PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)

[Grall]

Although it's a bit weird him saying opaque shadowmaps use vertex shaders on a unified shader architecture.

When he says vertex shaders (for simplicity's sake), just imagine he said vertex shader programs or somesuch, if you're feeling extra pedantic today.

The ALUs are still going to be tied up processing the workload.

I think his point was that shadow map rendering isn't going to load the ALUs very heavily as most of it is just rasterization. Some (or rather, a little) is going to be vertex processing, but not very much, compared to the full capability of the shader array...

Any dev on this board with experience of a US architecture will be well capable to tell us exactly how much idle time the GPU experiences

That would be interesting, but seeing as today's consoles use GPUs with much more limited capabilities than the next gen will I'm not sure how well the numbers would compare. After all, while xenos (xenon? shit, I can never tell 'em apart) can be used for GPGPU it's not very capable (and not optimized for fast switching between jobs in the middle of a frame), and the 7800GTX series in the PS3 reportedly can't do compute at all. So any frame time these chips' shader arrays are idle it's basically just wasted anyway.

 

[Scott_Arm]

Interested in the culling technique described

"This technique, he said, is "a mix of hardware, firmware inside of the GPU, and compiler technology. What happens is you take your vertex shader, and you compile it twice, once as a compute shader, once as a vertex shader. The compute shader does a triangle sieve -- it just does the position computations from the original vertex shader and sees if the triangle is backfaced, or the like. And it's generating, on the fly, a reduced set of triangles for the vertex shader to use. This compute shader and the vertex shader are very, very tightly linked inside of the hardware."

So they're culling triangles before vertex operations. OpenGL and DirectX pipelines do this after after vertex and tesselation, and before (during) fragment. It seems to me you'd want to eliminate as many vertices as possible before running your vertex operations, which seems to be the case here. Why is that not the standard way of doing it? There must be some reason OpenGL and DirectX pipelines are set up the way they are.

Overall, the interview wasn't too technical, but it seems like Sony have done a good job with their hardware.

Edit:
I'm guessing you can't cull before performing vertex operations because of shadows, or other things that would require the full "world" data, for lack of a better term on my part.

 

[3dilettante]

Suspend state can save the game progress by leaving the gddr5 ram pool powered while the APU is shut down. --- How power hungry will that be? Can they possibly downlock gddt5 ram down enough to meet "green" power laws of EU?

I don't think the full pool can be powered below something like the the .5W standby state, but the regulations make distinctions between that low-power state and non-active states that maintain machine state.

The power levels for that are allowed to be higher than standby.

 

[MrFox]

Another thing the PlayStation 4 team did to increase the flexibility of the console is to put many of its basic functions on dedicated units on the board
[..] For example, by having the hardware dedicated unit for audio
[..] The same thing for compression and decompression of video.
[..] the system also has a unit to support zlib decompression

Does that mean the audio, video and zlib units are on separate chips, not inside the SoC?

There's also another custom chip to put the system in a low-power mode for background downloads. "To make it a more green hardware, which is very important for us, we have the ability to turn off the main power in the system and just have power to that secondary custom chip, system memory, and I/O -- hard drive, Ethernet. So that allows background downloads to happen in a very low power scenario. We also have the ability to shut off everything except power to the RAMs, which is how we leave your game session suspended."

That kind of confirm the south bridge doesn't have it's own memory, it can use the GDDR5 (maybe they throttle it really slow?).

"One thing we could have done is drop it down to 128-bit bus, which would drop the bandwidth to 88 gigabytes per second, and then have eDRAM on chip to bring the performance back up again," said Cerny. While that solution initially looked appealing to the team due to its ease of manufacturability, it was abandoned thanks to the complexity it would add for developers.

It's interesting that they considered what is rumored to be in the 720, but the reason they dropped it is not related to performance. It also hints that using eDRAM could have possibly cost less for a similar performance. (assuming "ease of manufacturing" == "better yield, smaller chip")

 

[onQ]

It confirms a lot, including the compute rings being custom, simultaneous compute and graphics with almost however compromise you want, the RAM not being on the chip (thus questioning if we can still call it APU) and A LOT more...

Since when did a APU have to have RAM on the chip to be called an APU?

 

[Xenus]

Not only that but it sounds like if they were going to go that route they would of provided an order of magnitude more bandwidth then the 720 is rumored as he is talking terabytes. That could of had some downright interesting applications.

 

[patsu]

PlayStation 4: The complete tech breakdown from Mark Cerny

Huge amount of info.

Ph*ck ! On one of my busiest days. Will come back later.

Excellent interview.


OS will automatically cache the Bluray games on HDD when the game is not requesting contact to the BD drive. ---- Will those caches be permanent or will they be overwritten in time?

