By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Sales Discussion - The PS3 not performing as bad as some claim?

@ stewacide

E.g. a PowerPC G5, which these chips are often erroneously compared to, absolutely *DWARFS* these chips. They're not in the same order of magnitude. And that goes for performance as well.
To quote Mintaor (Nicholas Blanchford) from AmigaWorld.net: "There are some early PS3 benchmarks here. The PPE comes out with a score almost identical to a 1.6GHz G5. That's nothing spectacular but do bear in mind that's that's using only 1 core (which happens to be the weakest), it appears to be completely unoptimised and doesn't use AltiVec. It's also Geekbench which is a bit dodgy anyway... A quick port of OS4 to the PS3 should run faster than any A1 or Peg II." "A port to the PPE should be trivial, certainly no more difficult than a port to the G5 (they are binary compatible). Adding support to the SPEs is more difficult but there is a reference implementation and the docs are all available. It's not a traditional SMP system so adding support may not be that difficult." "The real point of Cell though is the SPEs, any real work should be done on them. Getting stuff to work on one of them should be pretty easy as the instruction set is very similar to AltiVec. Altivec is dead easy, I got an app to speed up 20X in less than 3 days last week - with no prior experience of AltiVec. Parallel programming is notoriously complex though so running on multiple SPEs will be rather more difficult." So if we only take into account the PPE, an equivalent to a 1.6 Ghz G5 would be a lot faster than the platforms (called A1s or AmigaOnes, which were only ever sold together with a developer pre-release) AmigaOS4 currently runs on, and it's already very fast! A video showing off AmigaOS4 on a 800 Mhz G3 with 256MB and an old graphics card: http://youtube.com/watch?v=qSA-q1qniMY Nicholas (Minator) wrote quite a few in depth Cell articles, you can find a link on the frontpage here: http://www.cellsupercomputer.com/ Regarding the SPEs here some info from Mikael Haglunds (technical specialist of IBM Sweden) at an Amiga event in 2005: "IBM is still leading in terms of the Watts/Performance ratio with regard to their processors compared to available x86 technology and he believes this will likely continue to be so for the foreseeable future. He also explained that the new Cell processor technology includes a simple yet highly clocked PPC compatible processing core, but for software to really take full advantage of the new Cell technology, software needs to be optimized for using the specialized SPEs. He compared the Cell to how the classic Amigas operated, as the more impressive Amiga software took full advantage of its specialized custom chips. He wasn’t only thinking about complex Playstation3 games taking advantage of such technology (for instance dealing with lots more complex and more quantitive collisions), but also with regard to possible desktop orientated usages, like for instance for enhancing the performance of multimedia software or for advanced spreadsheet calculations." Article here



Naughty Dog: "At Naughty Dog, we're pretty sure we should be able to see leaps between games on the PS3 that are even bigger than they were on the PS2."

PS3 vs 360 sales

Around the Network

Diomedes1976 said: Well ,third party games tend to be multiplatform thats true .But this doesnt hurt the PS3 more than it hurts the X360 .Plus ,it is easier that the oriental games dont appear on the X360 that the occidental ones cease to appear in the PS3 .And Sony first party software has advantage in the numbers over Microsoft one (huge list of developers acquired or under contract by Sony in the recent years ) so in the software stakes they will be quite close but if there is a small advantage in one plattform I would give it to Sony not to MS .
Actuly it does hur Sony more than Microsoft because Microsoft has a lead now, and they are showing better software sales. If Xbox360 continue to have better sales than PS3, we might get a time were most software is both on PS3 and xbox360 but xbox360 has some more unique game. Of course so far it is far from true, but everything can happen.



 

 

Buy it and pray to the gods of Sigs: Naznatips!

