Support NeoGAF

[Drek]

Saw this yesterday, its blatant bullshit spin with more selective logic and fabricated numbers than you see in a republican national budget plan.

Seriously, at least Sony's spin uses real numbers, even if they are completely meaningless.

 

[GhaleonEB]

Sony started the PR Fud with concept videos and a claim that the PS3 is 2x the power of X360.

Now MS is just playing the same game of dishonesty. Can you blame them?

I mean really guys.....these consoles will have nothing between them power wise, with a slight edge in cpu power for PS3 but probably not noticable.

It's looking like:

360 GPU >>>>>> PS3 GPU

CELL >>>>>>>>>> 360 3-core

How that translates to game performance it seems people are still bickering about.

 

[Rorschach]

Well, sure. The fanboys will believe everything that says good things about their system of choice or bad things about other systems. That's the nature of the beast, and frankly nobody cares about them - they can go rot.

Wavering casuals with no hardware knowledge may be swayed by this sort of thing to a certain extent, but they're fickle creatures - eventually it'll come down to what the games look like, and it'll be the machine that renders the most ludicrously misproportioned breasts that wins, as it should be.

Haha touche.

Those casuals you're referring to rely on visuals not numbers. Sony knows how to handle that properly, as they showed with their E3 presentation.

I can agree with that, but just to make it clear, when I was referring to "casuals" I meant as a result of word-of-mouth. I don't think many casuals visit game sites or whatever, but everyone has that 1 friend that is into games and he's going to go to school tomorrow talkin' 'bout teh pwr!

 

[Agisthos]

It's looking like:

360 GPU >>>>>> PS3 GPU

CELL >>>>>>>>>> 360 3-core

How that translates to game performance it seems people are still bickering about.

Yes. The only unbiased info on the whole web so far is a few sparse statements by that Deano guy doing the Heavenly Sword game and he seems to think the CPU's will be pretty close. It is a mistake to think the Xgpu will be better than RSX though.

I mean what is everybody expecting with two consoles launching only a few months apart?

I think the big dissapointment is for the people really believing that the PS3 was going to be so much more powerful. The MS document is also just FUD. Inbetween lies the truth

 

[gofreak]

It's looking like:

360 GPU >>>>>> PS3 GPU

It certainly isn't.

few sparse statements by that Deano guy doing the Heavenly Sword game and he seems to think the CPU's will be pretty close. It is a mistake to think the Xgpu will be better than RSX though.

You're taking him out of context. He was talking about one specific task, not how they'd match up generally across all game tasks. There's no doubt in my mind that Cell, if tapped properly, is a much more powerful chip for games than X360's. The headroom there to use it for graphics is a boon too.

 

[Forsete]

It certainly isn't.

Cant someone just kill these myths once and for all? Jesus this is annoying, false info being allowed to spread around like this.

 

[ram]

maybe this article meant: xbox 360 outpowers playstation 2? :lol

 

[Pug]

GOFREAK That stamement is as bad as some of the crap in that spec run down. Cell is certaintly better at certain situations for games. The try core is better in other areas for games. Cell is in no way better across the board for games, thats is completely wrong.

 

[DarienA]

You're all so quick to jump on the "Microsoft is spinning", "Microsoft with their damage control" bandwagons, but have we seen any actual PS3 games RUNNING and being PLAYED?

No.

Honestly, you may view it as spin, but I think they're being very honest with some of these charts.

It's not Microsoft's fault that Sony isn't delivering.

:lol you rock Mike.

 

[Elios83]

If Microsoft authorized the article then they are really scared by the negative feedback received at E3.The poor Allard was only asked everywhere about the hardware inferiority and about the games that don't look really next gen.
As for me I'm only waiting Sony/nVidia to release full PS3 specs and to see real games running on both the consoles.

 

[gofreak]

GOFREAK That stamement is as bad as some of the crap in that spec run down. Cell is certaintly better at certain situations for games. The try core is better in other areas for games. Cell is in no way better across the board for games, thats is completely wrong.

