SO this is what I am getting out of the 360 and PS3...

In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.
 
Shifty Geezer said:
In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.

But then when dedicating 48 ALUs to pixel shading, geometry will suffer, and CELL+RSX will be handling a whole lot of it in top conditions.
 
london-boy said:
Shifty Geezer said:
In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.

But then when dedicating 48 ALUs to pixel shading, geometry will suffer, and CELL+RSX will be handling a whole lot of it in top conditions.

between two extra cores and a tesselator on the xenos parent... im not sure if i agree...
 
3roxor said:
And in the end it's still most likely that the Ps3 will have the most impressive graphics.
Based on what? No system will have more superior graphics than the other. We will see Xbox 360 games that look phenomenal and PS3 games that look phenomenal.
 
blakjedi said:
london-boy said:
Shifty Geezer said:
In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.

But then when dedicating 48 ALUs to pixel shading, geometry will suffer, and CELL+RSX will be handling a whole lot of it in top conditions.

between two extra cores and a tesselator on the xenos parent... im not sure if i agree...

Geometry will suffer compared to not using 48 ALUs all for pixel shading.
 
3roxor said:
And in the end it's still most likely that the Ps3 will have the most impressive graphics.

And how do you come to that conclusion?

My guess would be #1 whoever has the best artists will have the most impressive graphics.

On top tier games, 1st party exclusives, it will be a wash. It is almost always a wash. No one need look any farther than RE4, StarWars, MP2, and Zelda on the GCN (usually considered the weakest of this gens systems).

And by all means this generations hardware is closer in performance than last generations.

And from a hardware point of view, my next guess is that with accellerated Z, Alpha, Stencil, FP10 HDR effects, and little performance hit for 4xMSAA at HD resolutions the medium will be higher, at least early on.

And if the C1 is as effecient as ATI indicates (and as the general design would make one believe, minus any booboos) it is not clear that the RSX is more powerful. Not having any way to benchmark either chip--let alone we do not know all the details on either design!--it is hard to say. It surely is not as easy as counting transistors (e.g. the 6800GT has ~40-50M more transistors than the X800XL) or theoretical peak performance in GPUs. And the designs are so divergent it is not as easy as comparing them side by side. It just doesn't work that way.

That is the very reason why some SuperComputers have better realworld performance than competing SuperComputers with more maximum theoretical peak performance. Peak numbers, and realworld numbers, are completely different animals.

Anandtech gave a good example comparing a P4 to an Athlon 64. The P4 has a significant edge in theoretical performance in the ALUs, but the Athlon kicks it butt.

Right now, without seeing REAL software developed from the ground up with FINAL hardware at hand it is hard to make any statements about realworld performance. We probably wont get a good idea of that until 2007.

Remember, developers just got Beta 360 kits this month (without XeCPUs either!) and developers are still working on early CELL dev kits. There is a big reason why 2nd/3rd gen software looks leaps and bounds better.

In general I think it would be fair to say that both designs have their pluses, and both designs have their hurdles. The PS3 should outperform the 360 in floating point, and developers should have an easier time getting traditionally GP code running on the 360.

Both systems post developers with the huge problem of serializing their code to work in threads, 6 and 9, respectively. Breaking your game code up into 9 threads is not going to be easy. BUT, that is the ONLY way to maximize the performance of the PS3 (9 HW threads) and 360 (6 HW threads). There is no magic way to get at all that power on each system. In that regards the PS3 has a larger top end for floating point performance, but also requires a developer to work with 8 distinct cores, while on the 360 there are 3 cores. One could look at it as 3 cores with ~38GFLOPs theoretical each versus 8 with ~27GLOPs theoretical (very CRUDE and simplified way of breaking it down). Will less threads be better than more threads? Who knows at this point.

But that all goes back to the issue of GFLOPs, which by all means is not the only important metric. As poor as the XeCPU may be at GP code compared to an x86 desktop chip, the CELL is even worse.

Both systems face serious challenges, and I am not sure anyone can say which will be most impressive.

Overall I think who the most impressive will be who has the best artists, most time, largest budget, and best tools to work with. That means top games on each platform will look great.
 
london-boy said:
The fact that Cell might handle vertex data doesn't necessarily mean that the RSX vertex shaders would just sit idle. The VS could very well handle geometry of their own on top of the information Cell is sending it. It's not an exclusive deal, both can work on geometry at once, therefore there is the potential for having very high polygon counts since both are calculating triangles.

For example, Cell might be handling geometry for the environments and RSX handles the characters. Still better than RSX hadnling everything on its own, which can be done, it's all up to the developers really.

