Intel's Harpertown (Penryn) quad-core processors benchmarked!

by Tarinder Sandhu on 18 September 2007, 18:37

Tags: Penryn, Intel (NASDAQ:INTC)

Quick Link: HEXUS.net/qaju4

Add to My Vault: x

Introduction



Like Intel informed us at the beginning of IDF this story may contain forward-looking statements, which might not be fact until they are released or occur.

On a sunny afternoon in San Francisco, California, Pat Gelsinger, Senior VP and General Manager of Digital Enterprise Group, pulled the attendees out of their collective slumber by announcing details on Intel's upcoming Stoakley server platform.

Stoakley is the platform name given to a combination of soon-to-be-released server/workstation-oriented processors and accompanying chipsets. In particular, Stoakley encompasses Intel's two-socket-capable, quad-core, 45nm CPUs based on the Penryn Core architecture - you may know them better as Harpertown (5400-series Xeons). Harpertown is designed to ultimately replace the current 65nm, quad-core Clovertown (5300-series) line. Harpertown Xeons, though, will carry 12MiB of shared L2 cache, compared to the Clovertown's 8MiB. Further, Intel will also release Wolfdale DP - a dual-core, 45nm part that's, effectively, half a Harpertown.



All but the range-topping X-series Harpertown quad-core processors are to ship with TDPs of 80W or less. Select models will run off an effective 1600MHz FSB, up from the maximum 1333MHz in present Clovertowns, but all will be endowed with the new SSE4.1 instruction set.

Supporting these new 45nm processors, then, is Intel's 2P Seaburg chipset, designed to appeal to the high-performance computing (HPC) sector. In most cases, Seaburg will support 8 FB-DIMM slots with 800MHz memory support, and at least two x16 PCI-Express 2.0 slots, crucially increasing graphics bandwidth.

That's all interesting stuff, we're sure you'll agree, but how does it perform? We were lucky enough to have a little hands-on testing with a Stoakley-based system, so read on to find out just how fast it is...