Upper limits

Khannie

The quads have been around for quite a while now. What have people found to be the safe upper limits for daily usage for fsb and voltage say on air or water?

There used to be a superb chart for AMD chips when they were king. It had "safe", "some risk" and "high risk" on voltage.

I think it'd be useful to gather all of the collective overclocking destruction / knowledge in this forum in some way.

Khannie

Oops....I'll start: My Q6600 is running a 9x multi @ 1.375V and 333FSB (giving 3.0GHz for daily use). I consider this in the "safe" range and doubt that it will negatively impact the life of the processor.

awhir

1.5v seems to be the safe are for quads while 1.6v seems to be the quite risky (but nothing ever goes worng)

Ur Q6600 is a 105w so i would say you wont get past 1.4v as of the temps.

Khannie

awhir wrote: »

Ur Q6600 is a 105w so i would say you wont get past 1.4v as of the temps.

I don't really understand this. Can you expand on it a bit? Will 1.4V = too much heat for my older Q6600? How would you determine this? (not you in particular...."one")

awhir

well the older Q6600 b2 ( non Go stepping) runs Very hot.So with Air cooling i cant see you going over 1.4v since you will most likly be hitting 70c+.

I dont have a q6600 my self but even if you look around boards you will see how hot the B2's are (lmrds rig/pog's/anti's).

also another problem with overclocking quads is the chipset.The realy only good chipset for clocking quads are the P35 and X38's, while p965 does allright they ussaly crapp out at around 300-330 fsb.

Anti

Great reply andy I have ot agree with it all.

L31mr0d

yeah can't really add anything onto Awhirs.

Also, we can't tell you what the safe upper limits will be as each chip is different. Also I think the problem with my QX6700 (Awhir, how do you know its a B2? I don't even know that ) was my P180. I've been leaving the side off and my quad idles at 25'c, with the side panel on it was hitting 40'c idle.

also, as Awhir said, it depends on the chipset of your mobo also, i'm finding at the moment that overclocking on my 680i is far from the same walk in the park that my old 939 was. From reading around, nvidia cut a few corners to be able to offer a QDR of over 1600Mhz, being that at each strap change the NB latencies loosen so much to accommodate the higer frequency that the overclock is pointless.

i.e. you could be getting better performance at FSB499 than you are at FSB500.

What motherboard do you have?

Khannie

I currently have a 975X chipset board (P5W DH Deluxe). There are two versions of this board from what I can tell and I have the newer / better one. I haven't hit any fsb barriers with it, but didn't get beyond 3.15GHz before deciding that I was happy with 3.0GHz for everyday use. I need to reapply the HSF as I discovered today that it's probably come loose (remotely, I ran mprime and it went to sh1t in seconds. I had cause to run mprime because parity repairs were failing and I ruled out ram and hdd).

So...just to summarise (motherboard complications aside):
Q6600 b2 ( non Go stepping) with air cooling -> safe up to 1.41, some risk 1.41-1.45? high risk > 1.45?

What are the other types of quad then?

edit: I am considering a new motherboard because I've ordered an SAS drive and controller and have read that the controller wont work with the P5W DH Deluxe.
edit;2: I should say that I'm happy with 3.0GHz. My biggest bottleneck is my OS HDD and I'm hoping the SAS drive will sort that out for the most part. Feck off Chosen. Don't even start.

TheThreeDegrees

go on tell us which ssd did you go for?

TonyM.

For everyday use 3300.00mhz and 1200 fsb.

Khannie

TheThreeDegrees wrote: »

go on tell us which ssd did you go for?

That's the thing....I didn't.....I went with an SAS hard drive because of budget. 15k rpm, 146G, 2ms seek time. Pretty sexy but it has its drawbacks. Still waiting for it to arrive.

Chosen

a) I have my G0 Q6600 folding 24/7 at 3600MHz with 1.328v.
b) P965 is the 2nd best overclocking chipset for quads. Had my old Q6600 and my QX6700 running at 450MHz fsb no probs.
c) There was never a B2 stepping on quads, just C2D's. C2D: B2->G0 & C2Q: B3->G0.
d) Best 975X board with quads, was the XBX2 with a stable 370-380 MHz on the FSB.

RoyalMarine

the qx6700 is b3 stepping.
no quads have b2 stepping. only b3 and go.

awhir

Chosen wrote: »

a) I have my G0 Q6600 folding 24/7 at 3600MHz with 1.328v.
b) P965 is the 2nd best overclocking chipset for quads. Had my old Q6600 and my QX6700 running at 450MHz fsb no probs.
c) There was never a B2 stepping on quads, just C2D's. C2D: B2->G0 & C2Q: B3->G0.
d) Best 975X board with quads, was the XBX2 with a stable 370-380 MHz on the FSB.

b)yes the late p965 boards warrent bad (dfi dark,asus something ?,ds3) but still you were lucky if you could get a board that could do 450mhz...

c)sorry my bad allways get confused with steppings :rolleyes:

d)The Intel board was one of the good ones but overall 975 was crap for clocking quads just look at the abit AW9D-MAX that crapped out at around 300fsb.The memory controllers couldent handle quads.

po0k

Khannie wrote: »

I don't really understand this. Can you expand on it a bit? Will 1.4V = too much heat for my older Q6600? How would you determine this? (not you in particular...."one")

TDP = fCV^2
f = Frequency
C = capacitance (smaller process = lower capacitance generally)
V = voltage.

Increases in voltage will increase your TDP exponentially.
Try and keep the volts as low as you can while stable.
Aside from that, high voltage will literally wear away the silicon (or cause deposition in the gates).

gline

G0's, safe is about 1.45V - 1.5V so i read online (with decent aircooling, not standard intel cooler).

