Dual core Pentium Presler XE runs fine at 4.26GHz

I FINALLY I finally had a look at a long-awaited Presler XE (Pentium 955) sample, the dual-core 3.46 GHz MCM beastie with dual 2 MB caches and 1066 MHz FSB to boot - all that put on a brand new Intel 975XBX mainboard (oh yes, don't confuse the 955X chipset with the 955 CPU, although both should be able to work together anyway).

As it is a dual-core hot rod able to heat even cold Beijing winters, and probably melt some icicles off the Great Wall along the way, I expected it to overclock only slightly, or not at all, when using the supplied Intel stock fan. So I minimised on all the rest, using just two sticks of Crucial Ballistix DDR2-1000 RAM, and Sapphire X1600Pro graphics.

Boy, was I wrong... by mistake, I changed the default x13 multiplier not to x14, but to x15 and, when I rebooted the system, the magical 4.0 GHz clock frequency status appeared, yet I only upped the default CPU voltage by 0.025 volts. The CPU temperature as per hardware monitor jumped from 80 C to 90 C, and then pretty much stayed there over those 15 mins as I checked all the new BIOS options, some of which I saw for the first time on an Intel board.

Anyway, let's go all the way to a nice, more harmonious x16 multiplier - yes it worked too, and I got an even nicer 4.26 GHz clock with an extra 0.025 volts on top. Hmm, counting dual cores, this is now a 17 GFLOPs PC system - once this was considered a supercomputing power. The CPU temperature went up to 91 C and, again, stayed firm there over another 20 mins or so - the system also rebooted to BIOS fine several times at this setting.

Watch the mainboard overclocking (or as they call it at Intel, "Configuration override") settings - you got everything from (fully working) CPU multiplier changes up and down, all possible voltage changes, FSB settings up to 1333 MHz, even FSB bus voltage settings! And remember, all that from the company that, well, designed those CPUs and chipsets in the first place - they should know their limits best...

See the screenshots for more... not to forget those large blue heatsinks on the VRM section.

On top of it, the board is nicely done, with Crossfire support (one PCI-E x16 or two PCI-E x8 config, plus one extra PCI-E x4 slot in PCI-E x16 physical connector, supporting triple-card config) and all the bells & whistles (including 8 SATA ports) that you'd see on a high-end Taiwan board. Hmm, Taiwan high-end mobo vendors better watch out... looks like Intel is after them.

Now trying to run Windows on this... let's see how that goes. µ

Presler speed margins seem to be high... is Intel keeping a secret emergency "performance headroom"?

Should AMD be on alert - what if Conroe also has better than expected frequency headroom?