Overclock.net › Forums › Video Games › Console Gaming › Playstation › PS4-some personal speculation...
New Posts  All Forums:Forum Nav:

PS4-some personal speculation...

post #1 of 12
Thread Starter 
Even though when it comes to consoles ,game developers are able to aggressively optimize their games for one specific hardware and thus extract seemingly magical performance from weak hardware, this time it seems the CPU is exceptionally weak(Sure it can process 8 threads simultaneously, but the cell could also process 6 threads, and so 2 more low-IPC threads don't seem particularly Next-gen power to me). Sure, proprietary API’s and optimization can work wonders, but the Jaguar architecture is simply not a very capable one and I am willing to bet my money on the fact that in terms of game development, it isn’t much better than the Cell processor in the PS3 (Well, this would also depend on the APUs operating frequency, which hasn’t been unveiled yet, but I am willing to bite the bullet and hypothesize that it isn’t going to be much more than 3.2 Ghz). For this exact same reason, some are wondering whether the developers will be using the GPGPU capabilities of the PS4′s advanced GPU to perform some functions of the CPU and thus be able to alleviate the problem. But how exactly will the GPGPU capabilities of the GPU be exploited? And here is where my theory steps in. If you are even partially interested in the covert world of tech prophesies, please read on, because unlike some of the most crappy conspiracy theories out there , mine fits the bill perfectly.
First, Bear in mind these points:

The PS4 employs GDDR5(in a pretty weird decision, as GDDR memory has never been used as the main system RAM before, so the decision has solely been made for exceptional speed, remember, they expressly wanted the maximum bandwidth for the CPU and the GPU) as it’s main unified memory, and is exceptionally fast @ 176 Gb/s (in comparison the NextBox’s been rumored to use the use the much slower DDR3 standard, peaking out at only 68 Gb/s).
The CPU and GPU are on the same die, because PS4 utilizes a custom AMD APU. If you are unfamiliar with an APU, it is mainly a processor which has a GPU directly beside it on the die and thus the GPU is called an iGPU (integrated GPU).
Sony curiously talked about how it’s leading the charge on development efficiency (remember this point).
PS4 unites a rather large GPU as an iGPU, something that rivals a HD 7850 and is an iGPU? That’s unheard of (the best iGPU currently only rivals the rather puny HD 6670) This has been done deliberately as we will soon see).

Now, I said that I have a theory about how the GPGPU capabilities of the GPU will be levied in the PS4 and surprise, surprise, it’s through tech AMD has really been pioneering since the last few years (and companies like Samsung and QUALCOMM are onboard and the reason why the PS4 has the CPU and GPU by AMD ) and it is HSA. Heterogeneous System architecture, so you don’t know what that means. Here let me explain:
“The computing industry is approaching a formidable obstacle course where anyone wishing to drive advances in computing technology must carefully negotiate several key trade-offs. First, reducing power consumption is increasingly critical across all segments of computing. Consumers want improved battery life, size and weight for their laptops, tablets and smartphones. Likewise data center power demands and cooling costs continue to rise.”
“To navigate this complex set of requirements the computer industry needs a different approach – a more efficient approach to computer architecture. We need an approach that promises to deliver improvement across all 4 of these vectors: Power, Performance, Programmability and Portability.”
So basically, with the way things are moving, AMD has already realized that the process of shrinking nodes down and reaping the resultant benefits won’t last forever. Intel is already at 22nm and soon the industry is going to hit the problem of the quantum tunneling effect (At 4-5nm) so this process of shrinking nodes can’t last forever. The returns are diminishing and the costs rising. The development model is becoming exactly like what it shouldn’t be. We need a fresh approach to computing,one that is efficient and cost-effective. Enter HSA.
Basically, HSA is about finding a middle path between the complexities of programming in GPGPU languages(such as CUDA and OpenCL) for a GPU and the relative ease of programming for x86-64 CPU’s. So basically to exploit the massive parallel power of a modern GPU without the horrors of coding solely in complex GPGPU languages because it is financially undesirable (more so when it comes to game development). And this is the point people are missing right now. Why would Sony save Devs from the horror of coding the cell, only to make the problem much worse, i.e. GPGPU coding? Without some other trick to make it easy of course. And that ‘trick’ is HSA.
What is “HSA”?

