Quote:
Originally posted by Destroyr:
Thank you all for bringing up one of my lifelong pet peeves: Mac People. Mac people always rant and rave about the superiority of RISC architecture, trying to get their chips associated with REAL computers (Alpha, Sun, SGI, HP-UX, etc.). The truth is, they have never had a leg to stand on. They always based their "superiority" over pc's on theoretical numbers and instruction latencies, neither of which have any real world meaning. We'll start with theoretical numbers. The basic thing that separates the G4 from the G3 is the same thing that separates the P3 from the P2, and the K6-2 from the K6; namely SIMD. AMD did it first, and it would be nice if someone would give them credit once in a while. They had 3dnow a good year and a half before SSE, and 2 years before "velocity" (for those of you who are unaware, that is what apple called their 3dnow wannabe instruction set). These 3 setups were all basically the same in the beginning, and both Intel and Apple decided to take all the credit for being ever so special, when it was, in fact, AMD's idea. All that gigaflop nonsense apple was preaching way back in the day was just sustained processing using code made specifically to look good on a mac, by running the SIMD unit in a way that would very seldom be possible in real life, and their results could just as easily been duplicated on a $50 k6-2 500, or a p3 500
(not to mention the fact that any athlon or duron over 700mhz really can sustain 1 gflop, in real world computing, at double precision). When the G4 came out, they claimed to have taken 6 benchmarks off of intels own website, and beaten the p3 by something like 300%. After digging and digging and digging through apple's bs propoganda, I finally managed to find these benchmarks. All 6 were slight variations of a single command in adobe photoshop, and the results were in theoretical instruction latencies (ie: 2 clock cycles to complete on a g4 vs 6 clock cycles to complete on a p3 = g4 is 3 times faster than p3). At this point, I reallized just how desperate apple was, and how depressingly far behind they were. And on top of that, they have barely managed to increase their speed since then, where AMD has doubled their clockspeed, AND increased their IPC substantially, and Intel has also mad some big performance gains. Anyway, it all came down to underhanded garbage, which apple had no qualms over shoveling at the computing community incessantly. Any way, I should avoid getting into the subject of macs sucking any further, or i'll be here all night. Have a nice day.
Destroyr, if I compared the Alpha and the G4, it was simply as a point of reference. I understand that the two are very different architectures, but I was simply making note of the manner in which both approach their problems. That is, while clock speeds are low, the best way to increase effective throughput is to throw more transistors at the problem regardless of cost. Apple has ignored the external bus (especially to RAM) for sometime now. For all the G4 is touted as, it can be that much better of a CPU with a new FSB revamp. 