On balance, I think it will be. Not necessarily at every task, but not all tasks are created evenly in terms of their load on the processor. Next-gen games will be physics/AI and simulation heavy..or at least they should be, right? Those things, particularly physics, would take up a large proportion of frame computation time. If you can accelerate that, even if you're slightly slower elsewhere, in the end you'll still end up faster overall.

We're also seeing comparisons between X360 cores and SPEs. No doubt X360 core has higher integer performance, for example, than a SPE. But higher than 2 SPEs? Even forgetting about the PPE, which should be roughly as good as one core on X360, Let's remember there's over twice the number of SPEs on PS3's CPU as there are cores on X360. Even if a SPE is only half as good as a X360 core at integer ops, two of them will have twice the floating point performance and the same level of integer performance. Let's match our PPE up against one core in Cell. Then you'd could compare 1 X360 core with 3 SPEs - the SPEs could afford to be only have a third of the "general" or integer performance of a core, and still level with it on that and have 3 times the floating point performance to boot. Sure, you have to split your task between them then - it's not as easy - but like I said, if tapped properly..

 

[Monk]

Gofreak said what I wanted to say best.

 

[Agisthos]

You're taking him out of context. He was talking about one specific task, not how they'd match up generally across all game tasks. There's no doubt in my mind that Cell, if tapped properly, is a much more powerful chip for games than X360's. The headroom there to use it for graphics is a boon too.

I know what context is. And the context was about whether the XCPU could software render the Getaway graphics like Cell can

"XeCPU has 3 VMX128 cores at 3.2 Ghz, at worst it could get within 30% of Cell doing the same job, because the cores of Xenon are more flexible then SPU in all likelyhood it could achieve results much closer than that.

Cell has a FLOP advantage, XeCPU has a flexibilty advantage... I suspect a tuned advanced software engine for both would be within 70-80% of each other. I'm not even sure that if you have lots of vertex and texture data, that XeCPU would lose..."

It is only the multiplatform developers which will tell us how much more powerful the Cell will be, not marketing BS

 

[Kuroyume]

Poor PS3 fans, kings of damage control...

 

[Pug]

Goifreak, I agree with your figures however we have no idea how well efficient the use of the SPE's will be and what I was trying point out that loads apllied across the CPU for games will be different depending on the game. And whilst Cell may be more favourable in a lot of cases it will not be across the board which is what you implied.

 

[JMPovoa]

We will problably have a bigger gap in graphics between games produced by the TOP Developers and the rest on the PS3 than on Xbox 360.

But anyway, PS3 games developed by TOP Developers will problably look better than Xbox 360's TOP Developers' games.

 

[SolidSnakex]

But anyway, PS3 games developed by TOP Developers will problably look better than Xbox 360's TOP Developers' games.

 

[Nostromo]

It's looking like:

360 GPU >>>>>> PS3 GPU

You asked to prove why Major Nelson stuff is flawed, I proved it to you, you ignored me and you're still writing crap like that. Nice :lol

 

[Cerebral Palsy]

GAFer: This report is a lie! *inserts own madeup reasoning why PS3 is more powerful*

You people are sad.

 

[gofreak]

I know what context is. And the context was about whether the XCPU could software render the Getaway graphics like Cell can

No, Deano was talking about software rendering generally - and specifically singled out rasterisation, something not being done on Cell in the Getaway demo. He wasn't talking about whether what was being done on Cell in the Getaway demo (lighting, texturing etc.) was as doable on X360, or how doable it was versus Cell, he was looking at the general case.

Goifreak, I agree with your figures however we have no idea how well efficient the use of the SPE's will be

True, but we don't know the same re. X360's cores either. We can really just take them at face value for now.

what I was trying point out that loads apllied across the CPU for games will be different depending on the game. And whilst Cell may be more favourable in a lot of cases it will not be across the board which is what you implied.