The main PowerPC core is *by far* the most powerful on the Cell: in fact it's the only one you can characterize as a real CPU core. The other elements are essentially vector units. Linux benchmarks on the PS3 are pathetic: equivalent to a well-sub-1ghz desktop processor of any sort. Of course that isn't leveraging the vector units, but it gives you an idea what kind of power the Cell has in processing highly-branched general purpose code. Hell, the PS3 has essentially the same flexible general-purpose power than the Wii CPU! (which is a 700-800ghz full-fledged G4): the PS3s edge comes as developers figure out how to off-load work to the vector units, and also because the GPU can do things the Wii CPU has to handle. The PS3 and 360 are *VERY* skewed towards flashy graphics, because this is where the best bang-for-buck is. The trade-off (i.e. why these systems cost so little vs. a PC) is in very weak gamecode processing. So while the 360/PS3 can run flashy but ultimately shallow games like Gears of War and Motorstorm, they wouldn't have a chance in hell of running CPU-heavy PC stuff like Cell Factor or Crysis.



@ robjoh

Actuly it does hur Sony more than Microsoft because Microsoft has a lead now, and they are showing better software sales. If Xbox360 continue to have better sales than PS3, we might get a time were most software is both on PS3 and xbox360 but xbox360 has some more unique game. Of course so far it is far from true, but everything can happen.
The development costs of top projects from multi-platform game designers have increased dramatically, so without special exclusivity deals with either Microsoft or Sony they will port games to as many platforms as possible. Such companies usually use cross-platform middleware, so if the middleware's platform are optimized for platform specific advantages this may translate into platform specific advantages, but middleware developers and thus users probably won't be able to get the most out of the system. So I think *if* XBox 360 games sales for the long run will significantly outperform PS3 game sales, this will not result in fewer game ports, but this would likely result into developers not spending much time on PS3 specific optimizations.



Naughty Dog: "At Naughty Dog, we're pretty sure we should be able to see leaps between games on the PS3 that are even bigger than they were on the PS2."

PS3 vs 360 sales

@ stewacide Can you provide some benchmark references (at some functions the PPE and/or SPEs are weaker or stronger than at other functions) ? And what Linux distro did you use on the PS3? IMO Yellow Dog Linux from Terra Soft currently performs best.



Naughty Dog: "At Naughty Dog, we're pretty sure we should be able to see leaps between games on the PS3 that are even bigger than they were on the PS2."

PS3 vs 360 sales

Around the Network

MikeB said: @ stewacide Can you provide some benchmark references (at some functions the PPE and/or SPEs are weaker or stronger, than others) ? And what Linux distro did you use on the PS3? IMO Yellow Dog Linux from Terra Soft currently performs best.
Benchmarks were widely reported shortly after the PS3 launch. I'm a bit lazy to look them up unless you really care, but what the benchmarking tools showed - as you'd expect if you understand the PS3 architecture - is that it really excells in a few (highly-parallel) tasks, but really lags in all the rest. So what developers have to do is exploit those few instructions that run well, and try to minimize the rest that don't: theoretically it can be done, but practically developers are lazy and if the PS3 doesn't sell gangbusters they won't bother to re-write everything for the Cell. Just in general, the Xbox and PS3 CPUs have all the logic stripped out: i.e. they're in-order processors. And also have next to no on-chip cache. Which means they stall out *constantly* without *extremely* clever programming and performance tuning. The Wii, in comparison, is a G4 - i.e. real desktop out-of-order-processor - with added SDMI instructions (i.e. over and above regular Altivec). The 360 CPU also has very beefed-up SDMI capabilities (probably lifted by IBM from the GC/Wii) which get overlooked, but which which largely balance-out the Cell's vector units (that and the fact is has 3 vs. 1 normal in-order core). Memory speed (and disc speed as mentioned) is also a big problem for the 360 and PS3: speeds haven't scaled with what they're being asked to move around, which is why loading times are so pronounced. The Wii has *very* fast memory (and also very clever cache-locking techniques) which, combined with the scaled-down memory requirements because of SD graphics, explain why there are almost no loading times, and why it can get along well with a fraction the total memory (the memory can be more dynamic) p.s. Load-times are something that isn't often addressed but I think are very important to the casual gamer: a lot of 360 and PS3 games have really unbearable load times, and things will only get much worse. Most Wii games have effectively no load times, which is something that's much appreciated for pick-up-and-play gameplay.



stewacide said: Memory speed (and disc speed as mentioned) is also a big problem for the 360 and PS3: speeds haven't scaled with what they're being asked to move around, which is why loading times are so pronounced. The Wii has *very* fast memory (and also very clever cache-locking techniques) which, combined with the scaled-down memory requirements because of SD graphics, explain why there are almost no loading times, and why it can get along well with a fraction the total memory (the memory can be my dynamic)
I just wanna say this really fast. *nothing* comes near the PS3 in memory speed. That's all I'm gonna add to this thread.



