The Inquirer-Home

Intel 45nm Xeon 3360 on MSI X48 Platinum tested

First INQpressions FSB pumped up to the max
Fri Jan 25 2008, 16:27

BESIDE THE uber-geek QX9650 and QX9770 parts, Intel has also gradually released the first XeonUP processors in the 45nm process. Similar to their desktop counterparts, but with supposedly tougher binning and selection criteria, these server-grade 775-pin CPUs still fit into any desktop mobo supporting the Penryn generation.

Here we havethe very first in the series - Xeon 3360, a CPU with an odd 8.5x fractional clock multiplier, resulting in a 2.83GHz CPU at FSB1333 default settings. Probably the first such test on the Web at this moment.

Now, as we all know, these Penryns can go far higher, and, if our experience with the 65 nm generation was anything to go by, the FSB benefits should show up, too. So, the multiplier might not be that much of a hassle anyway...

New things only get along with other new stuff - so we got an equally new MSI X48 Platinum DDR3 mobo, the first one with the yet unannounced X48 chipset from Intel.

Besides it's famed "rollercoaster" heat pipe chipset cooling, now in the second-generation "thicker" edition, this high-end unit has four GPU slots (two PCI-E x16 v2 and two PCI-E x4 v1) and the usual set of I/O interfaces - thankfully, the old PS/2 keyboard and mouse ports are still there.

MSI also has a sister "combo" version of X48 platinum with just two GPU slots and BOTH DDR2 and DDR3 on board... that one provides for more PCI slots in lieu of the 3rd and 4th GPU slot.

As the X48 chipset is supposed to be the highest-FSB bin of the X38 run, and the XeonUP usually has a bit higher FSB stability than the QX desktop parts, we were inclined to put the two together and see how far we can go FSB-wise.

So, we decided to forgo any massive CPU overclocking, and lowered the multipler to just 8x - easy to remember CPU - FSB - RAM proportions. To match the expected high FSB, I used SuperTalent ProjectX DDR3-1800 set - a proven performer since our first X38 tests.

As for the fan, we dug out the old big Intel reference fan once used with those hot Presler Pentium Extreme 965 chips - they are far hotter than this new XeonUP, and yes their fin did a good job on the newbie, keeping it cool by running at up to 4,000 rpm. Noisy? Yeah, a lot!

How far did we go? Well, first of all, the MSI BIOS is not nearly as exhaustive in overclocking features as, say, Asus ROG series one. Still, the timing and voltage features should be reasonably sufficient to get a decent spin.

The first try: go as far as you can with default voltage on all the stuff: went fine til 3.2GHz / FSB1600. For 3.33GHz / FSB1667, I had to up the NB and FSB voltages to 1.35 volts, while for 3.5GHz / FSB 1750, that had to go to 1.4 volts. In both cases, the CPU voltage was left on "default". In fact, the CPU ran on "default" just fine up to 3.6GHz / FSB 1800, except that the FSB had to go to 1.45 volts then, while the NB could stay at 1.4 volts.

Up to this level, the Windoze XP SP2 booted fine, and the initial Sandra and Everest bench runs completed fine - the memory benchmarks were pretty much the same as on our 4.5GHz FSB1800 QX9770 run, while the CPU runs were in scale with the clock. Keep in mind, though, that this was a simple air cooled setup with the oldie fan for the CPU and NO fan for the heatpiped chipset and VRMs, or the memory for that matter - all that in a aircon-off room at a balmy 30 C evening temperature.

Anything above 3.6GHz, we had to up the CPU voltage - at 3.8GHz FSB1900, 1.34 volts were required, and, for some reason, nothing above that booted into the BIOS. 3.8 GHz settings, even with adjusted-down memory settings, weren't stable in Windows. The 3.7 GHz FSB1850 setting was fine, though. We noticed the North Bridge "rollercoaster" was getting hot - we could probably get FSB1900 working if better NB cooling is used.

In summary, 3.7GHz FSB1850 run, booting Windows with usual benchmark suspects, is pretty impressive for an air-only cooled mobo in a hot. humid tropical setting - as far aswe know, one of better results around. Our next part will focus on the same CPU in other X38/X48 boards, and more detailed benchmarks as we try to go for even higher FSB there - say 2000. ?


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?