeds, lol, esram would have 500 GB/s paired with a whopping 3.4 TF of compute power.

 

[goonergaz]

I see what you're saying. Sony will drop PS4/Pro like a stone and launch their 2021+ games only on PS5 and PC going forward.

no, it’ll be like previous gens with Sony, when a new console launches it’ll have proper exclusives that take advantage of the new hardware and not held back by needing to run on 1TF GPU with a shit slow HDD.

Or are you saying Sony won’t have PS5 exclusives that take full advantage of the new hardware (like MS are doing?)

 

[Jay]

He wasn't lying, and he did explicitly also state that by increasing clocks there were limitations because memory wasn't going any faster. Most people choose to ignore his caveat. So while his statement is that a rising tide lifts all boats; the pipeline goes faster everywhere, this places additional pressure on memory being able to delivery on both it's bandwidth and latency.

Yea I explicitly said he wasn't lieing.
But people take what he says and extrapolate it to mean things he didn't actually say.

He specifically was talking about different ways to get to 10TF. Not that a slower but wider 12TF would not be better.

Hence why people keep saying that the road to PS5 shows that the frequency could be a huge distinction/advantage. When that isn't what was actually said in comparison to xsx.

I've not seen anyone say, oh but xsx has advantage of being so much wider.
That will come when MS 'explains' why wider is better i guess

 

[iroboto]

Yea I explicitly said he wasn't lieing.
But people take what he says and extrapolate it to mean things he didn't actually say.

He specifically was talking about different ways to get to 10TF. Not that a slower but wider 12TF would not be better.

Hence why people keep saying that the road to PS5 shows that the frequency could be a huge distinction/advantage. When that isn't what was actually said in comparison to xsx.

I've not seen anyone say, oh but xsx has advantage of being so much wider.
That will come when MS 'explains' why wider is better i guess

sorry forgot to write 'agreed'.

 

[Inuhanyou]

Do you have a link to that particular benchmark? I'd be interested to see what they were doing that was using so little bandwidth.

In actual games the X1X can be putting out anywhere from 40% to 100% more pixels [edit: in a given period of time], so I'm inclined to think that is most cases, in the real world, bandwidth is the real limiter.

Exactly. Bandwidth is the main killer of PS4Pro..its easy to see because the deltas we see between Pro and X are not anywhere near accounted for in many cases with the marginal CPU clock difference and GPU performance differential.

I dont know how efficient RDNA2 is with bandwidth or if Sony figured it was the sweet spot, but PS5 seems to follow a similar ethos(although perhaps not as bad as Pro which was super bandwidth starved from what i hear)

The best thing sony could do right now for PS5 externally is to increase memory bandwidth...if they choose to go that path its also lucky for them that its one of the easiest things they could do to the HW before launch as well. Hopefully they decide to go that route

 

[mrcorbo]

no, it’ll be like previous gens with Sony, when a new console launches it’ll have proper exclusives that take advantage of the new hardware and not held back by needing to run on 1TF GPU with a shit slow HDD.

Or are you saying Sony won’t have PS5 exclusives that take full advantage of the new hardware (like MS are doing?)

I disagree with the idea that you can't have cross-gen games that "fully take advantage" of the new hardware. Cross-gen does not have to mean parity. We've even seen the opposite in the past where next-get versions of cross-gen games were cut down compared to the prior-gen version. Fortunately, I don't expect to see that during this transition.

I understand that you could certainly create a game that fundamentally relies on an upgraded hardware spec that can't deliver even a down-graded experience on older hardware, but that's not going to create an inherently superior game, just a different one. I am eager to see the first example that delivers on both, but that won't preclude me from appreciating or enjoying scalable experiences simply because they are scalable.

 

[PSman1700]

no, it’ll be like previous gens with Sony, when a new console launches it’ll have proper exclusives that take advantage of the new hardware and not held back by needing to run on 1TF GPU with a shit slow HDD.

Or are you saying Sony won’t have PS5 exclusives that take full advantage of the new hardware (like MS are doing?)

That would be a first for sony, as for every gen it mostly where tech demos (order 1886) or just games that won't make any impression both graphics and gameplay wise (shadowfall, FC3 looks better). I rather have corporations put their resources and time into proper AAA titles, most don't have access to these new machines anyway at launch, which make it a niche market.

You're also forgetting scaling, which is something that didn't exist in the scope we know today.

 

[iroboto]

Exactly. Bandwidth is the main killer of PS4Pro..its easy to see because the deltas we see between Pro and X are not anywhere near accounted for in many cases with the marginal CPU clock difference and GPU performance differential.

I dont know how efficient RDNA2 is with bandwidth or if Sony figured it was the sweet spot, but PS5 seems to follow a similar ethos(although perhaps not as bad as Pro which was super bandwidth starved from what i hear)

The best thing sony could do right now for PS5 externally is to increase memory bandwidth...if they choose to go that path its also lucky for them that its one of the easiest things they could do to the HW before launch as well. Hopefully they decide to go that route

I'm not so sure if Sony can do that unfortunately. I think bus widths are locked. Capacities can change, but I don't think the bus width can. I know some folks will refer to the clamshell arrangement, but I believe that is a way to use the existing bus to increase capacity via more chips, but you aren't allowed to increase the bus width as a result. If I understand correctly, and feel free to correct me, with GDDR5 I think has 32 bit going to each chip, they can split that into two chips of 16 bits and double the memory capacity with clam shell mode, without having to buy higher capacity chips. This is what allowed PS4 to go from 4->8 GB of memory, but I think the bandwidth was always the same.

tldr; I think clam shell allows for augmentation of a data bus, but not the address bus.