PSN ID: Kwaad


I fly this flag in victory!

Kwaad said: stewacide said: Memory speed (and disc speed as mentioned) is also a big problem for the 360 and PS3: speeds haven't scaled with what they're being asked to move around, which is why loading times are so pronounced. The Wii has *very* fast memory (and also very clever cache-locking techniques) which, combined with the scaled-down memory requirements because of SD graphics, explain why there are almost no loading times, and why it can get along well with a fraction the total memory (the memory can be my dynamic) I just wanna say this really fast. *nothing* comes near the PS3 in memory speed. That's all I'm gonna add to this thread.
The PS3 memory architecture is a diaster/nightmare: that's all I have to say ;) It's segmented about 40 different ways and flying in all directions with about 328 potential bottlenecks to monitor. Again, a lot of *theoretical* potential, but in a day and age where computer code is still written by human beings I strongly doubt it will be managed effectively. The 360 and Wii memory architectures in comparison are about as simple as they come (the Wii has actually become a bit more convoluted than the 360 in how they scaled it up from the GC). Basically a big chunk of main unified memory, with a nice-sized chunk of super-fast memory on the GPU (or beside it on the 360): as I recall 10mb for the 360 and 3mb for the Wii assuming it stayed the same as the GC.



stewacide said: The PS3 memory architecture is a diaster/nightmare: that's all I have to say ;) It's segmented about 40 different ways and flying in all directions with about 328 potential bottlenecks to monitor. Again, a lot of *theoretical* potential, but in a day and age where computer code is still written by human beings I strongly doubt it will be managed effectively. The 360 and Wii memory architectures in comparison are about as simple as they come (the Wii has actually become a bit more convoluted than the 360 in how they scaled it up from the GC). Basically a big chunk of main unified memory, with a nice-sized chunk of super-fast memory on the GPU (or beside it on the 360): as I recall 10mb for the 360 and 3mb for the Wii assuming it stayed the same as the GC.
You my friend need to explain those bottlenecks to me, and why it sucks. Because I dont think you know what your saying.



PSN ID: Kwaad


I fly this flag in victory!

@ stewacide

So what developers have to do is exploit those few instructions that run well, and try to minimize the rest that don't: theoretically it can be done, but practically developers are lazy and if the PS3 doesn't sell gangbusters they won't bother to re-write everything for the Cell.
I don't think all game developers are lazy, especially not many coming from an Amiga background like the guys at Factor 5 (of Turrican fame, developing Lair for the PS3 right now). The classic Amiga is still popular within the demoscene, such developers tend to push every little ounce of performance out of ancient hardware configurations (within given memory limits) and actually have fun doing so. Look at this as a rewarding challenge. For instance the people behind Max Payne/FutureMark (demoscene group MatureFurk) made an excellent demo for the Amiga in 2001 winning the Assembly competition, the biggest demoscene event around the world. Lapsuus demo 50 mhz 68k Amiga with an AGA chipset (released in 1992): http://youtube.com/watch?v=2aoIsWxGWHQ For instance regarding Jorge (Jürgen Schober) a SW engineer posting at AmigaWorld, who also toured an early version of AmigaOS4 around Europe in 2003: Article here He is very excited about the Cell since he got his PS3 (he moved to the US), he thinks the Cell is a lot of fun to experiment with and offers lots of potential. He also thinks Linux already runs well on the PS3, OS4 should run even better since it better optimized to run efficiently (already runs on an old Amiga 1200 released in 1992, expanded with 166Mhz PPC and 128 MB, ancient video).



Naughty Dog: "At Naughty Dog, we're pretty sure we should be able to see leaps between games on the PS3 that are even bigger than they were on the PS2."

PS3 vs 360 sales