Theoretical vs Practical Power

The Cell was, for a while, difficult to understand, but the potential power was enough to deliver great games for 10+ years. Xbox 360, although much, much weaker, had more practical power at first and delivered better games the first 2 years or so.

LMAO. If you actually believe that bullcrap.

if you want to make the devs hate you, you'll give them Cell 2

The Cell CPU was used in super computers at Norrad and several other big military facilities. It was a bleeding edge niche processor for the best of the best.

Anyone who talks badly about the Cell didn't realize where and for what it was being used for. Not many processor's could stand up to it when the ps3 launched.

Well that's one the reasons ps3 did so poorly compared to other playstations, the Cell was expensive + hard to program for and they went with weaker gpu and ended up getting the d in pretty much all multiplats vs x360

So why shouldn't one point out that Cell didn't pay off?

OP must have just started gaming because the Xbox 360 was NOT "much much weaker" than the PS3.

The Cell wasn't a pure CPU, but more like a CPU and GPU combined. On the other hand, the PS3's GPU was weaker than the 360's GPU. The 360's design was much more practical, letting the GPU handle all the graphics processing. Whereas the PS3's design was unusual, splitting its graphics processing capabilities between the CPU and GPU, making it more difficult to tap into its power.

They're x86 machines, there is no theoretical vs. practical power, there is no unknown peak to the architecture, it's simple, it's easy to understand and we already know what each system is capable of doing.

Theoretical power is a rough translation for convoluted architecture that is a nightmare for programmers, it's garbage.

@ProtossRushX: Exactly my point. Cell was essentially a super computer, it's just the Devs couldn't always easily tap in to all the power.

Cell was not essentially a super computer, it had to be run in tandem with dozens, and dozens of other machines to fall into that classification.

@ProtossRushX: Yeah if you connected a couple hundreds of them it could be used as a supercomputer lol.

The US department of defense had to connect 1,760 PlayStation 3's for their supercomputer lol...

@dynamitecop: Well if you think about it, all The Cells were connected through PSN, so essentially all PS3 owners had their own super-super computer in a way. Lol. Folding@home combined all the power for research.

That's a pretty large leap in rationalization lol..

Indeed, games like skyrim, dragon's dogma and so many others were running worse on the ps3 and these were games late in the gen.

it's only the first party devs that actually could harness the power of the ps3 and this was mostly seen in animation cutscenes like the last of us and uncharted, but the xbox was the better system overall, not only because it was more performant but also because it was cheaper, had better network performance and had party chat.

Shame microsoft suffered from megalomania with the xboxone, and sony happily capitalized on it with the ps4, apparently ms found their thunder back with the scorpio (well, let's hope so), and sony seems the one to be screwing up now.

Apparently all console manufacturers suffer from the same problem when they are successfull and that is the underestimation of the customer and the competition.

the only thing to be learned from the PS3 is how not to design a console. sony leanred the lesson well.

it was expensive to produce. it cost the customer more yet offered inferior version of multiplats (and even at the end of gen it was mixed at best). it cost developers an absolute fortune to make games on. it was, frankly, unacceptable. sony are very very lucky they didnt completely bottle it in the sales department.

they basically through the PS brand under a bus to try and get the cell and blu-ray established. in the end the cell has been scrapped (and rightfully so) and blu-ray hasnt been worth it...streaming is taking over (somethign sony also acknowledge..no UHD blu-ray for the PS4).

any percieved advantage the PS3 had at the end from its exclusives had less to do with the hardware and more to do with very talented teams working on the same hardware for a decade. had these people worked on the 360 they would have achieved the same results (and probably for less money too).

the console itself was a complete shambles. there is nothing, absolutely nothing, positive to be taken from the PS3s design.