Suspend state can save the game progress by leaving the gddr5 ram pool powered while the APU is shut down. --- How power hungry will that be? Can they possibly downlock gddt5 ram down enough to meet "green" power laws of EU?

Cache means temporary.

 

[MrFox]

Cache means temporary.

But Cerny didn't use the word "cache" he said after an hour or two the game is on the HDD. I sure hope it's manageable, I don't want to waste the space used by FMV and cut-scene stuff (there's no reason to cache that). Maybe I'm being paranoid, he seem to imply it always copies the whole disk?

 

[Grall]

If it stops the optical drive from spinning up and making noise, why wouldn't you want the whole disc cached?

 

[Gradthrawn]

At least now we finally have an official name (albeit a branded, non technical, name) for the progressive download/install system we've been discussing (dubbed PlayGo by Sony).

At the New York City unveiling of the system, Cerny talked about PlayGo, the system by which the console will download digital titles even as they're being played.

"The concept is you download just a portion of the overall data and start your play session, and you continue your play session as the rest downloads in the background," he explained to Gamasutra.

However, PlayGo "is two separate linked systems," Cerny said. The other is to do with the Blu-ray drive -- to help with the fact that it is, essentially, a bit slow for next-gen games.

"So, what we do as the game accesses the Blu-ray disc, is we take any data that was accessed and we put it on the hard drive. And if then if there is idle time, we go ahead and copy the remaining data to the hard drive. And what that means is after an hour or two, the game is on the hard drive, and you have access, you have dramatically quicker loading... And you have the ability to do some truly high-speed streaming."

Looks like they're aiming to maintain the same level of "immediate" play from BD as what's available now (I use immediate loosely). Meaning playing directly from the disc (but apparently caching everything transparently in the background). I think that was confirmed in another recent interview as well.

And it appears that game suspend will simply keep the last memory state active and powered and not use any snapshot boot tech as was previously speculated. Seems like the simplest and most responsive method to accomplish their goal of immediacy.

 

[babcat]

That was a horrible interview, imo. There were many questions that were not asked!

a. What are the specs of the ARM chip that handles background downloading?

b. Does the ARM chip contain any type of embedded graphic functions?

c. Did they add any extra ALUs to the GPU or any other units?

d. Can he confirm, once and for all, that there is no bespoke graphics processor hidden in the system?

e. With eight gigs of RAM how many gigs of texture data does he think the GPU can actually utilize in a frame?

There are these and many other questions should have been asked, but they were not. I wish I had ten minutes to talk with Cerny -- I would have done a much better interview.

 

[ZiGgY]

"One thing we could have done is drop it down to 128-bit bus, which would drop the bandwidth to 88 gigabytes per second, and then have eDRAM on chip to bring the performance back up again," said Cerny. While that solution initially looked appealing to the team due to its ease of manufacturability, it was abandoned thanks to the complexity it would add for developers.

Lol..slight dig at MS methinks and where they think the Xbox will try and match the Ps4.

 

[anexanhume]

That was a horrible interview, imo. There were many questions that were not asked!

a. What are the specs of the ARM chip that handles background downloading?

b. Does the ARM chip contain any type of embedded graphic functions?

c. Did they add any extra ALUs to the GPU or any other units?

d. Can he confirm, once and for all, that there is no bespoke graphics processor hidden in the system?

e. With eight gigs of RAM how many gigs of texture data does he think the GPU can actually utilize in a frame?

There are these and many other questions should have been asked, but they were not. I wish I had ten minutes to talk with Cerny -- I would have done a much better interview.

It was very much a directed interview, by them. They spouted the facts they wanted to clap themselves on the back for. There doesn't seem to be any of the critical questions this community (and others) had following the event and leaks.

Still, it was nice to hear about the compute focus and vertex mitigation techniques in their own words to confirm the existing understanding.

 

[Gradthrawn]

That was a horrible interview, imo. There were many questions that were not asked!

a. What are the specs of the ARM chip that handles background downloading?

b. Does the ARM chip contain any type of embedded graphic functions?

c. Did they add any extra ALUs to the GPU or any other units?

d. Can he confirm, once and for all, that there is no bespoke graphics processor hidden in the system?

e. With eight gigs of RAM how many gigs of texture data does he think the GPU can actually utilize in a frame?

There are these and many other questions should have been asked, but they were not. I wish I had ten minutes to talk with Cerny -- I would have done a much better interview.

No offense, but by and large those sound like Console Warrior questions along with random speculation that few beyond said warriors care about. I would hardly call the interview terrible for not addressing those specific points, but that's just me. They should have asked about the super secret PLOPs acceleration units, though.

 

[babcat]

No offense, but by and large those sound like Console Warrior questions along with random speculation that few besidse said warriors care about. I would hardly call the interview terrible for not addressing those specific points, but that's just me.