HSA is a short term for “Heterogenuous Systems Architecture”. It is not something that is easy to describe because it is sort of both a software and a hardware thing.
How AMD describes HSA: The Heterogeneous System Architecture (HSA) provides a unified view of fundamental computing elements. HSA allows a programmer to write applications that seamlessly integrate CPUs (called latency compute units) with GPUs (called throughput compute units), while benefiting from the best attributes of each.
HSA is an integration that creates a compromise between traditional CPU-based coding and GPGPU coding. Traditional CPU coding (i.e. x86, C++, etc) is easier, less costly, and less time consuming while GPGPU coding is far more efficient but it is not common because it is limited, more difficult, costly and time consuming. HSA is a combination of both. Its application is a variant of GPGPU (general purpose GPU) processing in which some functions of applications are coded to take advantage of the better performance & power efficiency that the GPU can provide in some scenarios, but it retains the ease and flexibility of coding for the traditional CPU.
HSA is a totally different and new direction because AMD may be the first processor company to have made such significant investment primarily to improve ease of programming.

The reason why AMD is the only company who seems to have a product ready for HSA is because they designed their APUs to culminate at this (read on to find out why the APUs have been designed the way they have been). Now It makes sense why Sony opted for an APU, doesn’t it? And why the said APU was customized to include a much bigger GPU as an iGPU than even the best current commercial APU has. Let’s read a little more to make things clearer :
“Today, a growing number of mainstream applications require the high performance and power efficiency achievable only through such highly parallel computation. But current CPUs and GPUs have been designed as separate processing elements and do not work together efficiently – and are cumbersome to program. Each has a separate memory space, requiring an application to explicitly copy data from CPU to GPU and then back again.
A program running on the CPU queues work for the GPU using system calls through a device driver stack managed by a completely separate scheduler. This introduces significant dispatch latency with overhead that makes the process worthwhile only when the application requires a very large amount of parallel computation. Further, if a program running on the GPU wants to directly generate work-items, either for itself or for the CPU, it is impossible today!”
The purpose of HSA is to seamlessly share data between CPU and GPU, without memory copies or cache flushes because it assigns each part of the workload to the most appropriate processor with minimal dispatch overhead. The resulting performance is not possible using only Multicore CPU, only GPU, or even combined CPU and GPU with today’s driver model. Just as important, it is done using simple extensions to C++, not a totally different programming model.A middle path.

As mentioned, one of the key points that can be made from HSA is that optimization for HSA is going to be easy to code – this is because of the hardware optimizations AMD can make in their APUs. Traditional programming for OpenCL and CUDA to enable apps to take advantage of the GPU has provided the same kind of benefits HSA will provide for quite awhile, but coding to take advantage of OpenCL and CUDA is difficult, time-consuming, and costly.
Coding for HSA optimization will be almost as easy and as quick as coding traditionally while giving the same performance and efficiency provided by the more complex GPGPU coding methods available today. This means that HSA can be easily adopted by developers.

The above points explains why Sony stressed so much on how it is achieving development efficiency (this would not have been possible without HSA, as using GPGPU solely would be too cumbersome for game development, the budgets would skyrocket.).
Now, why use GDDR5 as the main system memory? Clearly, GDDR RAMs have been designed from the start to be used only as V-RAMs. This may be why :

AMD has exclusivity because HSA requires a specific level of integration that AMD has been preparing by designing its APUs a certain way. Intel’s “APUs” do not follow the same design principles.
For example: the average user might notice that the Llano and Trinity APUs are very bandwidth intensive and require fast RAM to get better performance.
This is actually because on AMD APUs, CPU & GPU communication is done via main memory. Llano and Trinity APUs accomplish a partial integration of this and the Kaveri APU which introduces HSA will feature true address-space sharing between CPU & GPU to enable HSA.