I ran that on my E6550

Anti

my q6600 is happy to run upto 1.6v but i am able to shift the heat. But i would recomend 1.5v to be the highest you should use anythign else should have water/ln2 etc etc

Khannie

SyxPak wrote: »

TDP = fCV^2
f = Frequency
C = capacitance (smaller process = lower capacitance generally)
V = voltage.

Believe it or not, I made money from that forumla for 5.5 years.

Yeah, I knew about voltage increases resulting in a square for power. My question was whether newer ones could handle a higher voltage because perhaps they were using lower capacitance or something (though I wasn't gonna ask about lower C values ). I'm not hugely concerned about the wear and tear on the silicon. I've never overclocked my chips far enough to shorten their actual lifetime.

The real purpose of the thread was to try to gather a table together of safe voltage levels for the various chips. At the time that I got my Q6600, they were a real rarity (at the time they were horribly priced, so everyone who was going quad was going with the qx6700's. I just got mine cheaply.).

Anti, how are you shifting that much heat?

Anti

125cfm fans

gline

Anti wrote: »

125cfm fans

probably sounds like a wind tunnel . eh??

Khannie

pics plz

Khannie

Righto lads....I'm a smidge freaked out.....

The CPU was happily running at 3.0GHz up until recently. I figured the heatsink had maybe come loose so reseated it. Now running 4 instances of mprime at 2.7GHz, 1.32VCore and the temperature sensor says 76C (there's one sensor says 95C, but it's not clear what it belongs to, so I'm assuming it's lying). This seems very high for such a low VCore. Am I wrong?

I should point out that even with the reseated heatsink, the cpu wasn't stable at 3.0GHz with the old vcore values. Have I maybe just made a balls of the reseat?

PogMoThoin

use core temp. Dunno what utility ur usin but i wouldn't trust it.

Khannie

I'm using linux. It's the coretemp module for the kernel, so should be accurate. Says I'm currently idling at 33C.

PogMoThoin

Khannie wrote: »

I'm using linux. It's the coretemp module for the kernel, so should be accurate. Says I'm currently idling at 33C.

I also use linux, whats that application?

Khannie

lm_sensors. The gui bit is just a gnome applet frontend for it.

Just ssh'd home. Here's the output from sensors directly (which I hadn't examined directly before....was just using the gui):

paul@computa ~ $ sensors
w83627dhg-isa-0290
Adapter: ISA adapter
VCore:     +1.26 V  (min =  +0.00 V, max =  +1.74 V) 
in1:      +12.41 V  (min =  +0.00 V, max =  +2.59 V) ALARM
AVCC:      +3.26 V  (min =  +0.77 V, max =  +0.83 V) ALARM
3VCC:      +3.26 V  (min =  +2.19 V, max =  +1.55 V) ALARM
in4:       +1.17 V  (min =  +0.29 V, max =  +0.61 V) ALARM
in5:       +1.58 V  (min =  +0.13 V, max =  +1.65 V) 
in6:       +3.84 V  (min =  +4.40 V, max =  +3.94 V) ALARM
VSB:       +3.26 V  (min =  +2.46 V, max =  +0.10 V) ALARM
VBAT:      +3.18 V  (min =  +0.54 V, max =  +2.30 V) ALARM
Case Fan:    0 RPM  (min =   48 RPM, div = 128) ALARM
CPU Fan:  1196 RPM  (min = 2518 RPM, div = 8) ALARM
Aux Fan:     0 RPM  (min =   79 RPM, div = 128) ALARM
fan4:        0 RPM  (min =   56 RPM, div = 128) ALARM
fan5:        0 RPM  (min =  390 RPM, div = 128) ALARM
Sys Temp:    +25 C  (high =   +96 C, hyst =   +68 C)  [thermistor]
CPU Temp:  +26.0 C  (high = +80.0 C, hyst = +75.0 C)  [CPU diode ]
AUX Temp:  +46.0 C  (high = +80.0 C, hyst = +75.0 C)  [thermistor]

coretemp-isa-0000
Adapter: ISA adapter
Core 0:      +45 C  (high =  +100 C)                   ALARM

coretemp-isa-0001
Adapter: ISA adapter
Core 1:      +44 C  (high =  +100 C)                   ALARM

coretemp-isa-0002
Adapter: ISA adapter
Core 2:      +40 C  (high =  +100 C)                   ALARM

coretemp-isa-0003
Adapter: ISA adapter
Core 3:      +40 C  (high =  +100 C)                   ALARM

edit: Yeah....there we go....the individual cores are reporting 40 odd, but the "CPU Temp" is reporting 26 now. I wonder if that 26 is a sensor on the motherboard. mprime ran for an 1:22:00 before I killed it, so it's stable at the moment anyway.

I'm gonna re-apply the heatsink again tonight. Christ I hate doing that. I need to take the whole mobo out to do it. X(

conzy

windows tbh

Lots of tools for probing temperature

Khannie

conzymaher wrote: »

windows tbh

Lots of tools for probing temperature

Ridiculous. It's a piece of hardware on the chip. They all do the same thing that lm_sensors does. l2c.

conzy

But yours seem as reliable as a <very unreliable example> !

L31mr0d

test if the base of the IHS or HSF is concave or convex. If it is then lap it.

Also, could you run a Windows temperature monitor from within WINE?

Khannie

To get back to my question a few posts ago...is 76C very high for 2.7GHz @ 1.32V?

L31mr0d wrote: »

test if the base of the IHS or HSF is concave or convex. If it is then lap it.

I did look at the base of the hsf when I had taken it off and cleaned it. It /seemed/ like a flat finish, but I'll give it a proper test when I reseat it for the final time.

Also, could you run a Windows temperature monitor from within WINE?

Haven't tried to be honest. I could give it a lash alright, but I doubt it'd work.