the best example of a console that had a nice balance of practical and theoretical imho is the gamecube. developers actually liked working on the system. those that tried it for the first time were surprised by it (i always remember the story of the creative assembly getting spartan total warrior running on the gamecube. they thought the CPU just wouldnt be able to deal with it and they would need to cut and do loads of hacking. it was a non issue and they were quite surprised). amusement vision got amazing resuluts out of it with Fzero GX with about 18 months of development.

but it did take some mastering to really make it sing. guys like rare and factor 5 seemed to be able to really make that hardware shine. nintendo too of course...twilight princess is quite an impressive looking game considering the host.

modern console development, from what i see in interviews n such like, has changed quite a bit though. its a bit more like PC development. there are now layers between the game and the hardware. nowhere near as many as, say, openGL or DX11 and less than vulkan/DX12 too. but still layers. the games have become more hardware agnostic and developers dont seem to be coding as much to the metal as they used to. in fairness this does make sense. the days of having to squeeze the best out of every KB and MHz are gone..the pay off is not worth it and the industry couldnt bare the cost of it now anyway.

this has its upsides too. BC should be easier to maintain going forward.

@osan0: blu ray was the one good decision they made with the ps3. No compressed audio, cgi, more varied textures for exclusives, no multi discs... not to mentinon that they're more durable than dvd.

The cell from what I remember was really awesome at floating point calculations, pretty average to cr@p at everything else. Was not a super computer.

@jereb31: It's power was never fully harnessed. Still can probably compete with the latest processors if some dev team could actually tap into it's full power.

Well, maybe, but i really doubt it. I'll see if i can dig out a link or something. I'm curious whether it does stack up in an ideal situation.

There was nothing to "harness". The real world is not an anime. There is no power levels nor is their true form. If you continue to speak in such a manner you are heading down the road of stupidity.

For PC GPUs with 4 TFLOPS or greater and general PC benchmarks, effective memory bandwidth has larger influence with practical performance.

Unlike NVIDIA Maxwell/Pascal GPUs, AMD didn't incease their effective memory bandwidth with TFLOPS increase. NVIDIA has improved their memory compression capabilities with ~1.56X memory compression factor for Maxwell GPUs.

Ideally, PS4 Pro should have raw 352 GB/s memory bandwidth or 280 GB/s effective memory bandwidth.

Polaris GPU has ~1.36X memory compression factor.

Scorpio's "more than 320 GB/s memory bandwidth" was done for a purpose.

Kind of agreeing with @Pedro,

Articles all sound nice and fancy, http://www.anandtech.com/show/1719, http://www.anandtech.com/show/1647/13

But at the end of the day, it was a moderately sized CPU, with very specific core functions. It excelled in parallelism but did not excel in much else.

This summed it up nicely enough: https://groups.google.com/forum/?hl=en#!msg/alt.games.video.sony-playstation2/iVcRIKgR-r8/qY6nbtmD_2IJ

Pretty sure the Cell processor was beaten by standard PC cpu's prior to the PS3 release as per normal. But the "True power" of the cell looks like it was mostly hype and didn't even hold up to the Pentium 4's of the time.

EA DICE was forced to use CELL's SPEs for their deferred rendering processes while X360's Xenos microcode mode was able to handle it.

From https://forum.beyond3d.com/posts/1460125/

------------------------

"I could go on for pages listing the types of things the spu's are used for to make up for the machines aging gpu, which may be 7 series NVidia but that's basically a tweaked 6 series NVidia for the most part. But I'll just type a few off the top of my head:"

1) Two ppu/vmx units

There are three ppu/vmx units on the 360, and just one on the PS3. So any load on the 360's remaining two ppu/vmx units must be moved to spu.

2) Vertex culling

You can look back a few years at my first post talking about this, but it's common knowledge now that you need to move as much vertex load as possible to spu otherwise it won't keep pace with the 360.

3) Vertex texture sampling

You can texture sample in vertex shaders on 360 just fine, but it's unusably slow on PS3. Most multi platform games simply won't use this feature on 360 to make keeping parity easier, but if a dev does make use of it then you will have no choice but to move all such functionality to spu.

