Yes! Because the difference can be seen by the human eye! It really can!

makingmusic476 said: superchunk said: So does this mean they won't launch a ps4, just tell everyone to buy another ps3 and connect them?!   They'll also be offering special deals on duct tape so you can stick them together.  :P

beat me to it.

 

@ selnor

I don't think it's from Microsoft, maybe some anonymous troll sending out fake Spam. Sure Microsoft likes to mislead, but that "email" is far over the top. I won't do a complete runthrough as I already addressed everything here on VGchartz before.

Just one for illustration:



Each SPE does have a tiny amount of cache, they all do have DMA units and a branch predictor  is not needed with properly designed SPE code (the programmer does so manually, like shifting gears, but more efficienty than a predictor can).

Why do processors need cache to perform properly? The answer is simply, operating speed, for working on data as fast as possible. One thing you will note when looking at each SPE's memory is that it's the same kind of RAM that is used for cache elsewhere. Just because this memory does share the limitations of cache when used, does not give other CPUs an advantage.



You wish... It's inaccurate, incomplete, biased and thus entirely useless.

MikeB said: @ selnor Each SPE does have a tiny amount of cache, they all do have DMA units and a branch predictor is not needed with properly designed SPE code (the programmer does so manually, like shifting gears, but more efficienty than a predictor can).

Its not cache, its like small ram for each SPE and you have program usage of that memory 'by yourself'(Ie theres no cache.). Uhm, try to do the ones with logical AI 'manually'. Logical AI is very hard to implement with SPE:s anyway(real time that is). On the other hand you could use PPE, but its also a bit limited when it comes to programs that use a lot of branches.

And wtf? Its not always better when programmer must do something manually. I bet for you whenever programmer must do something manually for X360 it will be crap and for PS3 it will be a frigging BLESSING. Just cut that biased crap. You seem to know something, but I(and I guess others too) can't take you seriously because you act like PS3 zealot.

Only useful for mainly AA? Hahaha.....

Sounds quite familiar especially when reading your posts... (heh...)

 

I challenge you to say something good about X360. :D

@ Deneidez

"Atomic Cache Unit

Never heard of it right?

Well its one of my favorite things on the PS3 and gets little love cos its one of those tiny features that make life so much nicer."

http://blog.deanoc.com/?p=96



It's extra work, but all game engine devs agree they are able to achieve superior results. Code designed for the SPEs will be a lot more cache friendly on other CPUs as well.

The Tomb Raider Underworld dev for example stated the Cell approach is superior to anything Microsoft proposes for both platforms. He does however acknowledge game assets wise and general game engine design for this game is totally geared towards the 360 (due to the 360 being on the market earlier) and thus will not get the most out of the PS3 (referring to a comparison with Uncharted: Drake's Fortune, a PS3 exclusive for which the devs explained to have an enormous amount of headroom for future games as well).



I said mainly not only, that´s what it´s currently mostly used for. However 720p together with AA already results into tiling, thus impacts framerates. 600p with AA can be done without really impacting framerates, thus a game like COD4 can run at 60 FPS on the 360.



Like I said before, it´s really cheap around here, 149 Euro (much less than a Sony PSP for example or a Wii) and I loved Kameo when I bought one a long time ago.

lol for people that can see things that fast it must really be annoying for them to play games in 30 or 60fps :P

MikeB said: @ selnor What I am posting is from a M$ e-mail from their own comparitive analysis I don't think it's from Microsoft, maybe some anonymous troll sending out fake Spam. Sure Microsoft likes to mislead, but that "email" is far over the top. I won't do a complete runthrough as I already addressed everything here on VGchartz before. Just one for illustration: The Cell's seven DSPs (what Sony calls SPEs) have no cache, no direct access to memory, no branch predictor, and a different instruction set from the PS3's main CPU. Each SPE does have a tiny amount of cache, they all do have DMA units and a branch predictor  is not needed with properly designed SPE code (the programmer does so manually, like shifting gears, but more efficienty than a predictor can). Why do processors need cache to perform properly? The answer is simply, operating speed, for working on data as fast as possible. One thing you will note when looking at each SPE's memory is that it's the same kind of RAM that is used for cache elsewhere. Just because this memory does share the limitations of cache when used, does not give other CPUs an advantage. his info certainly doesnt lie and is accurate You wish... It's inaccurate, incomplete, biased and thus entirely useless.

