Gamasutra: The Art & Business of Making Gamesspacer
View All     RSS
August 28, 2014
arrowPress Releases
August 28, 2014
PR Newswire
View All





If you enjoy reading this site, you might also want to check out these UBM Tech sites:


Inside the PlayStation 4 with Mark Cerny Exclusive
Inside the PlayStation 4 with Mark Cerny
March 27, 2013 | By Christian Nutt

March 27, 2013 | By Christian Nutt
Comments
    22 comments
More: Console/PC, Programming, Business/Marketing, Exclusive, GDC



Mark Cerny originally began to think about designing the PlayStation 4 in 2007. Over Thanksgiving weekend, a mere year after the PlayStation 3 was released, he began to read technical documents about the X86 processor -- the processor that ended up going into the system that was unveiled this past February, by Cerny, in New York City.

The fact that he spent so much of his personal time working on the question of just what hardware should go into the box made Cerny realize something important: "I probably have more passion about the next generation than anybody inside the Sony Computer Entertainment world."

With that in mind, he pitched his bosses on letting him lead the PlayStation 4 development efforts. To his surprise, he earned himself the role of lead system architect.

Though he began with the technology, "wanting to lead the effort wasn't based on any specific beliefs at that time -- other than that clearly we had had some issues with PlayStation 3, in that a very developer-centric approach to the design of the PlayStation 4 would just make things go more smoothly overall."

"The biggest thing was that we didn't want the hardware to be a puzzle that programmers would be needing to solve to make quality titles," says Cerny. He's referring here to the fact that the CELL processor, which powers the PlayStation 3, was extremely powerful by 2006 standards -- but also notoriously difficult to work with.

So in 2008, once he'd gotten the okay, Cerny began to canvass PlayStation 3 developers, asking them what they wanted from a theoretical next generation console -- yes, that early. "It's not like we could come out and say we were developing the next generation of hardware -- we certainly couldn't say that in 2008," Cerny recollects.

"My first tour of the developers, I had a questionnaire where I just asked them their thoughts on what the next generation might bring," he says. "The largest piece of feedback we got was that they wanted unified memory."

The PlayStation 4 will launch with an 8GB bank of GDDR5 RAM, which can be directly addressed by both the CPU and GPU of the system. Cerny is confident that this strategy brings flexibility and power to the console in both the near and long term.

The system also will ship with an eight-core CPU, another decision that came from the developer-questioning phase. "We quickly could tell that we should put either four or eight cores on the hardware," Cerny says. "The consensus was that any more than eight, and special techniques would be needed to use them, to get efficiency."

"It definitely was very helpful to have gone out and have done the outreach before sitting down to design the hardware," he says.

For all of its commercial shortcomings, the PlayStation Vita marks the first time the company put the software developer at the center of its hardware design efforts, something Cerny says paid off both directly on that system, and also in terms of laying the groundwork for the PS4's design.

"We took Vita as an opportunity to rework the tool chain and the development environment, and I think that you saw that the response from the development community [to those changes] was very good," says Cerny. "That meant that with PS4 we already had this philosophy in place -- that we wanted our tools to be much richer and much more accessible to our developers, even in the launch timeframe."

He didn't stop at game developers, either, he tells Gamasutra. "When I started talking to the development community, prominent middleware companies were in the mix at that time. It's very important to us to have those engines on our platform," Cerny says. "I have to say, also, the insights that you can get by talking to their top technology people -- It's quite nice to have those insights when doing the hardware design."

Throughout his conversation with Gamasutra -- which, in the end, lasted well over an hour -- the two threads that came through again and again were that Cerny wanted the console to be familiar enough that the barrier to entry for developers was very low, but at the same time, he wanted to be sure that the technical decisions he and his team made would ensure high performance over its entire lifespan.

In a forthcoming article, Gamasutra will share the many details of the PlayStation 4's architecture and design that came to light during this extensive and highly technical conversation.


