Bob F wrote:
Bob F wrote:
I finally got around to trying this processor on the P5K Premium
after my problems on the Abit IP35 Pro.

I modded the ASUS AMI bios using the following:
http://www.delidded.com/how-to-updat...e-in-ami-bios/

Booted it up to bios, reset to default, rebooted to windows. Windows
runs for awhile, then crashes. That's way further than I got with
the Abit board.
Speedfan seems to display temps about 15C below the real temp
displayed by coretemp on the E5472. It was right on with my Q8200.

Temps, still using the stock Intel heatsink, seem to be higher than
they should be, so maybe that is why it crashes. The E5472 seems to
have a max temp spec of 75C. Speedfan was showing temps in the mid
60's, so with the 15C offset, maybe too high. The E5472 is supposedly
80W max compared to the Q8200 at 95, so I'm a bit surprised. They
seem to both idle around 50C, compensating for the speedfan errors on
the E5472, with speedfan controlling the CPU fan speed according to
temp.
I thought these newer processors were suppost to slow down if they
get too hot, rather than just crashing. Could it be a memory error?
The DDR2-800 memory should be able to handle this speed, right? ASUS
claims this board is good to FSB 1600, which I understand matches the
DDR2-800 memory
I've got the Q8200 back in now, as this is my HTPC machine, so I
don't want it crashing, but will do more playing with the E5472 when
I have the opportunity.
I'd appreciate opinions and suggestions.

I just re-installed the E5472, and modified Speedfan to compensate
for the 15C temp offset for reading the E5472. With the compensation,
SpeedFan reads the same as CoreTemp.

I ran MemTest86+ with no errors, with my ram at both 6-6-6-18 and
then 5-5-5-15. Funny thing, memtest86+ on the Q8200 showed the RAM
speed as 4394 MB/s at 6-6-6-18 setting. The E5472 shows 3969 MB/s at
6-6-6-18 and 4197 at 5-5-5-15. So the E5472 is displayed as slower on
Memtest86+. The time for a whole pass testing the memory was faster
with the E5472 however. It took 33:42 on the Q8200, and 30:32 with
the E5472, both at 6-6-6-18. At 5-5-5-15, that was 29:33 on the
E5472. This suggests something funny about the MemTest86+ speed
reading. Just looked, MEmTest86+ has a new version. I'm at 4.1, so I
need to get 5.1 and try that.
I ran the Prime95 torture test for 35 minutes without a problem (at
5-5-5-15). Core temps maxed at 67C. Interestingly, coretemp shows
Tjmax as 100C, although I've found numbers from 71 to 85 online.
Intel lists case temp max of 67C, which I'm about 10C under.

It looks like this E5472 on my P5K Premium is a GO! I guess it's time
to try overclocking it a bit.

I am now running at 3.3 GHz (FSB 1761 - DDR2-880)) without problem. At first
when I got to 3.3, the temp under Prime95 was pushing the limits. The default
CPU voltage bounced around 1.25V, and since the spec is .85-1.35, I tried
reducing that voltage, and finally settled at 1.18V setting, which displays on
Speedfan as 1.16V. Running Prime95 at that voltage indicated no problems, and
the core temps now stayed below 70C. CPU case temp got up to about 62C. RAM
voltage remains at the default 1.8V. I'll keep my eyes open for a better
heatsink than the standard Intel assembly, but I'm quite happy with these
results. The PC definately is faster.