I'll agree with that. Games that skimp on physics in particular, and AI to another extent, might run as well or better on X360 as PS3 But seriously, it's hard to come up with a "typical" next-gen game and what it's distribution will look like. I think if you look at the things people really want a step-up in, Cell provides more potential for those things (imo). People say "yeah yeah, physics, AI, so what" (well, you know what I mean), but I think those are two key things people want a leap in. Epic is saying PS3's CPU is in the same ballpark for physics as a physx chip. Where is X360's advantage? Are the computational requirements of those advantages as high as for those Cell is better at? People are talking about a nebulous concept of "general processing"...I'm not even convinced it has a "general processing" advantage, or at least an integer processing advantage, overall (on a per core level maybe, but not overall).

 

[Panajev2001a]

No, Deano was talking about software rendering generally - and specifically singled out rasterisation, something not being done on Cell in the Getaway demo. He wasn't talking about whether what was being done on Cell in the Getaway demo (lighting, texturing etc.) was as doable on X360, or how doable it was versus Cell, he was looking at the general case.

For example we showed the demo that renders London City, it's not rendered in the GPU but the CELL does lighting and texture processing then outputs it to the frame buffer. Even without GPU, only CELL can create good enough 3D graphics.

It outputs to the frame-buffer, that would seem like doing rasterization too.

 

[gofreak]

It outputs to the frame-buffer, that would seem like doing rasterization too.

Ahhh, interesting. I didn't catch that.

Still, I don't think anyone's going to be using Cell for rasterisation in games generally given the GPU's presence Lend a hand where it makes sense? For sure. It could be a significant helping hand for certain things, I'd think. BTW, I'd love to hear your thoughts on that kind of co-operation Pana Your general analysis of the system will be more than appreciated when it's done!

 

[Panajev2001a]

Sony started the PR Fud with concept videos and a claim that the PS3 is 2x the power of X360.

Now MS is just playing the same game of dishonesty. Can you blame them?

Yes, I can blame them, PlayStation 3 over-all IS more powerful and Allard post E3's Press Conferences admitted it too basically.

There is a power greater than X.

 

[Panajev2001a]

Ahhh, interesting. I didn't catch that.

Still, I don't think anyone's going to be using Cell for rasterisation in games generally given the GPU's presence Lend a hand where it makes sense? For sure. It could be a significant helping hand for certain things, I'd think. BTW, I'd love to hear your thoughts on that kind of co-operation Pana Your general analysis of the system will be more than appreciated when it's done!

Keep the hype low, I do not get paid for this stuff, so I take a long time doing this: if people keep this thing going, by the time I am done, I will be expected to have also programmed some kind of FutureMark application for both consoles .

 

[gofreak]

Keep the hype low, I do not get paid for this stuff, so I take a long time doing this: if people keep this thing going, by the time I am done, I will be expected to have also programmed some kind of FutureMark application for both consoles .

Lol, don't worry, I shall refrain from mentioning it going forward. I'm not expecting it at all..

 

[madara]

Six pages!?? Gads is gaming even about games anymore?

 

[Brimstone]

Cause they are comparing R500 internal bandwith with PS3 external bandwith.
Do you want me to calculate SPEs bandwith to local sram? register files bandwith?
256 Gb/s my ass, I'm talking about terabytes/sec here :lol
Obviously it does't even count total external bandwith at all! even if we know RSX can read/write from CELL's xdr ram!
It's a flawed comparison, period.

The internal bandwidth of the register files vanishes if the SPE has to get information out of main memory. This is put into context by the article.

Sony's CPU is ideal for an environment where 12.5% of the work is
general-purpose computing and 87.5% of the work is DSP calculations. That
sort of mix makes sense for video playback or networked waveform analysis,
but not for games. In fact, when analyzing real games one finds almost the
opposite distribution of general purpose computing and DSP calculation
requirements. A relatively small percentage of instructions are actually
floating point. Of those instructions which are floating-point, very few
involve processing continuous streams of numbers. Instead they are used in
tasks like AI and path-finding, which require random access to memory and
frequent branches, which the DSPs are ill-suited to.

