Playstation 5 [PS5] [Release November 12 2020]

Also, I called it on the 3D audio as pretty evident given the acquisition of Audiokinectic. XBSX has no audio processor? I guess that might be a point of comparison.
PS5s audio is based on modified AMD CU (not specified which specific version) with caches ripped out at least. I'm pretty sure XSX can sacrifice a CU or two for audio if needed, unless it has some audio DSP solution (or to add on top of it), there should be no reason they couldn't use TrueAudio Next if they wanted which does exactly that.
 
Despite this presentation being "made for developers" there's a lot of information still missing and unconfirmed.
What about Variable Rate Shading, Mesh Shaders and DirectML?
Micosoft gaves us more clear information in their blog posts: https://news.xbox.com/en-us/2020/03/16/xbox-series-x-glossary/
Is this because Microsft had an advantage over Sony by using DirectX?
I'm mostly wondering whether their "primitive shaders" are just that, introduced in Vega, or just different word for the more advanced mesh shaders
 
Ye, could be anything. Between 9 and 10TF it is.

Double post somehow.

Well if it's 10% power reduction via 2% clock speed reduction, the only one that correlates directly to performance would be the latter. So it would be a rough bottom of 2.18GHz & 10.07TF (GPU) or 3.43GHz (CPU) , so perhaps these are just minor adjustments at the top-end so that you can prioritize CPU<>GPU workloads in a given situation, or they could be more dependent on the nature of the workload itself or even in the case of the CPU, the instruction types used (as he may have alluded to in regards to 256-Bit instructions in the presentation). I guess we won't really know until devs with hands-on experience start divulging info.
 
I don’t know about RDNA 2 but game clocks on Navi based parts are 100-150mhz from boost clocks. AMD game clock ratings are based on a panel of 25 popular games and ratings are conservative so that the majority of games will run in between game clocks and boost clocks.
 
Well if it's 10% power reduction via 2% clock speed reduction, the only one that correlates directly to performance would be the latter. So it would be a rough bottom of 2.18GHz & 10.07TF (GPU) or 3.43GHz (CPU) , so perhaps these are just minor adjustments at the top-end so that you can prioritize CPU<>GPU workloads in a given situation, or they could be more dependent on the nature of the workload itself or even in the case of the CPU, the instruction types used (as he may have alluded to in regards to 256-Bit instructions in the presentation). I guess we won't really know until devs with hands-on experience start divulging info.

We have no idea, that's probably one of the reasons to have this boost clock, to be able to claim 10TF. At closer to 2Ghz we would have had in the 9TF range.
 
The point of this technology is that the lowest speed the system runs at is higher than it would have been with conventional methods and much more predictable and consistent, so easier to optimize for etc versus when this would have been governed purely by temperature. Or am I misunderstanding something here?
 
When the stream started and Jim Ryan was stood there, stating that "unfortunately [they] had to cancel this year's GDC talk" I almost threw my telly out of the window. Then he introduced Cerny and I stopped frothing at the mouth.

Pretty disappointing really. They haven't ballsed it up to the extent that I won't be buying it, but it seems to be more of a PS4Pro than a PS4.

I'm not keen that they've gone cheap, but given that we're on the verge of another global recession, maybe going cheap will work in their favour. If so, we have confirmation that Sony are responsible for COVID-19.

SSD

The SSD tech is welcomed. It bodes well that he mentioned it taking roughly half a second to spin the camera, and that being adequate to feed memory ~5GB of new data. I can see that being taken advantage of to realise some very dense, very varied environments and enemies.

I'm curious to see how the Kraken decompression block fares, given that he stated it can decompress up to 20GB of data depending on how well it compresses. Specifically, I'm curious to see just how often that best case scenario comes to pass. I reckon it'll be somewhat similar to the performance gains of RPM, in that around 10% of textures will compress quite that well.

825GB strikes me as an odd number, but the fact that it'll be easily expandable means I don't much care. I upgraded the storage on my PS3 and PS4, and by the time I got my PS4Pro, I was able to plug in an external 4TB HDD. Hopefully the M2 drives will be pretty much hot swappable. I quite like the idea of having a couple of different drives with my more played games on one, and less played games on another. Most played on internal SSD of course.

Audio

I like playing games with surround sound headphones and I'm pretty happy that it's going to be a hardware feature. Hopefully they'll let me use any headphones I want, not just some PlayStation branded ones. Time will tell.

I don't quite get why these HRTF profiles are something to give a toss about. Everyone has different ear canals etc yes, but they're also the ones that you use every moment of your life. You're used to hearing with them, used to positioning invisible sources with them, and I struggle to believe that your brain can't make sense of 3D audio because you haven't let the Babel fish's less useful cousin have a rummage around.

I've yet to try the tech out, so maybe I'm talking out of my arse.

Backwards compatibility

They've tested the top 100 most played games, and they expect most to be playable at launch. What the fuck does that mean? Are they classifying 51 out of that 100 as "most?"

Hopefully any issues are minimal and easily patched out. We're looking at a minimum of 7 months before launch, so I would hope that Sony are working with developers to iron out any troubles.

Or maybe it's just a legal thing, and all games seem to work fine most of the time, but they can't risk instances of mis-selling.

I'm curious about the custom block for audio processing, given that Cerny was comparing it to the Cell's SPE's. Could it be the case that this customised CU (or customised CU's) have scope for emulating the SPE's of the Cell? Maybe not to the extent of just inserting your disk, but more akin to the XBoxOne method?

POWER

It doesn't really deserve to be capitalised, as it really isn't all that powerful. Maybe. Cerny seemed to choke a bit when stating that CU count and TF count aren't all that important. Like he didn't believe it himself. And why would he? He's an engineer (honest and precise) not a salesman (deceitful and manipulative.)