Also, Cell might just be helping RSX with cloth, water and hair animation, leaving the VS free to handle all sorts of other aspects in the scene. Certainly not sitting there idle while Cell finishes.

True, but my point was that the RSX does not necessarily have more Pixel Shading power (as fuod said). Even if the Vertex Shaders do not sit idle, the C1 can dedicate the majority of its time to Pixel Shading if the CPU is doing the majority of the vertex lifting. The RSX cannot dedicate its Vertex Shaders to Pixel tasks, so the "peak" performance of the RSX cannot be all counted toward Pixel Shading.

Obviously I mispoke on the technicality of the vertex shaders sitting idle--sorry. :oops: That is not necessarily true as you point out (but I think it would be a danger).

Good points, especially animation btw

My only concern with using CELL and the vertex shader units as outlined above would be complexity. Devs have very limited resources, especially time. Anything that is hard or too unique tends to get ignored :( And if the CELL is doing the world and the RSX characters you run into the issue of Load Balancing because there will be some big peaks and valleys in different scenes. Without very smart engine design, game design, and balancing you could look at under using your hardware because a scene with limited scenery would have SPEs fairly idle, and scenes with very few characters result in the RSX VS units fairly idle.

That is the goal of a Unified Hardware Shader Array => to dynamically loadbalance vertex and pixel shading tasks to get more performance out of the chip.
 
Personally i'm not sure how capable the extra core and a half are at geometry on X360. (one core dedicated to game data and half core dedicated to sound and IO)

It's all nice and easy to say it can handle all geometry, but i'm not sure it will be able to handle that much geometry, and surely not ALL geometry in a scene. That's just too much.

In that sense, Cell has double the floating point overhead to handle geometry in case it needs to. Which means a lot more geometry, if the GPUs only have to do pixel shading. In which case, X360 might have more pixel shading potential, but PS3 will definately have the geometry advantage.

Having said that, i expect devs to go the easy way and either let the GPUs handle geometry, or share the loads. Cell does, in paper, look a lot more capable than XCPU at handling geometry, if need be.

Developers already said they will only be using one core with the first batch of games, so we won't be seeing the results for a while.
 
Shifty Geezer said:
In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.

And history gives us a lot of reasons NOT to believe the numbers from either camp. We can make educated guestimations, but to proclaim anything as "iron clad" at this point about who will be better overall is pretty premature.

A very trustworthy developer told me that in the end a lot will depend on how quickly they can get a handle on multithreading games and that it is a huge hurdle on both sides. And that we will see people claim their system is underutilized or not enough effort is put into certain design features (as is always the case). In the end whatever system, if either or both or one or the other, lends itself to getting a handle on the hurdles they face quickest so they can make the games they want will result in better end products sooner. It is not just an issue of power this gen. Multithreading is a huge challenge, and adding additional complexity to the equation seems unliklely for most developers from what I have heard (which is not much).

But I think most agree that 1st party software will really exploit the final HW pretty well, we will see some nice 3rd party games, and a LOT of devs will bite the big one early because of the large leap from last gen to this gen expectations.

Seeing as the GCN did some great stuff this gen I am not sure one system pulling ahead here or there is that important for 95% of titles (exclusives are another issue altogether and are designed with specific system advantages in mind). So whether one can do more Pixel Shader ops or vertex shader ops wont change the end result much because both systems are very similar.

MS and Sony getting good tools to developers WILL be a bigger factor, in the long run, IMO. and ultimately the similarity of the designs is good because it means a more stable financial model for many publishers.

Expect to see lots of shovelware! Yay o_O If I had one sentance to say about this gen it would be: Buy the console that has the exclusive 1st party games you want.
 
And how do you come to that conclusion?

I base this conclusion on everything I've seen, heard, and read (ALOT)
It's a rather bold "conclusion" but I'll stick to it untill proven otherwise..

I'm not so naive to think that everything Ps3 related as seen on E3 was realtime (Killzone and Heavenly Sword were running 5 fps and speeded up towards 60? Fps) but I'm willing to give Sony the benefit of the doubt. We all could have read that atleast 1 developer (who visits this forum) is rather confident that when they are done it will run smoothly. I also read the same thing about Killzone. motorstorm also looked Kickass b.t.w.

On the other hand I've also seen playable games for the Xbox360 which all could be done on some high end PC.
 
3roxor said:
And how do you come to that conclusion?

I base this conclusion on everything I've seen, heard, and read (ALOT)
It's a rather bold "conclusion" but I'll stick to it untill proven otherwise..