Based on measurements of running next generation games, only ~10-30% of the
instructions executed are floating point. The remainders of the instructions
are load, store, integer, branch, etc. Even fewer of the instructions
executed are streaming floating point-probably ~5-10%. Cell is optimized for
streaming floating-point, with 87.5% of its cores good for streaming
floating-point and nothing else.

Yes, you can add up the bandwidth of the register files and such for a perfect streaming enviroment, but what is argued is that game code doesn't fit into this perfect enviroment.

 

[Panajev2001a]

The internal bandwidth of the register files vanishes if the SPE has to get information out of main memory. This is put into context by the article.



Yes, you can add up the bandwidth of the register files and such for a perfect streaming enviroment, but what is argued is that game code doesn't fit into this perfect enviroment.

SPE's are not DSP's and while the latency that a Software Cache Hit would be able to give is not ultra-great, it is possible that compilers and some opportune hand-optimization of SPU code could hide it to a semi-acceptable level at least (certainly much better than going to main RAM). An SPE with say 128 KB of LS and 32 KB of SL1-ICache and 96 KB of SL1-D-Cache might not be as horrible as you might think with general purpose code IMHO.

SL1 = Software Layer 1

 

[gofreak]

The internal bandwidth of the register files vanishes if the SPE has to get information out of main memory. This is put into context by the article.

The R500's daughter chip needs to get its data from somewhere too - it doesn't work in isolation, and it's likely not going to have all the data for the entire frame locally at any one time. Does its bandwidth also disappear when you have to "go outside" for data? You'll likely be tiling the frame into the eDram.

If you wanted to consider a like-for-like task with Cell, it too could tile the framebuffer into its local memory and work on the frame using its internal bandwidth. You'd need to do more tiling, since it has less local memory than the eDram has, but the bandwidth requirements of tiling is small versus that of the internal operations on the data.

 

[duckroll]

Six pages!?? Gads is gaming even about games anymore?

Nope. The best next-gen game will be the one where players connect online via either the Xbox360 or the PS3 and combat with other players with their stats being equavilant to their console's hardware specs. You will be able to equip various types of spin to increase your stats or lower that of you opponents. I hear the prototype title for the game will be called DAMAGE CONTROL WARS.

 

[Panajev2001a]

Nope. The best next-gen game will be the one where players connect online via either the Xbox360 or the PS3 and combat with other players with their stats being equavilant to their console's hardware specs. You will be able to equip various types of spin to increase your stats or lower that of you opponents. I hear the prototype title for the game will be called DAMAGE CONTROL WARS.

Laughing aside that would be pretty cool .

 

[callous]

You don't really need to read anything but the final conclusion to realize how much bullshit is being spewed here:

However, hardware performance, while important, is only a third of the
puzzle. Xbox 360 is a fusion of hardware, software and services. Without the
software and services to power it, even the most powerful hardware becomes
inconsequential. Xbox 360 games-by leveraging cutting-edge hardware,
software, and services-will outperform the PlayStation 3.

I lack the words ...

 

[CosmicGroinPull]

You forgot something:

However, hardware performance, while important, is only a third of the
puzzle. Xbox 360 is a fusion of hardware, software and services. Without the
software and services to power it, even the most powerful hardware becomes
inconsequential. Xbox 360 games-by leveraging cutting-edge hardware,
software, and services-will outperform the PlayStation 3.

Nelson 3:7

 

[COCKLES]

Phew.

I knew it all along.

SONY. PEDDLING LIES. WITH THEIR LUKEWARM PERFORMANCE. ON THEIR ALREADY SAGGING AND DYING PS3 DISASTER! GIVE UP NOW SONY. CONCENTRATE ON SAVING YOUR DYING ELECTRONICS EMPIRE! THIS GENERATION IS OVER! YOU GOT 360SERVED!

