Bernard et al,
With all due respect...
1) Hitting the swap is NOT a reliable indication that RAM usage is maxed -- instead it almost always means the programs you are running cannot efficiently utilize it. Most machines cannot efficiently use more than 8G or so of RAM. My machine has 16 and if I run C1 in batch (very effectively uses all 8 cores), Helicon Focus (also efficiently uses all 8 cores), CS4 (very definitely does *NOT* efficiently utilize all 8-cores) and AutoPano Pro (does not use multi-core well) ALL AT ONCE I tag maybe 12 Gig total of my ram. Yes, CS4 tags the scratch disk as does APP. (See note #4)
2) The new *BASE* machine is a 2.26 GHz processor, 8 of them yes, but they run at 2.26, or 30% SLOWER than the current 3.2... The *FASTEST* new Mac Pro at 2.93 GHz ran programs like Aperture about 20% faster than the previous 8-core 3.2 machine; however, most of that gain is likely due to the added throughput of DDR3 RAM, and not anything else. It is also a bad assumption to lump CS4 into this same class of software --- CS4 does NOT manage processor throughput or RAM nearly as well as Aperture. Frankly, I suspect that CS4 will run faster on the old machine due to the faster processors -- and probably proportional to processor speed faster -- at least until the time Adobe writes some modern code for CS4 that will utilize all the processing power and RAM available to it. To wit, a friend with a first generation Mac Pro with a single dual-core 2.66 processor and 8 G RAM can run most CS4 benchmarks about 20% slower than my 8-core 3.2 machine with 16G. (See note #4.)
3) DDR3 RAM, an interesting note... DDR3 is THREE channel RAM. The *new* 8-core machines still only have 8 RAM slots configured in a new but still 2 banks x 4 slots configuration. Can somebody explain to me how 2 sets of 4-bank memory slots efficiently use 3 channel RAM? Clearly they do, but it has a few of us surmising they are really only utilizing the full DDR3 in the first three slots of each bank, then let the last pair of slots fall to DDR2 speed or even let them act as DDR1 overflow memory. Again, most programs simply cannot utilize RAM well yet.
4) IMO disk I/O is still the significant limiting factor for most of what we as photographers do. It remains the major bottleneck in our machines. Here is where having a striped array (RAID-0) works wonders for boosting performance. I have 6 drives in my Mac Pro (see http://www.maxupgrades.com/istore/index.cf...Product_ID=158)
. I use WD 640's, but the newest high-density 1TB and 1.5TB drives are also screamers.
*** Of course RAID 0 is for speed and is *LESS* reliability than single drives, so redundant back-up is mandatory; one of the drives in either of my arrays WILL GO DOWN and when it does, I will be DOA on that array. But I can rebuild it in a matter of a few hours when that happens and the performance gained in the meantime is well worth the rebuild hassle.