We know the APU in PS4 is a Richland one, the predecessor of the forthcoming Kaveri APU, the latter will enable full blown HSA in the PC space. But we also know that the APU in the PS4 is a highly customized one, is this what Sony and AMD have been cooking up? Even if this customization is not present, even then the Richland APUs have partial implementation of HSA. Now we also know why memory bandwidth is very important in such implementation, from the above bullet points.
This explains why the PS4 uses extremely fast (176 Gb/s) GDDR5 memory as unified(needed for HSA) memory for the system. How fast is it really? About 7 times than the bandwidth found in the PS3 and about 3 times than the bandwidth found in modern PC’s equipped with DDR3 RAM. I think in the PS4, the cpu and the iGPU share the address space.
HSA provides exactly the kind of benefits consoles needed for this gen, as consoles have to provide a generational leap In power without breaking the bank and keep game development costs in check, so efficiency of development process is needed.

Technology access is cheaper for everyone (Since performance does not require expensive parts)-Cheap
Lowered cost of software In addition to hardware, game development budgets don’t sky rocket, exactly what Sony has been masquerading about (and what simple implementation of GPGPU CANNOT PROVIDE).
PCs/CONSOLES worldwide will be providing the same performance while using less power- Power efficiency.

Now we have seen why :

Sony chose GDDR5 as the main unifiedRAM
Chose to go with a custom AMD APU, the only product which can currently enable HSA.
Why it talked so much about game development efficiency on the new York conference, instead of bragging about PS4′s hardware power, as is customary for consoles PR.

The reason is: PS4 might be implementing HSA (calling it).
This again bodes extremely well for PC gamers, well at least for those with AMD APUs. Similar to how people speculate that due to these next gen systems having CPUs similar to PC one’s, we might see PC ports being more optimized and using more CPU cores, we also might see game developers utilizing HSA on AMD APUs on the PC side of things. If this is indeed the case, then this is a significant win for AMD and it’s customers, and we all know how badly both needed one ( Bulldozer- I’m looking at you). HSA also offsets AMD’s performance processors woes with FPU’s, as in they currently suck at it. BUT, GPUs are good at calculations regarding FPU’s, actually VERY GOOD, like TERA FLOPS GOOD. This means that AMD will sometime in the future convert it’s CPU line-up to a pure , APU one (since these are the CPUs that levy the benefit of GPUs FPU processing capabilities and other capabilities too), the current flagship line up will cease to exist, sometime in the future (I think the excavator line-up will be the first line up consisting purely of APUs, but that’s another can of worms).
AMD’s Announcement at GDC, related?

AMD’s HSA implementation is huge in terms of the performance benefit it brings, but how could AMD convince game developers to port games with such optimization on the PC? Considering gamers predominantly use Intel Hardware, developers would surely find it hard to justify such an investment. But what if the games were made to run on the tech in the first place, such that hardly any more investment has to be made for the PC? See AMD guys are really smart. Considering the NextBox has also been rumored to use AMD tech, they are also probably using HSA (they will have to, if they want to remain competitive with the PS4) so games, before being ported over to the PC, will already be running on x86-64 CPUs, similar modern GPUs and HSA! What is there to even ‘port’ now? Just some optimizations have to be made. So, on the PC side of things, AMD APUs will become the game developers choice of hardware (since porting games to their APUs will be the easiest for them), and gamers choice of hardware too (and Intel will surely be overshadowed but they will remain there). In light of these facts, the following lines by John Taylor, AMD’s Vice President of Global Communications and Industry Marketing,start making real sense :
“This is going to be a very exciting year for gamers, especially for those with AMD hardware in their PCs and consoles, as we have even more game-changing (pun intended) announcements still to come,” he said. “Look for some more exciting things happening at the Game Developers Conference (GDC) in March when we will provide even more info on how we are working with game developers to make AMD the hardware of choice for running the best games!”
Considering the industry’s perception and performance of AMD’s PC hardware right now (excluding their GPUs),they can hardly be considered the hardware of choice for gamers. The only way such a stark difference can be brought along, the only way the tables can be turned, is through HSA and the resulting optimizations their hardware will get on the PC (again, calling it). It can also be something entirely unrelated, but I hope it is this.
HSA is quite possibly Sony’s secret weapon, and possibly maybe Microsoft’s (and AMD’s saviour).
AMD's announcement at GDC could also be simply to announce that the NExtBox is also using their APU tech...either way,both cases support my hypothesis.