 

[McFly]

You forgot something:

Mwahahahaha ... and I just wanted to compare Nelsons reasoning with the reasoning of common creationists. :lol

Fredi

 

[SantaC]

omg clash of the ps3 fanboys and xbots :lol This is going to be a long generation.

people need to stop looking at the charts and start to care about the games.

 

[Doom_Bringer]

omg clash of the ps3 fanboys and xbots :lol This is going to be a long generation.

people need to stop looking at the charts and start to care about the games.

I agree 100%, this thread is pathetic.

 

[AndreasNystrom]

I just read a report where MS claimed the Xbox 360 to be faster then the NES also!.

Really. what is this bullshit?.. the total memorybandwith

 

[Cerebral Palsy]

omg clash of the ps3 fanboys and xbots :lol This is going to be a long generation.

people need to stop looking at the charts and start to care about the games.

I also agree. It will never happen unfortunately.

 

[McFly]

people need to stop looking at the charts and start to care about the games.

Give me final hardware and final games and I start to care about games. Right now there simple is no way to care about specific games, just about the possible power and features driving them.

Fredi

 

[Nostromo]

The internal bandwidth of the register files vanishes if the SPE has to get information out of main memory. This is put into context by the article.

Well..I bet R500 and X360 CPU can have data in their caches without gathering them from the external memory :lol
SPEs are good at streaming processing, this is fact, but a LOT of game code could be resorted to a stream processing programming model.
If it wasn't done before (cause it was not needed) doens't mean it will not be done in the future.
Moreover even X360 CPU will benefit from a streamlined processing.

 

[Hitler Stole My Potato]

God knows I don't want to start another thread about this idiotic system power war thing but I saw this blog from a MS employee that corrects some of the stuff that came from Major Nelson. Anywho, here it is:



Xbox 360 versus Playstation Performance - Corrections for Major Nelson

GENERAL CPU PERFORMANCE

Major Nelson's article states that the PS3 only has one general-purpose processor, giving the Xbox 360 an advantage because it has three general-purpose processors. I think this is not a fair comparison because the 7 SPEs (Synergistic Processing Elements) are general-purpose processors! Unlike the PS2, they are not purely vector processors and can handle general-purpose tasks. They are completely programmable in C/C++ and have arbitrary DMA access allowing each to take advantage of the 256 MB of main memory. I would give each SPE half a processor's. It doesn't have full credit because of low cache and a lack of branch prediction. This would change Major Nelson's chart to the blue one below.

MEMORY BANDWIDTH

I have no problems with Major Nelson's memory bandwidth chart. The EDRAM is a monster with 256 GB/s of bandwidth. The PS2 had embedded memory, and the PS3's lack of it is very much a cost cutting measure that will impact performance. It is just important to note that the 10 MB of ram is used solely for a complement to the frame buffer to eliminate the performance hit of anti-aliasing at 720p. My problem is that he shows this chart several times and never shows the speed comparison of the actual memory bandwidth that developers would use for their games. My chart (with blue bars like before) clears this up and shows the actualy comparison of memory bandwidth. The facts on bandwidth are here for you as well since this chart doesn't talk about bus speeds or I/O.

Xbox 360
22.4 GB/s memory interface bus bandwidth
256 GB/s memory bandwidth to EDRAM
21.6 GB/s front-side bus

PS3
Main RAM 25.6GB/s
VRAM 22.4GB/s
RSX 20GB/s (write) + 15GB/s (read)
SB< 2.5GB/s (write) + 2.5GB/s (read)


CONCLUSION

I have included the rest of the charts, and now with my corrections, this is really a solid comparison of the two system's core performance. Major Nelson did a great job in his article. I just really thought that the charts for general CPU performance and memory bandwidth comparisons in his article were not fair. Everything else was very accurate, and you should read the rest at his site. The article is in 4 parts.