 

[Inuhanyou]

I'm not so sure if Sony can do that unfortunately. I think bus widths are locked. Capacities can change, but I don't think the bus width can. I know some folks will refer to the clamshell arrangement, but I believe that is a way to use the existing bus to increase capacity via more chips, but you aren't allowed to increase the bus width as a result. If I understand correctly, and feel free to correct me, with GDDR5 I think has 32 bit going to each chip, they can split that into two chips of 16 bits and double the memory capacity with clam shell mode, without having to buy higher capacity chips. This is what allowed PS4 to go from 4->8 GB of memory, but I think the bandwidth was always the same.

tldr; I think clam shell allows for augmentation of a data bus, but not the address bus.

People were talking about switching 14GBPS chips to 16GBPS chips...your saying that's not possible?

 

[BRiT]

People were talking about switching 14GBPS chips to 16GBPS chips...your saying that's not possible?

How expensive do you want that console to be?

 

[iroboto]

People were talking about switching 14GBPS chips to 16GBPS chips...your saying that's not possible?

It is. Technically possible. They would move bandwidth to 512 gb/s.

but then we get into the business side of the equation. Price, supply and availability. Re-negging on an existing contract for the slower memory etc. Or perhaps they built in a clause to upgrade. Who knows. That would be awesome, a bit dishonest to do that, but awesome.

I mean for the sake of internet drama I would love to see it happen. Unfortunately I think it’s very slim. Sony is more likely to invest their time and effort elsewhere.

 

[Inuhanyou]

How expensive do you want that console to be?

I..i'm not sure.

I thought 500 bucks off the bat was essentially what we would have to contend with anyways...and i dont know how much Sony is will to eat on the console upfront with subscriptions and such...

I'll admit i dont know how much it would cost to do that, but if its something even remotely feasible from a logistics perspective...

Then again, i am aware one could make the argument that they are going to put out a Pro model at the halfway point again anyways so it doesnt matter, but i dunno..

 

[Sonic]

How expensive do you want that console to be?

I'm willing be to spend an extra $1 per 2 GB/s of bandwidth.

 

[MrFox]

14gbps allows them to multi source right now, from everyone.

It will also allow a power reduction sooner with lower voltage chips at 14gbps, while 16gbps would possibly require an additional memory node to make it feasible.. GDDR6 won't be a negligible amount of power, the slim revision might be able to drop to 1.25v at 14gbps if by then the top bin is 18gbps (or 20gbps).

 

[PSman1700]

I'm willing be to spend an extra $1 per 2 GB/s of bandwidth.

Why?

 

[iroboto]

Not sure if this will ever be useful. Just dumping.

https://twitter.com/i/web/status/1257381972831338496

 

[Sonic]

Why?

To be less bandwidth limited.

 

[Deleted member 13524]

Exactly. Bandwidth is the main killer of PS4Pro..its easy to see because the deltas we see between Pro and X are not anywhere near accounted for in many cases with the marginal CPU clock difference and GPU performance differential.

All I remember hearing about a bandwidth deficit on the PS4 Pro was regarding the theoretical utilization of all the 64 ROPs in the GPU, but I don't know if by lifting memory bandwidth the GPU wouldn't become compute shader limited.
Looking at Polaris 10/20 cards there seems to be a constant ~40-45 GB/s-per-TFLOPs. The PS4 Pro has 50. Take away whatever's needed for the CPU and the values should match.

It is. Technically possible. They would move bandwidth to 512 gb/s.

but then we get into the business side of the equation. Price, supply and availability. Re-negging on an existing contract for the slower memory etc. Or perhaps they built in a clause to upgrade. Who knows. That would be awesome, a bit dishonest to do that, but awesome.

I mean for the sake of internet drama I would love to see it happen. Unfortunately I think it’s very slim. Sony is more likely to invest their time and effort elsewhere.

IMO the only factors standing in the way of the PS5 adopting 16Gbps GDDR6 is a) either supply amount and/or b) how much more power the MCU would take for driving the faster memory chips, considering how Sony wants the GPU to take in as much power as possible.

Sony is in a position that allows them to cancel whatever contracts they want to, and it's not like GDDR6 makers will have any trouble finding other buyers for 14Gbps chips in 2020 / 2021.

There's also c) we don't know if 256bit GDDR6 is all there is in the chip. We know developers only have access to those 16GB GDDR6, but a 64bit DDR4 pool exclusive to the CPU for OS tasks isn't out of the realm of possibilities for me, yet.

 

[MrFox]

"32 ALUs local to shared memory, allowing some compute as a free byproduct of moving data in/out of LDS"

An amiga blitter if I've ever seen one!

 

[iroboto]

IMO the only factors standing in the way of the PS5 adopting 16Gbps GDDR6 is a) either supply amount and/or b) how much more power the MCU would take for driving the faster memory chips, considering how Sony wants the GPU to take in as much power as possible.

Sony is in a position that allows them to cancel whatever contracts they want to, and it's not like GDDR6 makers will have any trouble finding other buyers for 14Gbps chips in 2020 / 2021.

You may be right. Though I know you were not a fan of the GitHub times, @Proelite gave me a nod that the GitHub leaked testing bandwidth values for Oberon were in fact 512 GB/s indicating a 16gbps chip, and also a second bandwidth number of 528 GB/s, it would appear the technical parts are achieved already (at least with respect to at least a devkit), Sony could make a switch here last minute if they felt they needed to, it would appear that either Sony downgraded to 14gbps, or the devkits are intended to run with the 16gbps chips.