MORE PROOF:Guys,this just in and it's crazy,YO!!,AMD PLANS TO SELL A VERSION OF THE PS4 APU TO PC BUILDERS.How is that related to my theory,i hear thee ask?This is how:

SO get a load of what AMD’s head of marketing for Global Business Units,John Taylor,has told the inquirer:
“Everything that Sony has shared in that single chip is AMD [intellectual property], but we have not built an APU quite like that for anyone else in the market. It is by far the most powerful APU we have built to date, it leverages [intellectual property] that you will find in our A-series APUs later this year, our new generation of APUs but none that will quite be to that level of sheer number of cores, sheer number of teraflops.”

If you have already realized what the above means,you already are congratulating me…(because i was right),but if not,then let me explain!!
So,i had hypothesized that the APU in the PS4 sported HSA implementations which will be found in AMD’s upcoming APU’s,which will debut later this year. Now according to the above quote,where John is talking about the PS4′s APU,he says all the usual stuff about how it is the biggest and baddest APU ever made,but he also says and i quote:
“it leverages [intellectual property] that you will find in our A-series APUs later this year, our new generation of APUs but none that will quite be to that level of sheer number of cores, sheer number of teraflops.”

SO,He is effectively saying that the PS4′s APU leverages ‘tech’ from their upcoming APU’s(which will debut later this year),but the upcoming APU’s won’t be better in performance.Now what APU’s are debuting later this year?Well 2 are: Richland in the first Half and Kaveri in the second.Now,judging by how John says ‘later this year’,he is obviously referring to the Kaveri APU’s…AND WHAT NEW TECH DOES THE KAVERI HAVE THAT THIS PS4 APU MAY BE(I AM ALMOST A 100 PERCENT SURE HERE) ‘LEVERAGING’?TECH TO ENABLE HSA!!(what other 'intellectual property' could it even leverage?Both are APU's so both already have all the IP necessary for a APU to function...,the major new thing Kaveri introduces is full support and implementation of HSA).
and just as a consolation,think about why BOTH sony AND microsoft left Nvidia for none other than an AMD APU(the microsoft part isn’t confirmed yet,but i digress)…none of them want to miss out on the HOT HSA action..
Driving the POINT home.Over and out.

Didn't know where to post it,so posted this here. what do you guys think?
Edited by kriss super boy - 3/6/13 at 5:52am
blackbeard
(11 items)
 
 
CPUMotherboardGraphicsRAM
fx 8120 biostar ta990fxe hd 5970 kingston hyperx blue 
Hard DriveCoolingOSMonitor
WD caviar green  antec kuhler h20 920 windows 8 pro x64 32" lg 1080p display  
PowerCaseAudio
corsair tx 750 CM elite 430 phillips hts5520  
  hide details  
Reply
blackbeard
(11 items)
 
 
CPUMotherboardGraphicsRAM
fx 8120 biostar ta990fxe hd 5970 kingston hyperx blue 
Hard DriveCoolingOSMonitor
WD caviar green  antec kuhler h20 920 windows 8 pro x64 32" lg 1080p display  
PowerCaseAudio
corsair tx 750 CM elite 430 phillips hts5520  
  hide details  
Reply

Gear mentioned in this thread:

post #2 of 12
Good read (yes, I read the entire post). Assuming you're correct - and I certainly hope you are - this will mean good things for gamers of all platforms in the very near future.
post #3 of 12
Thread Starter 
oh thank you for reading the entire damn thing...(i get carried away sometimes),and it really is revolutionary.i certainly hope i am right.
blackbeard
(11 items)
 
 
CPUMotherboardGraphicsRAM
fx 8120 biostar ta990fxe hd 5970 kingston hyperx blue 
Hard DriveCoolingOSMonitor
WD caviar green  antec kuhler h20 920 windows 8 pro x64 32" lg 1080p display  
PowerCaseAudio
corsair tx 750 CM elite 430 phillips hts5520  
  hide details  
Reply
blackbeard
(11 items)
 
 
CPUMotherboardGraphicsRAM
fx 8120 biostar ta990fxe hd 5970 kingston hyperx blue 
Hard DriveCoolingOSMonitor
WD caviar green  antec kuhler h20 920 windows 8 pro x64 32" lg 1080p display  
PowerCaseAudio
corsair tx 750 CM elite 430 phillips hts5520  
  hide details  
Reply
post #4 of 12
This post was awesome, thanks!
post #5 of 12
long read and half of the things are still taking time to digest...but if i am not wrong then this ps4 is going to the craziest, most powerful machine to exist...wow !! really is exciting stuff !!

