Quest for 1400 FSB on P5WD2-e

I'll be spending the rest of the weekend attempting to hit a "stable" 1400 fsb on the P5WD2-e using a 570 lock free cpu x14 multiplier to see how much performance this mobo has. Already @ semi stable 1300 with slight bumps in FSB termination, MCH and ICH voltages. Any tips from members that have been in this area would be appreciated. The target is with AIR, and I'm trying now improve mb underside cooling to help StackCool get rid of a little more heat.

If I get up the nerve, I'll connect the scope to Vcore and truly see what is going on with V stability so we can figure out what's needed for a droop mod. If someone can point me to some schematics or tech info on ASUS's switcher designs, it would be appreciated. I've seen all the formus pics on the boards and these really don't help. Someone musta has some tech info to originate the fixes out there.

OK, Here's the box that the mb sits on. (and my cat Rosie!), just a little extra to maybe get a few more Hz. My buddy would laugh at this because he has high end air flow modeling software used in for design of semiconductor equipment. I'll be adding pics to this post as more progress is made, trying to keep them small and reduce server space and avoid ****ing off the mods. Let me know if there unreadable and I'll reformat them for ya.

EDIT: I'll be adding 1/2 spacers between the board and the "fixture", Hope they ain't very conductive


Here's the rig on the box. Will be spending the next hour or so, tuning airflow for the CPU and finding out what runs the coolest. Theres a small 40mm fan on the NB. Quiet !!!.. Who cares. My ace in the hole is tonights cold Jersey air and the rig is right next to the window.


EDIT: Best temps with defaults are 39-40c. Fan on "base" is blowing in, Fan on Scythe is blowing in. NB fan is blowing down, added 2 80mm fans to suck air from the base of the Scythe which got reduced it by 2c. Once again, totally impractical for everyday use. Let's get some runs in and check thermals on stock settings, performance can wait till later.

First run @ shows 40 millivolt droop and would really like to see the overshoot coming off load with a scope which Speedfan's sample rate can't catch this. Stock run started @40c and peaked @ 52c, ambient=70f. Vcore Idled @ 1.44 and loaded down to 1.38. All tests will be run a minimum Vcore to keep temps manageable. I have no experience with the 570 and not sure if these temps are below or above average. Also don't know how high a "safe" Vcore is for the 570 so I'll stay under 1.6 and 65c unless there's some data to back up going higher or some reputable members taking it higher.

EDIT: Next run @ 1200 with 14x multiplier




Here's 1200. Still running stock except for Vcore increase.
Load temps up to 58c. Vdroop is up to 70 millivolts


Here's 1250


Tried 1300 but it's going to be a tough SP8m
1st run gave rounding errors, slowed ram to min, DDR480 just in case.
After reboot, once again Ai not starting up the same so entered BIOS and boot @ 300. Same thing rounding errors

Bumped FSB term to 1.40, MCH to 1.55 left ICH auto because I/O seems fine. Same thing. rounding errors. Temps about 60c

Forgot 800ul was running @ 1.8 so bumped that up to 2.2. System errors in Windows startup and even writing Dr Watson log. Event viewer is useless but suspect CDM services failure. Will disable all unnecessary hardware in BION.

Now SATA failure finding ntoskernal.exe. Had this problem with the raptor earlier and swapped it out, been good till now. Increased ICH to 1.2v, reduced processes to 22. Same freakin thing rounding errors.

Disabled Dram Throttling (that can't be helping performance...LOL), Disabled HT suspecting internal context switch problem. Same thing either Windows services throws GP faults or rounding errors.

I'll try upgrading the BIOS to 0302, really don't like the intermittent delay in the BIOS after the ram gets init and it goes to init I/O. That plus the SATA access failure at high FSB smells like a BIOS problem. The guys @ Anandtech 1300 np and will go for 1400 in a later 875 mb shootout, maybe need to re-read that review.

placeholder 3...final results and summary

that should be no problem for that board - people have hit 380+ fsb on the p5wd2-p's

Sneaky said:

that should be no problem for that board - people have hit 380+ fsb on the p5wd2-p's

Click to expand...

Ok then we'll go for 1600

EDIT: I'm just doing this as a learning experience for myself to see what needs to be changed as performance is increased. Maybe this will also help some others. Don't get me wrong, not trying to impress anyone.

care to explain how that box will help? does cooling the back of the MB really help with it's performance?

DvBoard said:

care to explain how that box will help? does cooling the back of the MB really help with it's performance?

Click to expand...

The back of the mobo gets hot, especially sitting on the table. The StackCool groundplane on the back pulls heat outta components and just want to avoid heat buildup. Besides that it only took 10 minutes to do and it's Saturday so it's either play with the PC or do laundry

cool man, good luck with that. already looking alot more promising then my old prescott. that thing was a dud and im glad its gone!

Looks like my machine was attacked by quite a few viruses surfing around for Super PI 1.4. It may take a few hours to straighten this out. Over 50 processes running.

In any case it will not be possible to get 1400 fsb on AIR, at least not on this rig. Although 1300 is SP1M stable it fails @ 8M. Temps push 60 and my feeling is the CPU is the limiting factor, not the mb. Things get real weird around here @ 4.5GHZ regardless of the FSB speed.

Will rekindle this thread when water goes on.