Cant boot at 400fsb = P5W DH.

Quote:

---

Originally Posted by itznfb

well, this is directly related to the ontopic issue. if there is an increase in NBCC related to the multiplier defaul/set ratio, whether it resets with straps or not could be causing his problems. which would explain the cases were some can boot at 401, but not 400.

---

oh but Jimbo has a P5W, it sets 1066 nb strap starting from 200ish fsb and that's pretty much it. no 1333 strap and no difference between booting from 333, 400 or 401 fsb :(

his problem currently is vmch - or lack of it. 1.5v seems to be too low.

is there that much difference between the P5B and P5W? i thought some of my voltages were too low and then some people corrected me and had me put them back down to default to get me to boot at 400fsb. i'm able to boot at 400-420fsb on all default voltages fine except vcore.

Quote:

---

Originally Posted by caater

you can easily disregard that nbcc information, it's just sandra bug :)
also control panel->system and various older programs have trouble detecting real clockspeed.
they just don't know how to read fsb :)
just use clockgen/setfsb and see if sandra detects new fsb.. :)
but what sandra is capable of, is reading multiplier.
default multiplier is readable from cpu registers (msr 2Ah, bits 26:22), bootup frequency is also supposedly readable from msr (i think from E7h or/and E8h).
if multiplier is changed, sandra detects it, assumes that clock speed is the same (because it lacks support for reading from clockgen) and displays that weird clock speed. and that weird clock speed is indeed DefaultMulti/SetMulti * FSB but that's not nbcc :) the real nbcc still is FSB * multi_Defined_by_strap.. or smth like that. as explained by Tony.

---

Yes, i totally agree with you, moreover Sandra shows that the memory too is overclocked when you lower the multiplier because it computes it as FSB*DRAM_ratio, how we can see in this bugged information the frequency of the northbridge i don't know, moreover i don't see a single reason to increase the northbridge frequency while decreasing the cpu multiplier.

Quote:

---

Originally Posted by itznfb

is there that much difference between the P5B and P5W? i thought some of my voltages were too low and then some people corrected me and had me put them back down to default to get me to boot at 400fsb. i'm able to boot at 400-420fsb on all default voltages fine except vcore.

---

965 chipset is 90nm while 975 is 130nm.
that explains why 975 needs more juice..
also it's reported that 965 gets quite hot..
but the main difference is exacly what Tony explained, p5b uses 1333 strap from 401 fsb.. slacker timings but way higher fsb..

one thing's for sure - remounting stock nb hs and using proper TIM will gain both fsb and memory stability with both boards..

btw Jimbo, what temps you're getting? maybe you're too worried about temps..

P5W DH sets 800 strap with an 800fsb cpu, with a 1067fsb cpu it sets the 1067 strap from 267fsb but it does seem to be locked to 1067 when down clocking.

NBCC while you think bugged in sandra we actually feel is pretty close, when I have time I will experiment with the badaxe where I can force straps etc and see how performance ramps etc. I can tell you chipset core clock does rise with CPU FSB so the higher you clock the more stain the NB is under unless you force a different strap to force its base clock lower.

In all at this time the P5W DH is not as good an overclocker as the P5B, this may change soon through with a new bios or 2 ;)

Quote:

---

Originally Posted by Tony

P5W DH sets 800 strap with an 800fsb cpu, with a 1067fsb cpu it sets the 1067 strap from 267fsb but it does seem to be locked to 1067 when down clocking.

---

did a quick test, starting from 221 fsb, p5w switches over to 1066 nb strap (and 800 strap from 148), though 800 strap is always available through some dividers.
i wonder why asus didn't make 1066 strap mem dividers available for lower straps - i guess they afraid some reviewers would use them and make this board look bad in reviews?

Quote:

---

I can tell you chipset core clock does rise with CPU FSB so the higher you clock the more stain the NB is under unless you force a different strap to force its base clock lower.