I'm not so naive to think that everything Ps3 related as seen on E3 was realtime (Killzone and Heavenly Sword were running 5 fps and speeded up towards 60? Fps) but I'm willing to give Sony the benefit of the doubt. We all could have read that atleast 1 developer (who visits this forum) is rather confident that when they are done it will run smoothly. I also read the same thing about Killzone. motorstorm also looked Kickass b.t.w.

On the other hand I've also seen playable games for the Xbox360 which all could be done on some high end PC.

Comparing mostly renders on PS3 (with the exception of HS) to games running (at more than 5fps) on last generation's PC hardware is kinda far fetched dont you think? The most impressive next generation game (actual game running at more than 5fps) has been universally hailed as Gears of War.... so in reality your perception is inverse to reality.

BTW everything youve seen is running on high end pcs...

Which doesn't make your opinion wrong just not as informed as you make it out to be.

No doubt PS3 will generate amazing games but i dont think you can PROVE that based on what youve seen to date.
 
3roxor said:
And in the end it's still most likely that the Ps3 will have the most impressive graphics.

This is true. wait for GT5 to be shocked. :oops:
After GT4 on ps2, it will be impossible for an average gamer to differentiate between real TV footage and GT5 graphics. :oops:
 
1/ The CELL is 100 GFLOPS/s more powerful than the XeCPU of xbox360. And this is HUGE difference.

And the PS2 CPU did 2x the theoretical peak FLOPs compared to the Xbox. This did not prevent the Xbox from doing fine in these situations. And less we forget, floating point is not the end-all be-all aspect of code, it is an important part, but applications require more than FP performance (or Intel would have gone this direction a decade ago)/

Conversely, the Xbox CPU was far superior in general purpose code--yet the PS2 CPU did more than adequately in these areas.

Simply a single metric does not define the strength of a system, especially a closed box system.

2/ the bandwidth of the main ram of xbox360 is shared between the GPU and the CPU ( 22.4 GB/s shared ), and this is the biggest ERROR committed by the designers of xbox360. this made the xbox360 an unbalanced systme. The CELL of PS3 has dedicated 25.6 GB/s bandwidth, and this is a huge difference between PS3 ad xbox360.

Yet many reagard the 360 UMA as a strength.

And once again, the systems are designed quite differently. Framebuffer, specifically backbuffer, operations can easily saturate your memory bandwidth.

PS3.
Code:
Memory
   |
Memory Tasks + framebuffer

Xbox 360.
Code:
System Memory  <>          eDRAM
   |                           |
Memory Tasks            framebuffer

They are simply different approaches. Sony went for a large aggregate system bandwidth pool; MS went for isolating the largest bandwidth need, backbuffer, into a separate pool of memory.

As Shifty has noted, in simplified terms, the difference is brute strength (Sony) versus elegence (MS). In the end I would guess that we will find that each design has strengths and excells in different gaming situations.

To put it in simple terms for you: What use is a large memory pool if the bandwidth is saturated by the very small backbuffer? Basically you get a large pool of memory underutilized because a small amount (~60MB) is consuming most of the system bandwidth.

Memory bandwidth will be an issue on both systems, in the end they are different solutions to dealing with the same problem.

3/ the connexion between the CELL and the RSX is 35 GB/s, unlike xbox360 where the connexion is almost 22.6 GB/s ( i dont have exact number ) and this is a HUGE DIFFERENCE.

And yet that connection is, again, not an apples-to-apples.

The RSX has 35GB/s / 42GB/s to CELL+XDR. RSX has to balance XDR memory access and CPU interaction. This is not a bad thing, but it is very different compared to the Xenon-C1 link.

Remember, C1 does not need to bother the bus/memory bandwidth with its backbuffer.

4/ PS3 could do a lot more pixel shaders than xbox360, and this too a huge difference for genius developers.

Are you sure about that ;)

The 360 can dedicate all its resources to Pixel Shading, RSX (based on the little we know) cannot. Having the Xenon CPU tesselate HOS and directly feed the information to the GPU to shade them could create an environment where the GPU is dedicated almost entirely to Pixel Shading.

While the PS3 can do this with CELL-RSX, that would mean that the RSX VS sit idle. i.e. Some of the RSX performance would be neglected and useless.

And there is always that issue of effeciency. The C1 has a very different approach to shading, using an array instead of fixed pipelines. ATI is confident that this will be more effecient in the end than a traditional design.

5/ the PS3 has BLU-RAY capable of 50 GB ( dual layer ), xbox360 has norma DVDs capable of 9 GB, and this is more than huge difference, because this will allow exclusive ps3 games to be a lot more bigger with more extras, CGs, and contents than exclusive xbox360 games. ( GT5 the best example here )

There are obviously positives to large media.