3 times as fast as ddr3 sounds pretty rapid to me !
post #6 of 12
I'm really interested to see how fast the machine will be with that unified ram. Plus if throwing a 256GB SSD in there instead of whatever Toshiba 5400rpm drive they put in it, how much faster it would be.

Definitely going to get a PS4, not so keen on the next Xbox really, nothing special there. This is coming from someone who's had several 360's and original Xbox's. Anyway, yeah PS4 looks to be something very epic. Hopefully they tap its true power sooner than 2-3 years as was the case for current consoles.
SandySLI
(27 items)
 
iMac
(12 items)
 
Media Box
(22 items)
 
CPUMotherboardGraphicsRAM
i5 4570R Apple Logicboard Intel Iris Pro  8GB 
Hard DriveOSMonitorMonitor
1TB HDD Yosemite 10.10.2 iMac 21.5" 1080p ASUS VH228 21.5" 1080p 
MonitorKeyboardCaseMouse
ASUS VH228 21.5" 1080p Apple Numeric Pad Keyboard Unibody Alu Apple Mouse 
CPUMotherboardGraphicsRAM
Intel Pentium G640 MSI H61-I E35 Intel HD Graphics AMD Entertainment Edition CL9 1600Mhz 
RAMHard DriveHard DriveHard Drive
AMD Entertainment Edition CL9 1600Mhz Seagate Pipeline | 2TB | 7.2k | 3.5" Seagate Pipeline | 2TB | 7.2k | 3.5" WD Black | 1TB | 7.2k | 3.5" 
Hard DriveHard DriveHard DriveCooling
Hitachi TravelStar | 750GB | 5.2k | 2.5" Hitachi TravelStar | 750GB | 5.2k | 2.5" WD Blue | 500GB | 5.2k | 2.5" Akasa K25 CPU Heatsink 
CoolingOSMonitorKeyboard
120mm Stock case fan Windows 10 x64 n/a - Remote access via other devices n/a 
PowerCaseMouseMouse Pad
CoolerMaster GM550 BitFenix Neos n/a n/a 
AudioOther
n/a ASRock 2-port Sata-6G PCI-E-1x 
  hide details  
Reply
SandySLI
(27 items)
 
iMac
(12 items)
 
Media Box
(22 items)
 
CPUMotherboardGraphicsRAM
i5 4570R Apple Logicboard Intel Iris Pro  8GB 
Hard DriveOSMonitorMonitor
1TB HDD Yosemite 10.10.2 iMac 21.5" 1080p ASUS VH228 21.5" 1080p 
MonitorKeyboardCaseMouse
ASUS VH228 21.5" 1080p Apple Numeric Pad Keyboard Unibody Alu Apple Mouse 
CPUMotherboardGraphicsRAM
Intel Pentium G640 MSI H61-I E35 Intel HD Graphics AMD Entertainment Edition CL9 1600Mhz 
RAMHard DriveHard DriveHard Drive
AMD Entertainment Edition CL9 1600Mhz Seagate Pipeline | 2TB | 7.2k | 3.5" Seagate Pipeline | 2TB | 7.2k | 3.5" WD Black | 1TB | 7.2k | 3.5" 
Hard DriveHard DriveHard DriveCooling
Hitachi TravelStar | 750GB | 5.2k | 2.5" Hitachi TravelStar | 750GB | 5.2k | 2.5" WD Blue | 500GB | 5.2k | 2.5" Akasa K25 CPU Heatsink 
CoolingOSMonitorKeyboard
120mm Stock case fan Windows 10 x64 n/a - Remote access via other devices n/a 
PowerCaseMouseMouse Pad
CoolerMaster GM550 BitFenix Neos n/a n/a 
AudioOther
n/a ASRock 2-port Sata-6G PCI-E-1x 
  hide details  
Reply
post #7 of 12
This was a good read and a very interesting theory you have, I'm hoping you're correct.
post #8 of 12
Holy crap I can't believe I read all of that and only got a few (low IQ).. still interesting..

if HSA is the way then the next xbox would be gimped so much? Or am I getting this wrong?