Related Jobs

Square Enix Co., Ltd.
Square Enix Co., Ltd. — Tokyo, Japan
[08.28.14]

Programmers
Monochrome LLC
Monochrome LLC — Aptos, California, United States
[08.28.14]

Senior Programmer
Infinity Ward / Activision
Infinity Ward / Activision — woodland hills, California, United States
[08.28.14]

Build Engineer-Infinity Ward
Cloud Imperium Games
Cloud Imperium Games — Austin, Texas, United States
[08.27.14]

Lead Network Engineer










Comments


[User Banned]
profile image
This user violated Gamasutra’s Comment Guidelines and has been banned.

Freek Hoekstra
profile image
staying with a powerPC architecture wouldn't have helped anyone, and ARM is not an option yet. so going x86 does have some advantages, lets see whether the disadvantages are outweighed though, as the original Xbox had some problems with being a x86 chip (mainly costs)

Now the fact that this is an APU from the start will help costs though (and maybe the amlunt of drawcalls as well!),

The real question is, is the jump from very High end hardware at the time, pretty much as fast as any pc of its day (PS3)
to mid-high end hardware at this time (PS4) going to hurt.

because a 10X increase in processing power is nice, but nothing like the move from PS2 to PS3, and PC's are already as fast as this console is.
again I'm not saying we cannot do great things with this, it will be a marked improvement for us but the thing is, it isn't about us...

The question is, will the average consumer notice the difference, and more importantly, is he willing to pay for it.

if that answer is no, we have a big problem.

Marc Schaerer
profile image
There is another reason why it makes sense to go with X86: its an established platform with working tool chains present already. No need for complex and hard to use stuff.
Also it saves Sony the trouble of recouping 2B USD of investment as with the CELL which will potentially prevent the PS3 from ever truely getting out of the lose sector, as the expected usage of CELL never materialized outside the PS3, neither for smart TVs nor for other usages, thanks to things like ARM as well as CUDA / GPGPU which removed the need to use a custom, PITA style processor that costs more and delivers less bang for the buck.

Kujel Selsuru
profile image
@Freek Hoekstra I'll wager the average consumer wont notice much of a difference between ps3 and ps4 so unless they can offer more then just gaming for that price I see trouble for these "high-end" machines. Oh and let's not forget the that the economy is still no where pre-2008 and so people are still going to be careful with their entertainment money.

Mathieu MarquisBolduc
profile image
Staying with powerPC would have helped. PowerPC assembly is so, so much easier to work with,debug,optimize, and console developpers are very proficient in it by now.

Nathan Mates
profile image
Even if you think PowerPC is so much cleaner, the problem with PowerPC is that IBM has really not committed to supporting it. The writing was on the wall a few years ago when Apple couldn't get IBM to commit to a low-wattage chip for laptops. IBM's primary use for their PowerPC chips is their big iron servers -- and that has two problems for console use: power consumption and no native GPU. Some other PowerPC variants exist for embedded systems, but those are much lower performance. And still no native GPU.

POWER7 has a 200 watt TDP. That's much higher than the Bobcat CPU which is ~18W or less. (Might be higher in the PS4 variant, as there are 2x the cores than shipped consumer variant, but that still gets to ~36W at a worst case). Then, add in the necessity for a GPU in a console -- Sony (mistakenly) thought that the PS3's cell would be sufficient to be a GPU last time around before enough developers told them that they really didn't want to be writing triangle rasterizers in CELL, thank you very much. IBM has no mass-market GPU available; only Intel, AMD and nVidia are in that market. nVidia's only got ARM chips of insufficient performance compared to Power/x86. In 5 years, that might change, but when shipping a console for hardcore graphics fanboys in 2013, ARM isn't in the running. Anandtech has done reviews of tablets showing even low-power x86 being competitive with ARM on per-watt performance, and midrange x86 smokes current ARM.

