fpnc said:
In a sense, Barefeats is correct.
No they are not. They are basically looking at the specs of the single processor and comparing them, when they should be comparing the specs of the core (which matter in this particular benchmark). the processor in Quad has two cores, with each core having 1MB of cache. If they want to do a real apples to apple comparison, they should compare the core on the Quad, to the core on the dual.
In the 970MP each core has 1MB of cache and the quad-core Power Mac thus has 2 x 2 x 1MB or 4MB of total L2 cache. The single core 970FX has 512MB of cache per processsor and the previous Power Macs only contained two cores (two processors). Thus, the old dual Power Macs had 2 x 512MB of L2 cache (1MB total) while the new quad Power Macs have 4MB total.
Since UT is single-threaded, it does not matter how much L2-cache the other core has. UT is faster on the Quad because the app has twice as much cache at it's disposal (1MB as opposed to 512KB), and propably because of PCI-E and faster RAM. Yes, the entire processor might have four times the cache, but since that cache is spread across the cores, and the app only uses one core, what matters is the amount of cache single core has. It has four times the cache, but that cache is spread to several cores, and it has twice as many cores as the dual hase.
If they say that "UT is propably faster because Quad has 4 times the cache", they should then also say "UT is propably faster because Quad has twice as many processor-cores as the dual has". But they can't say just one, and not the other. And in any case, since UT does not use multiple processors, what matter is the specs of the individual core.