To: Doren who wrote (34920 ) 9/13/2002 6:53:51 PM From: Cogito Read Replies (1) | Respond to of 213182 Doren - While I don't think it's really disputable that the fastest PCs are currently faster than the fastest Macs, at least for most tasks, I'm confused about some of the early benchmarks on the new G4 line. I am not implying that BareFeats has some sort of anti-Mac bias, since it's more likely that the opposite is true. What's bugging me is that I can't resolve what I've read there with what Apple says about the new machines. For example, on the subject of memory bandwidth, here's what Apple has to say, "The resulting throughput between main memory and the system controller is up to 2.7GBps, more than double the throughput from the previous dual 1GHz Power Mac G4." I remembered BareFeats, on the other hand, talking about a 1.3 GBps path between the main memory and system controller, which they identified as a bottleneck. But when I just looked at BareFeats again, I realized that the 1.3 GBps memory path theory was simply that, and that it had been proposed by some of their readers. Everything I have seen from Apple about the new systems indicates that a lot of work has been done to eliminate internal bottlenecks, including raising the system bus speed to 167 MHz, using a new system controller providing a 2.7GBps pipe from main memory to the processors, adding 2MB of DDR L3 cache on a dedicated bus making, again according to Apple, "the full 4 GBps data throughput always available." So on paper, they look great. One would expect such systems to excel in real-world benchmark tests. Yet according to BareFeats, they don't. Perhaps it is true that the CPUs themselves are incapable of receiving data at a rate above 1.3GBps. If this is all borne out in other benchmark tests, I'll be quite disappointed. That would show that while Apple did take care of a lot of bottlenecks, they either ignored one that would render all their other work far less meaningful, or they just decided that since they couldn't fix that one they would just use the other improvements as marketing fodder. I will say that some rather silly statements commonly appear on BareFeats. They seem to believe that a system with a 25% faster CPU should perform 25% faster, all else being equal. Of course that isn't true. Overall system performance almost never scales that way. Look at any of the hundreds of reviews of new Intel systems over the years. So one does have to wonder about them. I really want to see some more benchmark tests of these new machines. - Allen