LINK

 

[Argyle]

I'm not even going to bother talking about the two systems in this post - but do you remember those old ads Nintendo used to put in EGM back in the 16-bit era? The ones that they laid out to look like an article (you had to look closely for the "advertisement" notice) and basically tried to convince you (post Sega "blast processing" ads) that the SNES was much more powerful than the Genesis (this post is not about whether the ad was correct or not - so please don't derail this thread into SNES vs. Genesis )?

That's what all this reminds me of. I don't remember seeing ads like that in the last two generations so much...interesting that it's all coming back now

 

[Stinkles]

This is retarded. Both systems are a near technical wash (Moore's Law rearing its ugly head), and what's going to make that even more evident, is third party cross-platform support. EA games are going to look identical on each platform, Activision games are going to look the same. It'll be interesting to see subtle differences on exclusive titles - I would like to see the kind of "charismatic" technical differences we saw with SNES/Genesis, where you could tell which platform was which from things like color-palette and sprite performance.

 

[Pimpwerx]

This is retarded. Both systems are a near technical wash (Moore's Law rearing its ugly head), and what's going to make that even more evident, is third party cross-platform support. EA games are going to look identical on each platform, Activision games are going to look the same. It'll be interesting to see subtle differences on exclusive titles - I would like to see the kind of "charismatic" technical differences we saw with SNES/Genesis, where you could tell which platform was which from things like color-palette and sprite performance.

They are not a technical wash by any means. The end result of games may not evidence the gap quite as well, but there is one. I think it's already funny that people talk about PS3 having rendered demos when it could just as likely be evidence of the performance gap we will see. Notice, I'm saying "could" b/c I don't really know if Guerilla or even Ninja Theory can get to what they want. But I'd like to think devs aim for certain targets for their first gen titles, and don't try to shoot the moon on the very first outing. We'll see. In the meantime, we really need more info on RSX (it is not just a G70 dammit) and more details on Xenos (the block diagram is useful, but not the full picture). Because the CPU advantage is pretty moot. Cell will outperform the xCPU in most tasks, and by a significant amount for some of them.

The performance gap in GPU should be in RSX's favor as well, but by how much? And more importantly, what efficiency can either chip hope to sustatin at 720p+? Xenos already has to tile at 720p due to memory cap issues, and it's unknown if/how that affects its ability to fp16 blends or its fillrate. I'm pulling some of this stuff from ongoing discussions at other boards too, so I'm not really confident enough to make claims for or against. For perspective, it's been less than a week now since the forum got turned on its head. PEACE.

 

[VictimOfGrief]

Here's the problem....

"A document MSoft sent IGN..."

:lol

 

[MetalAlien]

This is retarded. Both systems are a near technical wash (Moore's Law rearing its ugly head), and what's going to make that even more evident, is third party cross-platform support. EA games are going to look identical on each platform, Activision games are going to look the same. It'll be interesting to see subtle differences on exclusive titles - I would like to see the kind of "charismatic" technical differences we saw with SNES/Genesis, where you could tell which platform was which from things like color-palette and sprite performance.

Well you can tell a PS2 from an XB1 game just from looking at it... LOL

But as for the PS3 Vs XB360.....

The PS3 has as much bandwidth to it's 512MB main system RAM as the PS2s GS (graphics synthesizer) has to it's 4MB on chip embedded ram.. That's something that comparison (linked in first post) failed to mention.

The XB360 has only half as much BW to it's main ram... only the frame buffer has access to the 256GB of BW... and even that isn't between the GPU and the EDRAM... it's between the EDRAM and it's on chip memory controller... LOL

 

[GhaleonEB]

Here's the problem....

:lol

They sent to lots of places, not just IGN.

 

[FiRez]

The PS3 has as much bandwidth to it's 512MB main system RAM as the PS2s GS (graphics synthesizer) has to it's 4MB on chip embedded ram.. That's something that comparison (linked in first post) failed to mention. L

WHAT???? :lol :lol :lol

 

[MetalAlien]

WHAT???? :lol :lol :lol

Laughter, at 512MB of ram available at around 50GBs of BW... Okay.