The article has little additional data in it and some specific points he made are questionable at best. So I really don't see the point in it other than "me too" journalism. That is, Anand had to have an article about it because the lack of such an article would be a black eye.
Points I dispute:
1) "With each core being able to execute two threads simultaneously, you effectively have a worst case scenario of 6 threads splitting a 1MB L2 cache. As a comparison, the current dual core Pentium 4s have a 1MB L2 cache per core and that number is only expected to rise in the future."
The comparison is flawed. P4s have how many competing threads to run? And the OS interrupting them all the time? Worst case, each thread gets 170 KB of L2 cache which is more than the X1 and it seems to do okay (the L1 caches are much bigger too).
2) "Regardless of how it is done, obviously the Epic team found the SPEs to be a good match for their physics code, if structured properly, meaning that the Cell processor isn’t just one general purpose core with 7 others that go unused."
Epic is using a physics middleware layer, so the point is kind of dulled. If Sweeney expects SPEs to perform physics work (which I think is reasonable) it's more because Ageia (sp) can make it work, not Epic.
3) "So with the Xbox 360 Microsoft used three fairly simple IBM PowerPC cores, while Sony has the much publicized Cell processor in their PlayStation 3. Both will perform absolutely much slower than even mainstream desktop processors in single threaded game code..."
So... categorizing the XeCPU as he does is a little flippant, but I digress. The main point here is that his claim is very objectionable. Given a single threaded game, I don't see my P4 2.8 GHz creaming the XeCPU or Cell. He needs way more support for this claim to ring true.
4) "With in-order execution as well as a small amount of high speed local memory, memory access becomes quite predictable and code is very easily scheduled by the compiler for the SPEs."
Memory access becomes predictable when the algorithm implemented makes predictable memory accesses. In-order + LS does not do that.
5) "Compilers are horrendously difficult to write; getting a compiler to work is a pretty difficult job in itself, but getting one to work well, regardless of what the input code is, is nearly impossible."
Compilers are definitely complex beasts anymore, but I think he overstates the difficulty a bit. Compiling for a single CPU has got to be simpler than compiling for x86. Between IBM and MS, I would imagine a decent compiler to be available for launch and it get better as time goes on.
6) "On the other hand, looking at all of the early demos we’ve seen of Xbox 360 and PS3 games, not a single one appears to offer better physics or AI than the best single threaded games on the PC today."
Launch games are always fractional representations of what the hardware can do. And anyway, I was pretty amazed at what Heavenly Sword's physics can do. As an aside, is HL2's AI a huge step up from HL1? This parallel's ERP's point.
7) "A single thread is used for all game code, physics and AI and in some cases, developers have split out physics into a separate thread, but for the most part you can expect all first generation and even some second generation titles to debut as basically single threaded games."
Animation is a big deal in terms of CPU time and it's not specifically referenced making me wonder if MS expects it to be spun off into another thread. I was under the impression that at least some PC games do this already, and I would expect it to be a huge win on the 360 where it would have its own core.
8 ) "This means that all the ops per clock could either be dedicated to geometry processing in truly polygon intense scenes. On the flip side (and more likely), any given clock cycle could see all 240 ops being used for pixel processing."
I'm not sure he realizes that the bottleneck on the GPU shifts on an intraframe basis. Games aren't necessarily fillrate bound the entire frame. And even if that's the case, the 360 has a lot of render target bandwidth.
9) "At 720p, the G70 is entirely CPU bound in just about every game we’ve tested, so the RSX should have no problems running at 720p with 4X AA enabled, just like the 360’s Xenos GPU. At 1080p, the G70 is still CPU bound in a number of situations, so it is quite possible for RSX to actually run just fine at 1080p which should provide for some excellent image quality."
This is almost a worthless point. Today's games? Today's PC games are likely to have a fraction of the demand PS3 games will have, especially after launch. I mean, did we judge the Xbox according to Quake 3 and that was it?