4) Shader patching

Changing variables in shader programs is cake on the 360. Not so on the PS3 because they are embedded into the shader programs. So you have to use spu's to patch your shader programs.

5) Branching

You never want a lot of branching in general, but when you do really need it the 360 handles it fine, PS3 does not. If you are stuck needing branching in shaders then you will want to move all such functionality to spu.

6) Shader inputs

You can pass plenty of inputs to shaders on 360, but do it on PS3 and your game will grind to a halt. You will want to move all such functionality to spu to minimize the amount of inputs needed on the shader programs.

7) MSAA alternatives

Msaa runs full speed on 360 gpu needing just cpu tiling calculations. Msaa on PS3 gpu is very slow. You will want to move msaa to spu as soon as you can.

Post processing

360 is unified architecture meaning post process steps can often be slotted into gpu idle time. This is not as easily doable on PS3, so you will want to move as much post process to spu as possible.

9) Load balancing

360 gpu load balances itself just fine since it's unified. If the load on a given frame shifts to heavy vertex or heavy pixel load then you don't care. Not so on PS3 where such load shifts will cause frame drops. You will want to shift as much load as possible to spu to minimize your peak load on the gpu.

10) Half floats

You can use full floats just fine on the 360 gpu. On the PS3 gpu they cause performance slowdowns. If you really need/have to use shaders with many full floats then you will want to move such functionality over to the spu's.

11) Shader array indexing

You can index into arrays in shaders on the 360 gpu no problem. You can't do that on PS3. If you absolutely need this functionality then you will have to either rework your shaders or move it all to spu.

Etc, etc, etc...

PS3's GPU was aging during unified shader era GPUs.

NVIDIA Kelper has aged badly during this game console generation while AMD GCN 1.0/1.1 PC GPUs was able to keep up with NVIDIA Maxwell and Pascal.

And then there's stuff like this...

Sony's First party was no better at it then others. Nothing struck me as "impossible on XBOX 360" even though there might technically have been things. But there was nothing jaw droppingly beautiful or running at insane frame rates.

@Pedro: No but there are often times "hidden secrets" if you will when it comes to processors. The power of The Cell wasn't fully realized until developers could understand it better. Only toward the end did they begin to wrap their heads around it. The Cells maximum efficiency has probably yet to be tapped.

Don't know don't care, what I do know is the ps pro has quality exclusives whereas the xboned Scorpio is weaker than my current rig and will most certainly be a lot weaker than my future rig so really theres no point for me to own an inferior system

Why would I ever wish for theoretical power in a medium where my enjoyment stems from its practical implementation, however lessened? It being possible in theory gives me nothing. The pleasure of gaming comes from results given, not power theorized.

@MirkoS77: Couple reasons:

1.) Bragging rights of having a more powerful system

2.) Knowing your system has the potential to last a very long time

3.) Value in knowing you paid for something which has a lot of power as opposed to something which has less power

The Cell was exceptionally powerful for its time, with 230.4 GFLOPS of power, comparable to a powerful GPU in its time, or a powerful CPU today. Combined with its GPU, that gave the PS3 a total 476.8 GFLOPS of processing power, the highest in its time.

But the problem was that the Cell wasn't really designed for games, but was designed for supercomputing applications. It was difficult to actually harness that power for gaming, but over the course of a generation, developers were eventually able to utilize much of that power.

PS3 = 5 SPE at 3.2 Ghz

PS4 CPU = 6 CPU at 1.6 Ghz

XBO = 6 CPU at 1.75 Ghz

AMD Jaguar has superior effective work IPC (instructions per cycle) when compared to IBM CELL.

FLOPS comparison means little when comparing different architectures.

NVIDIA RSX's FLOPS number is a load of b*llshit since it has problems with sustained 32bit FP compute.