AMD has both a CPU and a GPU, already integrated, already running, and with a decent wattage. And, AMD is hungry (desperate?) enough to promise a system that works, and build things. AMD is also rumored to be the CPU+GPU maker for the next xBox; the Wii U is Power CPU + AMD GPU. You don't sell your GPU to Nintendo, MS & Sony w/o being decent. You don't sell your CPU+GPU to both MS & Sony without being good enough -- and I think that MS & Sony have evaluated the whole package, not just the perceived cleanliness of POWER debugging. Nintendo apparently had some HW development 'fun' with the Wii U with so many different HW makers all trying to integrate their pieces. That's what MS or Sony would get if they tried a POWER cpu also.

Why do I mention wattage? Because too much wattage in a small box can cook everything if it's not ventilated. See: X360's red ring of death. That's a *BILLION* dollar mistake where the systems overheated and died. Everyone pays attention when that much money is at stake. If your console form factor is about 2x a laptop in size, you are severely limited, and can NOT just throw in some monster POWER CPU.

As much as someone might think that Power is a cleaner/nicer architecture, amd64 cleans up enough rough edges of x86, and there's a ton of PC developers who know the rules. Smart developers can switch architectures. Going forward, POWER is a dead end in the consumer space, except for embedded CPUs in cars. x86 (i.e. amd64) and ARM are on a growth trajectory -- and that's been the path over the past 5-6 years.

Mathieu MarquisBolduc
profile image
Thanks for the information, Nathan. Seems like a big loss for IBM, to not have pursued that market more aggressively. Still, im going to miss RISC assembly...

Keith Thomson
profile image
The new systems have enough power to where hand optimizing assembly shouldn't be necessary. Personally, I'm hoping that developers DON'T immediately try to max the hardware out on these things, and instead just evolve the graphics gradually. What we have now on PS3/360 would be great if it was just bumped up to a higher spec graphics card and run at 1080p/60fps, and all the texture pop-in was eliminated.

Tommy Hearns
profile image
Well, A) they realized after the PS3 was released that they needed a more developer friendly strategy, so it's not like they can snap their fingers and make the Cell go away.

B) There isn't a low power yet powerful x86 chip yet and won't be until Intel releases Haswell (and probably not perfected until Broadwell) and the next Atom chip that are truly capable of existing in a fan-less mobile environment later this year. And then consider AMD is at least a step behind Intel because Global Foundries just doesn't have the fabs that Intel does. And seeing as hardware development takes years, you wouldn't have seen an x86 PSP until late 2014 at the earliest, which would have been a nine year life span for the first PSP. With the 3DS released that's business suicide.

C) ARM isn't powerful enough for console/desktop use and won't even go 64-bit until next year. Again, considering hardware development cycles, you would see an ARM PS4 in like 2016? I mean, really? And it probably won't even come close to the offered x86 hardware. The unified memory is exciting and extremely visionary IMO. It's like the first mainstream device to finally do what AMD has been talking about for years and what the fledgling GPGPU community is pushing towards.

So given the timeframe and state of hardware where x86 isn't really fan-less mobile ready yet and ARM isn't desktop ready yet, their decisions make sense.

Freek Hoekstra
profile image
@ Keith Thompson,

taking a game from currentgen and just rendering at higher resolution and higher framerate will be the first steps, but then the hand optimizing will return as teams will fight over best graphics etc. as it was with every generation of consoles...

I just hope people have the guts to move away from trying to reach for graphical realism and try and copy Dreamwokrs/Pixar art styles more. I think there is much to be gained going in that direction, as these characters are much easier to relate to, and allow for much greater differentiation and creativity.

we have seen Knack do this, as well as Epic's Fortnite (that is a game from THE grey brown shooter company that was epic btw...)

so I hope more developers will follow this trend.

Mathieu MarquisBolduc
profile image
@Keith Thomson

I was thinking less of hand-optimizations, and more of those hard-to reproduce issues that only happen in fully-optimized builds, and where you need to dig in the assembly to debug it.