I just want all the rumors put to rest and to know ALL of the specs of the system so I can go into hibernation until the PS5 comes out.

 

[babcat]

It was very much a directed interview, by them. They spouted the facts they wanted to clap themselves on the back for. There doesn't seem to be any of the critical questions this community (and others) had following the event and leaks.

Still, it was nice to hear about the compute focus and vertex mitigation techniques in their own words to confirm the existing understanding.

Actually, I would not have agreed to such an interview, if I could not ask questions. For goodness sakes, the answers to the questions that really needed to be answered would not have taken long to provide. Some of them could have been answered with simple one sentence statements. I wish someone here who is in contact with Sony would ask for an interview with someone who knows the specs of the PS4.

One interview could put almost all the speculation and rumors to rest.

 

[Gradthrawn]

I just want all the rumors put to rest and to know ALL of the specs of the system so I can go into hibernation until the PS5 comes out.

Ah, I see. Then what you probably want is Dev level documentation, in which case I suggest you contact VGLeaks and see if they'll share with you their raw documentation (unlikely they'll provide it at random or someone would probably already have shared all of it by now). You could probably also look into PSN self publishing and see what documentation you can get a hold of from there (absent licensing fees and what not).

Below are some guidelines and contacts for PSN self-publishing:
SCEA (America): Devrel-scea@playstation.sony.com
PubFund: Ted_Regulski@playstation.sony.com

What's required in a PubFund pitch:
1. Brief design doc (1-3 pages)
2. Budget and schedule
3. A build is great, though not required
4. They usually like to see a build on Playstation hardware before signing anything, though PC/tablet builds have been approved in the past

MAJOR TIP: Take advantage of hardware in a unique, crazy way!

SCEE (Europe): Devrels_support@scea.net
Market covers 32 countries, and is expanding into India, Middle East, and Asia, where Vita sales are blowing up. Content does not have to be localized, but metadata does, in eight European languages.

Steps to publish:
1. Email address above to start registration process
2. Sign agreements
3. Develop game on PS hardware
4. Submit metadata and code
5. Goes live 3 weeks after final QA approval
6. Monthly payouts, no minimum

Top SCEE genres (in order):
Shooter, Sport Games, Action, Racing, Family Entertainment

Top SCEE countries (in order):
UK, France, Germany, Australia, Spain, Italy

SCEJ (Japan): bradley_douglas@hq.scei.sony.co.jp
SCEJ has featured space for non-Japanese games, which helps western developers gain visibility. Content localization is not required, but highly recommended. The SCEJ QA process is stricter than its western counterparts. PSP sales are great in Japan, so consider releasing or porting to PSP in this region.

Steps to publish:
1. Get licensed with SCEA or SCEE
2. Email address above to start registration process
3. Sign agreements
4. Get rated with CERO (requires some Japanese language, companies exist to help English developers through this process)
5. Have a way to support Japanese-speaking users (website or email)

Top SCEJ genres (in order):
RPG, Action, Rhythm Action, Simulation, Adventure

 

[babcat]

Ah, I see. Then what you probably want is Dev level documentation, in which case I suggest you contact VGLeaks and see if they'll share with you their raw documentation (unlikely they'll provide it at random or someone would probably already have shared all of it by now). You could probably also look into PSN self publishing and see what documentation you can get a hold of from there (absent licensing fees and what not).

All the information I want could be obtained in a ten minute conversation with Cerny. If I could get ahold of Developer level documentation that would be great, but I don't think I need all of it. I mean, for example, why can't Cerny just provide a simple set od specs for the ARM chip? For example, this is a hypothetical answer...

600 mhz ARM Cortex 15 with 4 core rogue power VR embedded GPU at 300mhz and 128mb of ddr3 ram.

 

[Deleted member 13524]

http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php

This one?

 

[bkilian]

All the information I want could be obtained in a ten minute conversation with Cerny. If I could get ahold of Developer level documentation that would be great, but I don't think I need all of it. I mean, for example, why can't Cerny just provide a simple set od specs for the ARM chip? For example, this is a hypothetical answer...

600 mhz ARM Cortex 15 with 4 core rogue power VR embedded GPU at 300mhz and 128mb of ddr3 ram.

Why would he? It's part of the system irrelevant to gamers, and irrelevant to devs, other than the API they use to interact with it.

If it's a 400MHz ARM, or a 600MHZ ARM, will that make _anything_ in the system different for the purpose of making games? No. I would be very surprised if the developer documentation said anything about this chip other than it exists, and how you interact with it.

The only reason the forum warriors are interested in the ARM chip is because they believe it will house the full OS and allow Sony to completely allocate all the main resources to the game. If it did something like this, it would have been called out in the reveal or the interview. The guy talked about the number of compute queues, for goodness sake.