PS3's 476.8 GFLOPS didn't NOT deliver Geforce 8800 GTX level gaming results.

Intel Core i7-4790K at 4 Ghz has about 500 GFLOPS without factoring it's HD 4600 IGP with 432 GFLOPS.

Intel Core i7-4xxx "Haswell" has Intel AVX v2.0 SIMD extensions with GPU style gather instructions. Intel HD 4600 IGP has 20 IEU (SIMD stream processors) at 1.25 GHz and has GPU gather instructions..

@ronvalencia: That's a bit old info tho, Sony unlocked 7th core for devs in SDK update last november

CELL has 6 SPEs + PPE for game.

PS4 has 7 CPU cores for game.

was reference to this "PS4 CPU = 6 CPU at 1.6 Ghz"

Ubisoft hasn't update their benchmark.

Bro , this generation consoles are typical mid to low end PCs. Theres nothing THAT complicated or untamed power for developers to work with like CELL was etc.

PS3 was the most moronic system SONY ever made . It didnt make ANY sense at all and if you want further explanation why i can tell you.

I must admit I do miss the surprises we had last gen due to optimisations and enhancements of the SDK.

Pretty much what we got on release is all we got this gen graphics wise same as PC.

Practical, of course. And no TC, 360 wasn't 'much, much weaker'. In fact, considering it's superior GPU and RAM solution, it was the most capable console last generation. Same exact reasons why PS4 is more powerful than Xbox One (GPU + RAM vs CPU).

The cell was like a cpu and gpu combined as mentioned in the thread. Ultimately both the ps3 and the 360 stayed in the same ballpark.

And the ps3 was hopelessly outmatched by a hi-end pc back in 2006. A Core2 Duo coupled with a 8800 GTX ran games at resolutions and frame rates that the ps3 could only dream of, not to mention dx10 support.

"hidden secrets" O_O

As an engineer the thought of this offends me. There is no such thing on a precisely engineered instrument.

@jereb31: I'm an engineer too, and I can tell you i'm always finding new ways to optimize CPUs which the developers never quite knew about. The Cell was almost it's own life force in that it was so advanced for it's time, the developers couldn't even fully understand it's power and capabilities.

So all the bullshit you and other lemmings ran about DX12 and the cloud were lies.? You did knew it would do nothing and you continue to use secret sauce even that you knew it was x86 and that its simple easy to understand..hahahahahaa

Nice to know..

2013 CPU is more efficient than a 2005 CPU hold the fu**ing presses...

Do you even look at your pathetic biased argument before you type.? You hate for Cell in particular is quite pathetic man.

Cell is a 2005 CPU even full blown Intel CPU back on 2006 kiss Cell ass when it came to compute,is the reason you so patheticly use Nvidia GPU vs Cell in Folding home because CPU lick sony's boots back then,on 2006 there simply wasn't a CPU capable of handling GPU task as Cell or crunching those numbers Cell did,so true it was the Army got many and in the medical field they were quite popular to.

Give fu**ing credit where is do man a Jaguar barely beating Cell in 2013 is not an achievement,fact is on 2005 when Cell was completed there was nothing like it nothing,and no CPU held a candle to it compute wise,so true it is that 6SPE now can beat the jaguar inside the xbox one,a old CPU which started development in 2001..lol

@Friendlyfire53:

It waasnt advanced for its time, it wasnt even really the first of its kind. Read that 3rd article i posted.

Yeah, I don't believe you and neither does the articles i posted. The Cell was a Single core processor driving 7 super specific weak cores. So specific in there use that they could often never be used.

Nothing secret about a micro controller, no secret sauce, no "life force". And certainly dubious about you being an engineer who believes in magic processors.

You not an engineer of homeopathic medicine are you XD

it's no longer a relevant argument since both consoles use the same architecture, same name brand APU, and are basically computers

The cell was like a CPU and DSP combined not CPU and GPU combined. The first CPU and GPU fusion was XCGPU in Xbox 360 S in mid-2010.

