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

[mrcorbo]

I don't think so, that was just speculation that maybe it needs more than 5.5 to compensate variations in controller implementations, we have no idea how much.

And the storage expansion doesn't need that many priority channels if those related to the OS are only present on the primary drive. The second drive is used exclusively for the game.

Apparently it does. Here's the relevant clip starting right at the beginning of the 3rd party drive portion. The needing "a little bit" more performance due to the lack of priority levels on NVMe is discussed shortly in.

Mr. Cerny sure does love his vagueries, doesn't he? "A little bit", "Most of the time", etc.

 

[psorcerer]

They're expensive now but within a few months these will be supplanted by the next gen drives based on the upcoming PS5018-E18 controller capable of 7GB/s so should drop considerably in price. Lexar is the fist vendor to start promoting it's upcoming 7GB/s drives:

I don't see any 7GB/sec there.
It's 639.9MB/sec on the screenshot.
SFS/DXS pages are 64K though, so it's not clear what the real speed will be.

 

[Shifty Geezer]

Mr. Cerny sure does love his vagueries, doesn't he? "A little bit", "Most of the time", etc.

Maybe it's a marketing ploy to generate discussion?

 

[pjbliverpool]

I don't see any 7GB/sec there.
It's 639.9MB/sec on the screenshot.
SFS/DXS pages are 64K though, so it's not clear what the real speed will be.

7GB/s sequential read which is an equivalent form of measurement to the PS5's 5.5GB/s raw throughput. In Cernys own words, these drives have "a little bit more [raw] performance" than the PS5's native drive. That's not to say the end result will be faster when the rest of the system is taken into account.

 

[psorcerer]

7GB/s sequential read which is an equivalent form of measurement to the PS5's 5.5GB/s raw throughput.

Not really.
Unless these too have a custom FS with the block index fully in SRAM.
Which I doubt.

 

[Deleted member 11852]

Mr. Cerny sure does love his vagueries, doesn't he? "A little bit", "Most of the time", etc.

True, but you can't be more specific without actual specifics. I mean he can't say more than 95% of the time unless he had data to back that up. It is vague, but more honest that making up statistics.

 

[Kaotik]

Not really.
Unless these too have a custom FS with the block index fully in SRAM.
Which I doubt.

So you're suggesting Cerny was BSing when he said PS5 games could be installed on separate SSD without issues when they're quick enough, with PCIe4 SSDs going to cap at about 7GB/s?

 

[mrcorbo]

True, but you can't be more specific without actual specifics. I mean he can't say more than 95% of the time unless he had data to back that up. It is vague, but more honest that making up statistics.

True. But in some of these cases they surely have "some" data and are choosing not to present it even as "preliminary" or "early" results. Meanwhile MS are showing games running and allowing FPS to be shown. The difference is striking. I'm not criticizing, BTW, or trying to paint one approach as better or worse. I'm just struck by how different the reveal strategies are.

 

[chris1515]

So you're suggesting Cerny was BSing when he said PS5 games could be installed on separate SSD without issues when they're quick enough, with PCIe4 SSDs going to cap at about 7GB/s?

I think he talks about the real speed of this drive out of PS5 before DirectStorage release.

 

[mrcorbo]

So you're suggesting Cerny was BSing when he said PS5 games could be installed on separate SSD without issues when they're quick enough, with PCIe4 SSDs going to cap at about 7GB/s?

I mean, that's as fast as drives will ever be on that interface. So, if it's insufficient it would have been grossly incompetent to add it as an option for expansion. When evaluating the PS hardware design team, "grossly incompetent" is not a phrase I would apply to them, so I'd fully expect their solution to work fine.

 

[Deleted member 11852]

True. But in some of these cases they surely have "some" data and are choosing not to present it even as "preliminary" or "early" results. Meanwhile MS are showing games running and allowing FPS to be shown. The difference is striking. I'm not criticizing, BTW, or trying to paint one approach as better or worse. I'm just struck by how different the reveal strategies are.

I don't know any more than you, but as somebody who has worked in several engineering roles, when I have been in the position of stating facts related to similar issues I have never provided statistics without a significant body of evidence. As you say there is the option to go with "preliminary" or "early" but Mark Cerny chose not too. He's never been a person to be drawn into speculative engineering throughout decades of presentations on hardware and software.

 

[mrcorbo]

I don't know any more than you, but as somebody who has worked in several engineering roles, when I have been in the position of stating facts related to similar issues I have never provided statistics without a significant body of evidence. As you say there is the option to go with "preliminary" or "early" but Mark Cerny chose not too. He's never been a person to be drawn into speculative engineering throughout decades of presentations on hardware and software.

He did say things, though, that must have had data behind them to back them up. He just chose not to be precise with the things he said.

Edit: In fact, how do you even determine the clocks that will hit those power limits "under normal operation" without massive amounts of data to drive that decision?

 

[BRiT]

True, but you can't be more specific without actual specifics. I mean he can't say more than 95% of the time unless he had data to back that up. It is vague, but more honest that making up statistics.

So how can anyone even say "most of the time" when they don't have data?

EDIT: I see @mrcorbo is saying what I was thinking too.

 

[psorcerer]

So you're suggesting Cerny was BSing when he said PS5 games could be installed on separate SSD without issues when they're quick enough, with PCIe4 SSDs going to cap at about 7GB/s?

Dunno. Is there a PC-style block storage controller in PS5?

 

[psorcerer]

Edit: In fact, how do you even determine the clocks that will hit those power limits "under normal operation" without massive amounts of data to drive that decision?

So for example the lowest clock is 800Mhz (which is probably the case). Then Cerny should have said that. Am I right?

 

[manux]

Dunno. Is there a PC-style block storage controller in PS5?

Cerny said the additional ssd has to be somewhat faster than the ps5 internal ssd. PS5 io-controller will still handle priorities of access ,decompression, dma etc. and issue commands to extra ssd. There is some amount of inefficiency leading to that extra ssd having to be bit faster than the ps5 internal ssd.

How much faster, we don't know. Luckily sony works with vendors and will certify drives that pass the bar.

 

[Deleted member 11852]

So how can anyone even say "most of the time" when they don't have data?

Data =/= a significant body of data. You can engineer something and do testing and conclude it works most of the time, but you have to do an exhaustive amount of testing to it works xx.xx% of the time. I would have though Sony would want to be absolutely certain of things like this before throwing such statistics about publicly. It may be something they don't intend to make public, like many other specifics regarding their console's workings.

 

[mrcorbo]

So for example the lowest clock is 800Mhz (which is probably the case). Then Cerny should have said that. Am I right?

No. A graph illustrating a worst-case of clockspeed reduction observed while running an actual game would have, assuming all the claims being made are valid, silenced any criticism. Or at least made that criticism obvious for what it was.

 

[psorcerer]

How much faster, we don't know. Luckily sony works with vendors and will certify drives that pass the bar.

Yep. But we don't really know what the interface between PS5 I/O controller and a certified SSD controller there will be.
And surely it will not be the same as the PC-oriented interface.

No. A graph illustrating a worst-case of clockspeed reduction observed while running an actual game would have, assuming all the claims being made are valid, silenced any criticism. Or at least made it obvious for what it was.

I'm not sure. Is there any graph of freq/temp/power usage for XBSX?
Why nobody whines about that?

 

[Deleted member 11852]

Luckily sony works with vendors and will certify drives that pass the bar.

Hang on, did Sony say that? I don't remember Sony saying that.