The Inquirer snagged some interesting information about AMD's upcoming K8L architecture at one of the company's recent keynotes. K8L is expected to supplant the current K8 architecture some time next year with a native quad-core design and better floating point performance. According to The Inquirer, the new architecture will double floating point and SSE resources, add new instructions, increase pre-fetch from 16 to 32 bits, and introduce a shared, expandable L3 cache. Core-dependent voltage settings are also said to be on the table, allowing different voltages for each of the K8L's four cores. The K8L architecture is reported to add 48-bit physical memory addressing-an 8-bit step up from current Opterons-that will boost physical memory support from 1 TB to 256 TB. K8L will also bring DDR3 support "when the spec 'settles down'," as well as optional FB-DIMM support. There is also official co-processor support, strongly hinted to be on a HTX card.