GPU's definition was defined by NVidia.

SPU is missing fix function and other GPU hardware features e.g. GpGPU usually has thousands of registers and SPU only has 128 registers.

SPU is not even unix type CPU nor GPU you fool. SPU will fail the simple 68000 test i.e. SPU lacks the supervisor and user mode instructions. IBM even stated SPU is "DSP-like". IBM >>>>>>>> YOU

SPU can not even self host it's own operating system e.g. needs PPE or X86 (for Toshiba's quad SPU solution).

SPU has DSP like branch feature i.e. weak branch.

X86's LEA instruction can load an array of data which is nearly akin to GPU's gather instruction. LEA instruction can do math operations.

The advantage of using X86's LEA over a series of MUL and ADD operations is primarily code size, as LEA allows for several distinct operations (e.g. multiply/add, or multiply/subtract) to be combined into a single compact operation. This is an example of CISC's advantage. It should sound familiar to GPU's MIMD (multiple instructions, multiple data) designs. Intel and AMD spent multiple billions of $$$$ to get this instruction faster and hardware optimized. This instruction is usually covered by AGU and ALU in AMD CPUs and on K8, there's 3 set of ALU+AGU combo.

Intel Core 2 Duo can act like a GPU e.g I can run Quake 3 with playable frame rates over OpenGL-to-Direct3D API bridge over Swiftshader 3.0 i.e. pure software Direct3D 9c render with similar LLVM software technology as ATI/NVidia LLVM based drivers. PS3 Linux run Quake 3 via software OpenGL on SPEs.

My early OpenGL-to-Direct3D API bridge over Swiftshader test with Quake 3 OpenGL. I recompiled OpenGL-to-Direct3D API bridge with Intel compiler.

The CPU has other things to do than render raster graphics i.e. this workload should be off loaded to a GPU.

AMD Jaguar has X86's CISC compact instruction advantage and 128bit hardware Intel AVX improvements.

@undefined: lol just no

FYI, GTX 8800 has DX10.X i.e. NVIDIA's DX10.0 with DX10.1 like extensions while not a proper DX10.1. DX10.X can be accessed via NVAPI.

GFLOPS comparison for 2006:

PS3 - 451.2 GFLOPS

• Cell - 204.8 GFLOPS
• RSX - 246.4 GFLOPS

Xbox 360 - 355.2 GFLOPS

• Xenon - 115.2 GFLOPS
• Xenos - 240 GFLOPS

PC - 430.6 GFLOPS

• QX6700 - 85 GFLOPS
• 8800 GTX - 345.6 GFLOPS

The Cell processor itself (not the cutback version on the PS3) is a powerhouse processor.

The problem was that it wasnt designed to be used in a console or function like a traditional PowerPC core.

That was the issue.

Even so, the PS3s architecture held the processor back even further.

The cell is an amazing processor for processing data, encoding/decoding data, maybe math calculations, and just crunching numbers.

This is because the cell SPEs were in-order processors and could only handle one task. Game development functions better with out of order processing, where instructions or what the CPU is needed for constantly changes.

This means that unless you give the cell specific instructions, the SPEs could sit idle, which is wasted cycles for a developer.

EX.

A,B,C,D will represent cycles.

Cell will execute, in order, ABCD. If for any reason B isnt ready to go, that will sit until the next cycle.

With a traditional processor, B would get picked up by the next free processor instead of having to wait.

This is what made game development so hard and long for PS3. Devs had to always make sure the processor SPEs were always busy with a task, extremely hard with only 1 PowerPC core giving instructions.

Devs had to offload single tasks to each SPE, which means breaking up or rewriting game code.

It just wasnt built for gaming.

(I know this is an extreme over simplification but I have to try to break it down for those who arent as familiar with IT)

You are using single precision values for the cell processor and double precision for the pc cpu.

Edit: whoops maybe not,