Bit-Tech, HardwareZone, Legit Reviews, Neoseeker, Overclockers Club, PC Perspective, TechReport take a further look at the new AMD Phenom X3 triple-core processors. Aside from the fact that all Phenom X4 and X3 processors are free of the TLB erratum present in the previous revision of silicon, the new Phenom X3s will feature 64K of L1 instruction and 64K of L1 cache per core (512KB total L1 per processor). These processors will have 512KB of L2 cache per core (1.5MB total L2 per processor) and 2MB of L3 Cache (shared). Pricing will be $195 for the AMD Phenom X3 8750 processor / 2.4GHz, $165 for the AMD Phenom X3 8650 processor / 2.3GHz and $145 AMD Phenom X3 8450 processor / 2.1GHz.

A three-core processor is a little bit oddball, but I've made peace with the concept. For multitasking or more parallelizable workloads like image and video processing, choosing three lower speed execution cores over two higher performance ones might make some sense. This is a debatable proposition because desktop PC software has been annoyingly slow to migrate to multiple threads overall, but I see the potential merits. I can even see past the momentary road bumps we hit in programs like Windows Media Encoder, where only two cores were put to use. Those problems will be ironed out in due time.

The Phenom X3 processors' problems aren't in the concept, but the execution. The three cores simply aren't quick enough, individually, to make this triple-core product look appealing. They're a liability in single- and dual-threaded tasks, where the X3 8750 sometimes falls behind the much older Athlon 64 X2. The X3 8450 almost always does so. And the cores aren't quick enough to really sell the three-way concept when they are all working together. More than once, we saw the Core 2 Duo E8400 outperform the Phenom X3 8750 in intelligently multithreaded applications. Not only that, but the savings in peak power draw from deactivating a core weren't enough to put the Phenom into the same league as Intel's 45nm chips, which are astoundingly power efficient.