MorganX said:
Only and idiot relies on benchmarks as opposed to real world performance and applications. Are you and idiot. I've used a 1.6Ghz G5 (Photoshop, Keynote, Office X) and you and Steve Jobs can't convince me that it's superior to a 2.8Ghz Pentium IV in the same price range. Troll must = disagrees with me.
you insist on using the lowest end cpu in your comparisons. I understand this is what you have, but you are talking about the bottom of the barrel PPC 970. The underlying system isn't even the same as the other models.. different PCI bus, different memory bus, different FSB. You are also comparing a low end G5 to a mid range P4 and the P4 has the benefit of robust compilers. At the time of the G5 release, gcc didn't even know what a G5 was. They 'shimmed' in G5 support by duping the compiler into treating it, in some respects, like different architectures that functioned in similar manner to the PPC 970. Even now, your software is being compiled on sub-par compiler technology (GCC isn't even all that well optimized for the P4 yet since the gcc steering group puts way more value in cross-platform support than specific architecture optimizations). As I mentioned, we regularly see a 30+% increase in code performance when using IBMs XL compilers (which are optimized for the PPC 970, but which were just released a couple months back).
Regarding the alleged Xbox 2 CPU, I doubt that all 3 cores will be able to fully utilize each clock cycle, or bus bandwidth to support the notion that it will have the performance of 3 G5 cores running at 3x the current single processors. Just as dual CPU systems rarely if ever produce 2x the performance of their single CPU counterpart.
dual cpu systems don't generally double performance, and one issue is cache coherence. They have to keep tabs on the status of data in the caches of other cpus in the same system. This processor would solve this issue (for the most part) by sharing a L2 cache. Also, the interprocessor latency is almost nil with a multi core cpu [compared to more common smp systems]. If this processor comes into being, it will be much more efficient than a system with 3 seperate CPUs in it.
I think it comes down to this.. (if this is real) you have a system with three cores on the same die and a shared 1MB L2 cache. It's hooked to a HT like bus with 10.8 GB/sec up and 10.8 GB/sec down unidirectional bandwidth. It will perform MUCH better than a typical (think Xeon) style SMP box. Not only that, but since we assume M$ has input into the design.. do you really think they would build in so much silicon that it would be rendered usless by an insufficient bus?
Furthermore, I doubt they will be full blown processors. With MS' customization, they won't need to be. They're not going to be running Office. They're going to be running games, media player, and XP very lite.
what exactly is a non-full blown processor that runs xp, games, media player? do you think it won't support the full PPC (or Power) ISA? I find that fairly unlikely. I'm guessing that you mean that it will just be less 'wide' than something like a PPC 970? That a given core won't be able to support as many execution units? Maybe 1 FP, 1 Int, a Vector unit or two? this could be.. but then if you are lowering the potential IPC, don't you think that would decrease FSB bandwidth requirements? Also, even CPUs with very high max theoretical IPCs like a PPC 970 don't get close to that number in real world use. One more thing to consider.. .if IBM is providing CPUs that are already wide (lots of int, lots of fp units, multiple vmx paths), why would M$ choose to totally reinvent the wheel by first stripping the parallelizm out of an existing design to build it back in with a transisitor expensive multi-core design? That seems odd, unless the real goal is a system with a lot of virtual paralellism.. a somewhat simple processor that appeared as 6 virtual CPUs to the system. I don't know why this would be so much more favorable to what they are planning to use this box for.. but I don't work for M$.
Based on that, the performance of the current 1.6GHz G5, and I think the only spectacular thing about this rumor is that IBM could be ready to produce multiple cores by the end of this year or early '05.
you mean, based on my impression of the performance of the low end PPC 970 processor running fairly unoptimized code....
As for cost, MS has entered into a partnership with Taiwan Semiconductor who has a patent for PC on a chip, so who knows how cheaply MS may be able to manufacture this thing, especially since they are assisting with the chip design/optimization.
It is my impression that IBM is manufacturing the cpu for the xbox2, as well as providing the technology. In fact, there was early confusion from the AMD fans because it initially sounded like IBM was MANUFACTURING the chip.. but they were initially unclear about whether or not they were desiging it too [as in was it a Power or not]
I still don't think you've made your case as to why this isn't an impressive cpu.. if it is real. Even considering your experience with the 1.6 GHz, you'd have to realize that this is a 3 core cpu based off a new architecture that may be clocked roughly 3x as fast as your machine. Even with simplified cores, I'd be shocked if a box with this design wasn't 6-10x as powerful as your machine.
edit: BTW, where's the audio processor?
yet another good question and another reason to think this is a fake.
