OCing Winchester 3200 / doesnt like voltage?

E

el rolio

2[H]4U
Joined
Jul 8, 2004
Messages
2,495
hey all, this weekend as i was waitin on more parts to finish my WC rig build i decided to play around with settings at my 250 x 10 setting that i have been using pretty much since 2k5 started. or before even. anyhoo, on to my question

it seems that runnin at 1.425 + 5% or whatev (since im using a msi neo plat) is a voltage setting it likes, and i like it too cuz its not that much heat coming out.
now i decided, lemme throw a smidgen more volts at this thing and i can see what happens with some long prime runs (the previous setting would go about 15- 20 hours then gimme a rounding error or expectation error im not sure).

well this more voltage thing made prime95 fail pretty wuickly, usually between 5- 6 hours. so i tried various settings:
1.425 + 8.3%
1.425 + 10%
1.45 + 5%
1.45 + 8.3%
1.45 + 10%

none of them lasted longer or shorter and none of them pushed my swiftech storm past say 51 max load as read out by mbm or everest.

so whats the deal folks? winnies like less voltage? is there some way to give the chipset more voltage cuz of the 250mhz? does it want more?

note: specs in sig / ram iss on 166 divider resulting in 208mhz and its memtest stable over 24 hours holla. the power components seem to be doin well cuz i got a big 120 blowin some decent air across the the whole ram and mosfets section of hte board then out hte back (no exhaust fan so i put it nearer to the front of the case)

let me know if you guys have any iddea whats up or other general OC tips here. thanks!
 
did you drop the ht link multi? sounds like it's not the cpu limiting the rig, but something else.
 
set it to 1.55 + whatever percent, 1.45 on a winnie is actually 1.35 because the board was meant for claws and newcastles, winchester and venice run differnt voltages and your running VERY low right now
 
eclipse, link is at 4 so ending up at 1000 as it should

omega, so how do i see what the real voltage is thats being used?
 
el rolio said:
how do i see what the real voltage is thats being used?
Click to expand...
best way would be to take a voltmeter to the voltage regulation circuit. simpler way would be to use something like speedfan or everest that reports temps and voltages, just look for vcore
 
yea but eclipse, i mean that must not be reporting right. i consistently monitor with everest or mbm and speedfan in my other windows boot, but hte point is they report what cpu-z reports which is basically what i asked the bioss to do, when i was using hte 1.425 etc etc


however, when i tried this suggestion today before work, 1.55, it did report around where omega said it would then +3.3% then realised that was too low then +5% and well, thats where im testing at now.

they all are reporting 1.5~ with variation of .02 volts so we shall see. the temps looked about right for what shoudl be 1.5v

i just want osme further explanation cuz before omega came in here i didnt even know about this. i only knew about the fact that alot of selectable values were useles and you have to use the percentage over VID system instead.

uhm currently usin the 1.36b modded bios if that matters. have tried others, this one seems decent enough. 1.8 and 1.9 tried to kill my machine when they came out.
 
el rolio said:
yea but eclipse, i mean that must not be reporting right. i consistently monitor with everest or mbm and speedfan in my other windows boot, but hte point is they report what cpu-z reports which is basically what i asked the bioss to do, when i was using hte 1.425 etc etc
Click to expand...
well as long as it's not fluctuating a whole lot, i wouldn't really worry about it. if you must know what the actual values are, i can probably dig up some pictures of a neo2 and circle the spots where you should test for vcore :D
 
ha, but then i would have to go buy this tool you speak of to measure stuff etc hahah

anyhoo another thing is im not sure if i feel right about runin a high vcore for my lil ol 2.5ghz. i wanna run as least as possible you know what im sayin? ha sigh.

anyhoo, i guess i gotta go research this funny bios voltage thing on the neo2. you knew about this before eclipse?
 
eclipse wouldnt, hes a s754 man IIRC,

the NORMAL voltage, is the VID - .1V + X%

or you could just use corecenter and check, i boot at max voltage (1.55 + 10%) and simmer it down to 1.5 in core center, its just easier and is rock stable :D
 
so why does the computer report the voltages that you get when you do any other combination like i did before (1.45 + 8%) yet it proves to be incorrect?

how am i to trust the current voltages of 1.55 + 5% even tho it reports wat you said it would. i guess i wanna know how to get my voltages right without having ot do changes in windows. this is something i want to set, test, and live with 24/7

ps: i guess if it proves to be stable finally then thats all the proof ima need right
 
0mega said:
eclipse wouldnt, hes a s754 man IIRC
Click to expand...
true, but i own a k8n neo plat, which is nearly identical to it's s939 brother, quirks included... like how when trcd = 2, the bios automatically sets 2.7v, unless they fixed that recently with a new bios. :D

as for the VID, it'll make the 1.5v option whatever your default vid is. in my case, i would set it to 1.5v with my newark, and it would boot with just over 1.35v reported in windows. as a general rule of thumb, the voltage it tells you in windows is going to be relatively accurate, but if you really want to know for sure, a decent voltmeter is nice.
if you're gung-ho about it, hook an oscilliscope up to it and see how much vcore is rippling. less ripple = better. generally the cooler the mosfets are that supply the power to the cpu, the better voltage regulation will be, and the longer they'll last (as with all cmos, ~10c decrease in temps = ~2x longer life, assuming same voltage)
 
cool beans eclipse.

well basically that explains it a bit more, i guess i will always set to the max (1.55) then work with percentages from now on. and yes that bullshit quirk is still there so i have to set my bh-5 ram to 2-3-2-x since i have to run it at 2.85v just ot make sure my OC works since im usually up in the 210mhz range with ma ram. i evne set it to the 2-2-2-x this weekend thinking to myself "maybe it will change itself to 3 but no harm setting it correctly in the bios". this while i completely forgot it would reduce my voltage which of course reduced my stability. big time hahaha.

oh and my vcore doesnt fluctuate that much, or at least that wasnt a staying point with any of my statements here, plus i got that area decently cool and heatsinked. it was more about getting a firm grip and control over what voltage im feeding to the cpu and finally allowing me control to make sure my OC was stable..... so i can post in one of your database threads... haha. actually if im not 100% stable, and i mean 100%... then dvdsanta crashes when im trying to make a dvd to watch in bed. nothing else will crash or act up. so i could prime over night easily with other programs running, but dvd santa would find hte flaw. so im lookin forward to a REAL proepr OC (even though i guess i have been running 2.5ghz UNDERvolted al lyear)
 
yeah, that voltage bug on the ram really perplexed me until i said "screw this" and took a multimeter to the ram mosfets :p

also.. if other programs crash and prime95 doesn't throw an error, i'll bet the ram or memory controller is at fault
 
You must log in or register to reply here.
Back
Top