---

for that part i fully agree with. only thing i disagree of is that nb core clock is in any way connected to cpu multiplier.
the question is - how low can a chipset go :)
is there any higher possible straps than 1333 (eg lower nb multipliers)?
975x seems to have only 2 bits for strap settings,
01=533, 10=800, 00=1066 and 11=1333 (i presume, i'm still waiting for someone to give me badaxe mchbar register dump or just fed14e08h value)..
though 965 might be different..

Quote:

---

In all at this time the P5W DH is not as good an overclocker as the P5B, this may change soon through with a new bios or 2 ;)

---

well i sure do hope that they make p5w bios on par with p5b..
there'll be lots of happy allendale users then..

one more question to Jimbo though:
do you have hyperpath3 turned off?

If the P5W-DH is to support the X8000 (3.33Ghz, 1333Mhz FSB) which is supposed to be coming out in Q4 2006 then wouldn't it have to support a 1333 strap sooner or later?

Or am I being stupid?

Hi Cater,

Yes I have hyper path turned off, temps are about 50c idle and 60c load with Everest/coretemp. Can anyone reccomend any settings I should try?

Quote:

---

Originally Posted by Jimbo

Hi Cater,

Yes I have hyper path turned off, temps are about 50c idle and 60c load with Everest/coretemp. Can anyone reccomend any settings I should try?

---

first try with your vcore:
fsb 333, ram 667, pcie 100, pci 33, vmch 1.65v, vfsb 1.4v, vich auto. in cpu configuration lower your multiplier to 7x (or lower it with crystalcpuid later on).
chipset settings - mem timings by spd. for testing.
in windows, use clockgen or setfsb to find your limit with those settings (multiplier 7x of course).
use first 1M, then sp2004 for stability test.
test in small steps, first like 10mhz, and later on, even for 1-5mhz.
if you'll cross 400 without much problem, you'll need more vcore to boot at 400.
write down all your stable frequencies. :)
if you don't.. then maybe write down all your bios settings and let us know.

Brilliant thanks Caater, I'll do some testing with the above.

Ok I must come clean here and say Im a bit of a noob at overclocking. Havent done so to any serious extent for a while, however I have the following settings working 100% stable:

Ai Overclocking Manual
CPU Freq 333
DRAM Freq: DDR2-833
Performance Mode: turbo
PCI Express Freq: 101
PCI Clock Synch Mode: 33.33
Memory Voltage: auto
CPU VCore Voltage: auto
FSB Term Voltage: auto
MCH Chipset Voltage: 1.5
ICH Chipset Voltage: 1.05
Configure DRAM Timing by SPD: 5-5-5-5-15
Hyper Path 3: Disabled

But since Conroe came out it like any wannabe overclocker caught my imagination. Problem I have though the likes of Crystal cpuid looks a bit daunting to use, I'll try though.

Edit:

Above settings on default multi of x9 of course on the 1403 bios.

just use these settings i suggested.
get setfsb from here: http://www13.plala.or.jp/setfsb/down...15a3_P5WD2.zip
and remember to check "ultra" before clicking "get fsb"
to change multiplier in crystalcpuid: open program, function-intel enhanced speedstep control and choose your desired multiplier (6-9x)
also it wouldn't do much harm to find your cpu limit.
put all the voltages like i wrote down, fsb 333, but ram this time to ddr2-500 speed.
leave the multiplier at default 9x.
then in windows use setfsb or clockgen to clock cpu and superpi 1M for testing.
if 1m went through, add fsb for ~3mhz (~27mhz more cpu speed) and repeat 1M. if superpi gives error, reboot computer and up vcore voltage.
suggested steps auto, 1.35v, 1.4v, 1.45v, 1.5v, 1,55v..
once back in windows, set fsb to the last failed one and continue testing.
remember to write it down :)
also after superpi crashes, remember to delete all *.dat files from superpi directory (ia.dat, iy.dat etc).