Printable View
I'm trying to run 7x400 fsb on my E6600 and I was hoping it would work at that lower multi so I could reduce vcore and therefore bring down temps at load on the cpu.
Dosent even get to the post screen and I have to power off to get back in the bios.
I'm running at the mo - 9x333fsb mem @ 833mhz vcore and mem volt auto, and nb/sb 1.50v/1.05v
Any ideas?
Try 401MHz ;)
I can boot at 8x400 fine here ... no raised vMCH or anything ...
t
Well I treid 400mhz and ram at 800mhz but it wont work, cant see 401mhz working mate.
Ram is Gskill 6400 NR. @ 5/5/5/15.
Seems like your mobo not from a good batch as mineQuote:
Originally Posted by Jimbo
can boot@3.954 ghz but not at 400 fsb :p: :p: :p:
vnb 1.65v and mem volt to.. 2.1-2.2v maybe. dunno what chips NR kits are using..
Sooo, your currently @ 9x333 = 3GHz but you want to run @ 7x400 = 2.8GHz? If you want to run at a lower GHz, lower Vcore, and lower volts, why not just try 9x300 = 2.7GHz.
If that would let me lower vcore I suppose I could try it, thanks for that. Just until I get a Tunic Tower 120 lol.
See this post:Quote:
Originally Posted by Jimbo
http://www.xtremesystems.org/forums/...5&postcount=23
Long and short of it is - I need 1.3250v vCore to dual Prime stable at 10x340 (3400Mhz), but I need 1.4250v vCore just to POST at 6x410 (2460Mhz).
Quote:
Originally Posted by Jimbo
You say you already have Vcore set to auto? How much lower are you trying to get?
Does this mean that the P5WDH boots up with a 9x multi and then switches back to the user-setting?Quote:
Originally Posted by Durzel
yup.Quote:
Originally Posted by CopperCAT
p5w and p5b change multiplier by modifying exactly the same cpu register as, for example, crystalcpuid does.
first, post at default values and then bios applies new multiplier and boots up.
ES and XE chips are different - they are true unlocked cpus.
i totally missed that one from first post :)
indeed, to post at 7x400 you have to give cpu enough vcore like post at 9x400.
My P5W DH couldn't boot at FSB 400 MHz, and the max value was 390 MHz. That's quite normal, that some can get past 400 MHz and some not.Quote:
Originally Posted by Jimbo
that's weird, 'cause I tried 7x450 1.3V with an E6400 and 8x400 1.3V with the same CPU. Both ran fine. Yet it couldn't boot at 450x8 with 1.3V. So what you've written doesn't add up, as far as my experience with P5B DLX and P5W DH is concerned.Quote:
Originally Posted by caater
Yup, you're right.. except ES CPUs are the same. Don't know about XEs though - Im guessing from that fact they've always been unlocked that there's something special about them internally.Quote:
Originally Posted by caater
From my experience (I've got an E6700 ES) I can't boot at 6x410 unless I give it 1.4250v vCore - which is way more than is required for the clockspeed (2460Mhz). I can dual prime stable for 18+ hours at 3.6Ghz (10x360) at that vCore.
My hunch, like you said, is that Asus multiplier unlocking works the same as CrystalCPUID - but that when you first POST it will try and boot at whatever your default multiplier is, which in my case is 10x... this would explain why I need so much voltage to boot if its trying to boot at 10x410 (4100Mhz).
What CPU have you got? E6400? If you're giving the CPU 1.3v vCore, which you say is fine for either 7x450 or 8x400, then maybe your PC is capable of booting at 8x450 (3.6Ghz), even if it wouldn't get into Windows at that speed (which wouldnt matter anyway because Im guessing the multiplier gets changed as soon as it properly POSTs).Quote:
Originally Posted by mach82
Jimbo on the other hand has an E6600, and if hes trying to boot at 7x400 then I think its initially POST'ing at 9x400 (3600Mhz), which is clearly too much for stock voltage.
7x400 = 514MHz NBCC, maybe you need to up the volts on your NorthBridge?
I have only so far been succesful botting at 1.3v anything below that and my current oc doesnt work.Quote:
Originally Posted by Icer
i'm afraid that multiplier has nothing to do with nbcc..Quote:
Originally Posted by itznfb
but that's possible that his nb voltage is too low for 400 fsb.
jimbo, have you tried to bootup at 333, mem at 1:1 or 4:3, lowering mult with crystalcpuid and using setfsb/clockgen to determine your max fsb with your current settings?
NBCC = DefaultMulti/SetMulti * FSBQuote:
Originally Posted by caater
http://www.xtremesystems.org/forums/...d.php?t=114998
you can easily disregard that nbcc information, it's just sandra bug :)Quote:
Originally Posted by itznfb
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.
yea, so you're saying all the information that Tony and FCG provided is false?
well, it was FCG's post, so unfortunately he made an error this time :)Quote:
Originally Posted by itznfb
Tony's posts about straps and so on are very true and quite helpful.
besides, if you think carefully, this statement also defies logic.. A laptop with T7600 in low-power mode (6x mult) would have 387MHz nb instead of 166MHz :)
not exactly power-saving..
well i can't find any evidence of NBCC=xDefault/xSet * FSB being incorrect. the only posts i can find prove it to be true. i don't have time however to read through all the posts right now. do you have evidence pointing to anything that disproves the NBCC or bandwidth results?
bandwidth tests (tonys findings) are correct for p5b and there's no need to disprove these. and they are totally unrelated to fcg's theory about nbcc.Quote:
Originally Posted by itznfb
and nbcc - read my earlier post about that.
but let's concentrate here on Jimbo now :)
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 :(Quote:
Originally Posted by itznfb
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.
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 caater
965 chipset is 90nm while 975 is 130nm.Quote:
Originally Posted by itznfb
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 ;)
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.Quote:
Originally Posted by Tony
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?
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.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.
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..
well i sure do hope that they make p5w bios on par with p5b..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 ;)
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?
first try with your vcore:Quote:
Originally Posted by Jimbo
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).