Sounds like NV and Intel are screwed for game optimizations in the near future.
Intel System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500K 4.3Ghz @~1.344 V ASUS P8Z68 V PRO Sapphire HD5870 2GB Eyefinity 6 G.Skill Sniper 1866 Mhz 9-10-9-28@1T 
Hard DriveOptical DriveOSMonitor
Samsung F3 1TB, Maxtor 500GB, SG 1.5 TB LG Lightscribe Windows 7 64bit Samsung 2333HD (TN Panel) 
KeyboardPowerCaseMouse
generic wireless keyboard from Logitech Antec TruePower New 650 watts NZXT Gamma dell optical 
Mouse Pad
simple 
  hide details  
Reply
Intel System
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5 2500K 4.3Ghz @~1.344 V ASUS P8Z68 V PRO Sapphire HD5870 2GB Eyefinity 6 G.Skill Sniper 1866 Mhz 9-10-9-28@1T 
Hard DriveOptical DriveOSMonitor
Samsung F3 1TB, Maxtor 500GB, SG 1.5 TB LG Lightscribe Windows 7 64bit Samsung 2333HD (TN Panel) 
KeyboardPowerCaseMouse
generic wireless keyboard from Logitech Antec TruePower New 650 watts NZXT Gamma dell optical 
Mouse Pad
simple 
  hide details  
Reply
post #9 of 12
All of this is just reminding me of what was said about The Cell and what actually happened afterwards.

Dont get me wrong, I love my PS3, but Sony just made things too difficult for devs. In terms of money, time, knowledge, tools, logistics, return on investment, everything.

I even had Yellowdog installed on my console until they took OtherOS out. The Cell was a beast, only restrained by the 256MB of GDDR3.

Long ago I read a thread on the Yellowdog forums, Fixstars had theorized that if the Hypervisor could be circumvented, the PS3 would be able to transcode HD Video something like 108 times faster than a Q6600. Provided the software was optimized for the Cell SPE's "In-Order" computational method.
    
CPUMotherboardGraphicsRAM
FX-8350 @ 4.6GHz 1.48v 990FXA-UD3 Pegatron HD6450 1GB 2x4GB G.Skill Sniper 1866 
Hard DriveOptical DriveCoolingOS
WD Blue 500GB Samsung DVD-RW Corsair H100i Win 7 x64 
MonitorPowerCase
LG 32" 720p HDTV SeaSonic M12II 620 Bronze CM 690 II Advanced 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
FX-8350 @ 4.6GHz 1.48v 990FXA-UD3 Pegatron HD6450 1GB 2x4GB G.Skill Sniper 1866 
Hard DriveOptical DriveCoolingOS
WD Blue 500GB Samsung DVD-RW Corsair H100i Win 7 x64 
MonitorPowerCase
LG 32" 720p HDTV SeaSonic M12II 620 Bronze CM 690 II Advanced 
  hide details  
Reply
post #10 of 12
Quote:
Originally Posted by Baskt_Case View Post

All of this is just reminding me of what was said about The Cell and what actually happened afterwards.

It's the exact opposite, Sony are moving to X86 arch, well known, easy to develop for, (Better for PC aswell) although the PS4 isn't a powerhouse at launch like previous consoles have been, it will be interesting to see what dev's can get from the PS4, encoding direct to the metal and 8GB DDR5 to play around with, I'm expecting very impressive games.

Will probably pick up a PS4 at launch, can't resist Killzone and Drive club haha.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Playstation

Gear mentioned in this thread:

Overclock.net › Forums › Video Games › Console Gaming › Playstation › PS4-some personal speculation...