[User Banned]
profile image
This user violated Gamasutra’s Comment Guidelines and has been banned.

Jarod Smiley
profile image
where is this general consensus of the "average" user not noticing the jump coming from? Are you guys browsing forums (hardcore/critical/addicted gamers) and making these conclusions?

Every person in my small circle who I would call "casual" has been blow away by Killzone PS meeting Debut, and are even impressed with Beyond. So I seriously don't get the pessimistic idea of casuals not being impressed with the upgrade, they are impressed rather easily, and PS4 being easier on developers should mean the power is harnessed even quicker than PS3, and impressive games will show up sooner. And really, with the indy boom, its more proof in my opinion that amazing graphics aren't even necessary to sell a platform, good games will sell it, and Sony seems to be trying to get a lot more games on the platform.

Gern Blanston
profile image
Just what I was thinking, well said. :)

I'm starting to think that those saying "the casuals this" and "the casuals that" don't actually have any casual gaming friends. And they just assume all of it.

Nick Quackenbush
profile image
We always have to remind ourselves of this very obvious, but elusive fact. Internet commentors != general consensus. If internet commentors were the majority of the gaming public, DLC wouldn't exist, F2P would have lost relevance with Zynga, and EA would be bankrupt.

The fact of the matter is internet chatter is the VAST MINORITY of the consumer base, but it is all we hear. So many of us get wrapped up in it, and even start to take it as fact, even though we know better.

Freek Hoekstra
profile image
however Beyond two souls is a PS3 game, not a PS4 game,
thus them being impressed with it (as am I btw) does not help sell PS4's

the thing is how big is the graphical leap, does it warrant sales.
if it isn;t big enough something else must warrant the sale.
gameplay mechanics/OS improvements/ new games that aren't released on the previous console etc etc.

but the thing is impressive as these games are they are not miles ahead of the PS3.
atleast not as many miles as the PS3 was ahead of the PS2.

Antonio Restivo
profile image
It seems rational that Sony would choose to go by the X86 architecture. But honestly I still feel disappointed that Sony "chickened out" and decided to throw away their multi-billion dollar investment with Cell architecture for the X86 architecture. IBM did make the cell architecture more user friendly to work with. From my memory(I could be wrong) the next cell cpu would have 32 cores running at 4 GHZ. And the GPU for the PS4 has near identical computational power to the RSX graphics processor. For example PS4 GPU has 1.84 TFLOPS as the RSX has 1.8 TFLOPS? I do not understand hardware specs that well but would the PS4 gpu be near identical to the RSX computationally? It seems like the PS4 was what the PS3 was hyped up to be back in 2006. Sony promised 1080p for the PS3 and CGI movie style graphics. But due to ram restraints the majority of the games had to look like 720p near identical xbox 360 games. It seems like this is the PS3 with the ram and bandwith limitations removed- not bad- but not a radical change as I hoped.

Artur Godlewski
profile image
@Antonio, actually PS4's GPU is a significant step up from RSX, with 1,84TFLOPS vs. 0,176TFLOPS. Not only that, but the architecture of the GPU is significantly enhanced, making the jump actually higher than x10.

Here's info on RSX: http://en.wikipedia.org/wiki/RSX_'Reality_Synthesizer'

Eric Garant
profile image
You said in in the end of this article: "In a forthcoming article, Gamasutra will share the many details of the PlayStation 4's architecture and design that came to light during this extensive and highly technical conversation."

Just wonder when are you planning to publish it ?

Thanks.

Christian Nutt
profile image
Within the next couple of weeks, but no firm date just yet. Need to make sure all the technical information is correct, etc.

Eric Garant
profile image
I really looking forward to read this article. If it's possible, let me know when you have a firm date.

Thanks.

Eric Garant
profile image
Hi Christian !

Do you have a firm date now for the forthcoming article ?

Thanks.


none
 
Comment: