The new chips are part of the Opteron 4100 family, codenamed Lisbon, and they’re pretty much chip-for-chip replacements for their predecessors, achieving the same power performance with a smaller hunger for power. It’s a shift from what we’re used to in a CPU race where everything gets denser, better, faster with every generation.
But the energy-sipping CPUs powered by the new Lisbon chips bring huge advantages.
For one, they’re mainly destined for use in servers. The two six-core CPUs in the range, the 4162 EE and 4164 EE suck in just 32W of power, which is the lowest achieved by any generation of AMD server CPUS, according to senior product manager Brent Kerby speaking to the New York Times. The other benefit the chips bring is price–the quad core 4122 chip is just $99, about $67 cheaper than its nearest Intel rival. When you’re talking about huge numbers of CPUs racked up in server farms, ready to bring the world the cloud computing revolution, the price differences like that (combined with lower power consumption, which translates into lower running and cooling costs) really are significant.
It’s great news for AMD, but it’s also got us pondering. Remember that innovative small-footprint server from last week, the SeaMicro? That was a system that took a lateral-thinking approach to producing powerful servers by combining lots of low-power, low power-eating cheap Atom CPUs in one box with some clever support circuitry. It results in what you could consider to be a (large) desktop-sized supercomputer. Imagine what SeaMicro could do with these new, low power-eating, cheap and much more powerful AMD chips? I once had to use, for professional purposes, a huge (multiple server cabinets-sized) old supercomputer that rocked 64 cores–it’s crazy to think that just 11 of AMD’s new chips could outperform it, for probably 1,000 times less cost.