Your very wrong Mr MikeB. It is from M$. (and I trust their Boffins much more than Sony or you, becuse they have been working with this sort of hardware far longer than Sony or you). It's from IGN where IGN state the following "We decided to put this Microsoft-provided information up undigested. Thus, enclosed is a Microsoft-made comparitive analysis. We have not altered, added, or tinkered with the data. We have not interpreted the specs. We have cut and pasted this information from the email into our site and formatted it."So you can blabber about Cache all you want. Bottom line is I trust them more with this stuff. It's like either listening to Ferrari about performance of cars or Rover about performance of cars. One has been doing it alot longer than the other.

 

 

@MikeB

Hmm, interesting blog article and yes I haven't heard of it ever. I wonder why there isn't much information about it. Its really small(~4 variables), but still makes it all a bit different.

 

Sometimes whatever you do/how optimal you try to be the best result is when used automatical/already implemented stuff. This is especially the case when all programmers in project aren't gurus and as Groucho said theres only few of these guys. (You like the word superior, don't you?)

 

Can I get sources for all of these claims/articles you refer? (No, I don't imply that they haven't said such things etc. I just want to see what else did they say and what kind of stuff they are talking about. Superior... Hmm... I wonder did they really use such a word. :D)

 

And how about shaders for example? I am quite sure its stupid to use ram for shaders when you have ten times better eDRAM you can use to play around with those shaders. Or are shaders some sort of inferior part so you can't really count them? And yes you are right. One must use tiling at higher resolutions with AA and it will slow things down a bit.

 

Not sure can I count calling something cheap as saying something good about it, but I guess thats best I can get from you.

@ selnor

It does not take a rocket scientist to understand it's a completely bogus comparison.

For example 278.4 GB/s for the 360 (of which 256 GB/s is internal Xenos bandwidth being fed data (and moved back and forth if tiling occurs) from the main RAM much slower vs the 48 GB/s for the PS3 (split up for XDR and GDDR3, which can be used simultaneously).

Now add the Cell internal memory bandwidth of 307.2 GB/s. So not taking into account technical restrictions for both platforms it's:

~355 GB/s for the PS3 and 278.4 GB/s for the 360 (BTW most modern PCs have A LOT less bandwidth using such rediculous measures). This is entirely useless though for comparison sake let me provide you one very simple example.

Let's say you have a full PC on a daughtter board capable of achieving 40 GB/s in a full PC (also 40 GB/s total memory bandwith) and would limit the communication between them to 1KB per minute. Total system bandwidth using your way of thinking results in a 80 GB/s memory bandwidth system. I used an extreme example to get the point across, at 1 KB/minute data feeds this daughter board PC is completely useless.

@ Deneidez

Some of the dev quotes from e-mpire forums:

"asset-wise 360 was around first, so we made stuff keeping the 360 in mind first."

"Well, that all depends on your definition. Writing code optimized for the PS3 and using threading policies that are suited the SPUs is a given, because not doing so would not be acceptable at all. All our multithreading is done on PS3 first without exception, and other platforms emulate SPURS."

"Secondly, the matters of multithreading policies, the whole job queue architecture, encapsulation of jobs and their corresponding data packets, etc. that work on the PS3 are indeed more than applicable of the 360/PC. And as I've mentioned before, they work better than anything and everything that Microsoft recommends (so far without exception for us). The problems lie in the fact that that work is an absolute necessity on the PS3, whereas they're not entirely necessary on any other platform."

The eDRAM approach is much cheaper than having seperate RAM with seperate buses. It's not an approach used for PCs because eDRAM becomes expensive if you want to have enough of it (which is not the case for the 360, but would be a requirement for a gaming PC, it's an approach which mostly makes sense on last gen low res consoles).

The PS3's low latency XDR ram approach is also expensive compared to PC solutions, but more important than costs it's because the Cell really requires (mostly, it can also be used by the RSX) dedicated low latency RAM to get the most out of this architecture (like being used as stream processors).