The 2.23GHz GPU clockspeed is all well and good, but the variability makes me a little nervous. It's probably the best generation ever in which to take this approach, given the abundance of dynamic resolution scaling that we've seen take off in the current generation, but I'm still unsure of what this will mean in practice.

We'll see, but I don't want to get one of the shit ones, and be stuck with a console that only reliably hits 2.1GHz at the same rate that an ideal console hits this touted 2.23GHz. There are plenty of unknowns, and so I don't want to jump to any conclusions, but my gut says that I would've preferred 40CU's at a reliable 2GHz.

44CU's per chip, with 4 deactivated for yields, then leaves open the possibility - with otherwise defective chips - the combination of deactivating a further 4 CU's and lowering the clockspeed to PS4Pro levels, for use in PSNow data centres.

If this results in a PS5Pro in a few years' time, and they can take the same approach of butterfly wings for the GPU, and clocking both halves higher, then I can accept this as a smart decision: games will be built around a dynamically changing GPU frequency, so will automatically see resolution/quality boosts (or resolution/quality stability.) Right now though, it seems a little bit lacking.

It's the bandwidth that disappoints me most though. I'm an HBM fanboy, so high bandwidth makes my nipples tingle. And for Sony to go with quite pedestrian GDDR6 for a meagre 448GB/s bandwidth bothers me. The 528GB/s in the GitHub gospel was a lot more appealing. It is quite funny though that the only appealing part of the gospel is the part that didn't come to pass.

So with Sony PS5 having ability for consumer M2 NVME drives that meet as yet to be determined threshold, will those too suffer from the heat related slowdowns that DF covered on the SeriesX deep-dive? If so, what will be their throughput after hitting the slowdowns?

Given that, as he was discussing the matter, the screen next to him displayed the example of 7GB/s, I suspect that such speeds will be required for compatible drives, which will then run below their bandwidth (and therefore temperature) limit.
 
Then how is nobody going to design a game with larger, better, faster loading worlds if you can play from an slower SSD?.
No, the SSD in the expansion bay must be as fast as the native SSD. The HDD can be used for run PS4 games from and, my hope is, that it can be used to store (or backup if you prefer) PS5 games that you have but won't fit in the solid state storage. All indications are PS5 games must be run from SSD and that SSD has to be as fast as Sony's option.
 
If it basically ran with boost mode all the time there would not be a boost mode.

There are a lot of things that raise concern. In the presentation Cerny mentioned a few times that the developer should be aware how he uses the hardware. Then there is boost, variable frequency and temperatures. Then there are the clear disadvantages compared to X. And also no mention of the various other technical features that help efficiency on the X.
The PS5 sounds like a weaker console that is also less clear cut for the developer to reach maximum performance.

As far as I can tell, going of eurogamer, this is the "boost" implementation,

https://www.amd.com/en/technologies/smartshift


Another thing, with the PS5, will we get expension cards like the xsx will offer?
Why do they need that when adequate NVMEs can be used?
 
Despite this presentation being "made for developers" there's a lot of information still missing and unconfirmed.
What about Variable Rate Shading, Mesh Shaders and DirectML?
Micosoft gaves us more clear information in their blog posts: https://news.xbox.com/en-us/2020/03/16/xbox-series-x-glossary/
Is this because Microsft had an advantage over Sony by using DirectX?
Everything about Sony's PS5 has felt rushed and not all there tbh.

Compared to MS, Sony hasn't shown the physical console while MS already has people with hands on videos and taking it apart.
MS has shown demos running on the console and demoed many of the advantages over last gen while Sony hasn't shown anything.
MS has a huge advantage presented with their back compat solution being improvement on the old game and saying things will work, Sony's message is most popular games will probably work..
MS has released much better detailed specs while Sony's technical presentation has very little in the technical side with lots of unanswered basic questions.
 
It will definitely be interesting to see how they manage this dynamic clock situation but doesn't seem good on the face of it. Basically what they are saying is that as more of the actual GPU or CPU subcomponent are engaged, thereby increasing power draw, they have to lower clocks to keep max power draw constant. If the mix of operations and instructions are such that not all subcomponents are engaged, then they can let things run at their max clocks. For example, if you were to create a GPU benchmark that could enable maximum parallelization across the available ALU's then it would never actually be able to hit 10TFLOPs since it would have to slow down to limit power consumption. What Cerny is saying is that most games aren't that efficient, and can't effectively use all of the available hardware compute for various factors; so then they are free to let the clocks ramp up higher. And the theoretical compute power at that frequency becomes a number that the system is not technically capable of. But certainly looks a lot better in a head to head comparison. Very tricky indeed.

The question I have is how frequently is this changing. As a developer, do I choose from a power profile from a predetermined set based on my games compute usage or is it constantly changing? And how do I manage performance consistency in that case?
 
Another thing, with the PS5, will we get expension cards like the xsx will offer?

No, dont think we’ll get over expensive Sony branded cards.

Guess like PS3, PS4.. you can just buy third party M.2 PCIe 4.0 NVMM cards. They only have to be ok’ed by Sony, meaning they are fast enough.

Any card comes on top of the built-in 850 GB.
 
No, the SSD in the expansion bay must be as fast as the native SSD. The HDD can be used for run PS4 games from and, my hope is, that it can be used to store (or backup if you prefer) PS5 games that you have but won't fit in the solid state storage

I'd hope they'd allow for that part and would be a bit surprised if they didn't support an option like that since SeriesX seems to support it.
 
Back
Top