But in the end: How many developers will have the money to fill a 9GB DVD, let alone a 25GB BR disk?

And there is the issue of a HDD standard on the 360.

Overall most of your points are mangled and contrived to push forth your agenda of "PS3 > 360". Having an opinion is fine, but so far every post has been the same. e.g.

because those games just cant be made for xbox360, xbox360 doesent have the power to run them, so the experience of gameplay will be totally different betweeen exclusive well developed ps3 games, and exclusive well developed xbox360 games.

the main difference in next generation games will be not in graphics quality, but on animations, physics, and AI. And the PS3 is far better in those domains than xbox360, and you will experience this by yourself in the end of 2007 ( the photo realistic GT5 than MGS4 with a full simulated world by ps3). PERIOD

You claim the above, without any proof in final software.

I am glad you are a fan of the PS3, and by all means enjoy your future purchase, but your posts lobbying for the PS3 and misrepresenting the market is tiresome. Some advice: If you want to stick around much longer, change your tune or you may find out that the mods really dislike trolls. Just some friendly advice.[/quote]


Acert93 said:
fouad said:
I see them as more similar than disimilar.

-3.2GHz PPC core(s)

XBOX 360 CPU will be only running at 3.00 GHZ, unlike the CELL 3.2 GHZ.

And where did you get this news? :rolleyes:

The officially specs are for a tricore custom PPC CPU operating at a frequency of 3.2GHz. This is what MS announced, this is what everyone is reporting, and there has been no official (nor unofficial) statement otherwise.

So far your post has not started off on a good first step...

Unfortunately I cant tell you where I got this ( NDA ), but I could give you a glimpse if you want. And anyway you will discover by yourself when microsoft will announce final specifications of xbox360.

I dont have time to respond to the rest of your comments, but I will do it as soon as possible.
 
fouad said:
3roxor said:
And in the end it's still most likely that the Ps3 will have the most impressive graphics.

This is true. wait for GT5 to be shocked. :oops:
After GT4 on ps2, it will be impossible for an average gamer to differentiate between real TV footage and GT5 CGI. :oops:

Fixed it for you there guy.
 
fouad said:
I dont have time to respond to the rest of your comments, but I will do it as soon as possible.

When you do please use the "quote" feature. Without using this feature it is near impossible to keep track of who said what. You can do this manually by typing:

Code:
[quote="SPEAKER_NAME"]

What they said...

[/quote]
 
Sorry blakjedi, but you are completely wrong :!:

Comparing mostly renders on PS3 (with the exception of HS)

Why..because it happens to be that deanoC is the only developer who confirmed that here.... :rolleyes:


to games running (at more than 5fps) on last generation's PC hardware is kinda far fetched dont you think?

Ehm..with last generation PC hardware you mean those 2 Powermac G5 where the xboxBeta kit is currently running on... :rolleyes:

(actual game running at more than 5fps) has been universally hailed as Gears of War.... so in reality your perception is inverse to reality.

Yup at a fluent 15fps.. It also wasn't "universally" hailed as the most impressive next generation game.... so in reality your perception is inverse to reality.

BTW everything youve seen is running on high end pcs...
"Comparing mostly renders on PS3" = you :?

Which doesn't make your opinion wrong just not as informed as you make it out to be.

says you :LOL:

No doubt PS3 will generate amazing games but i dont think you can PROVE that based on what youve seen to date.

Not in the court of law "but I'll stick to it untill proven otherwise.."

try again[/b]
 
Shifty Geezer said:
In terms of pure pixel shading though, RSX will have 24-32, whereas Xenos has 48 (in simple terms). The key point is that where RSX may have overall a shader strength advantage (perhaps), Xenos can dedicate all it's resources to pixel shading which will likely outperform RSX's pixel shading, is numbers are to be believed.

Well last time I looked, even raw figures for just pixel shaders on RSX were higher? (but they may not be directly comparable, taking into account texture addressing etc. but at least performance need not be sitting idle on RSX when texture addressing isn't needed).

That aside, there is a clock difference and an unknown, but highly likely, performance degradation on unified shaders vs a dedicated shader. It's a little complicated to weigh up all the plusses and minuses..

And I think your geometry performance would drop below what Cell + VS could offer. Asides from the tesselator, which can't be used for vertex shading, how much power can you spare on Xenon for such processing vs Cell+VS?

After Kirk's claim that some devs were pushing VS off Xenos and using it only for pixel shading, the more I thought about it, the more I consider that to be a very undesireable situation. It would suggest some bad things for Xenos, in fact, if that were more desireable in the typical case.
 
Back
Top