AMD 45nm median voltage limits:

[chuchnit]

I just use what voltage is needed until it stops scaling. It's not like I plan on keeping a CPU forever anyway.

[chew*]

1.475 is max i would bother with most chips for 24/7. TBH if I had the time and ripped my rig apart and put the amd system in thats been waiting for quite some time, I would be happy with 3.6 24/7, then again my 24/7 pc is mission critical, Quiet and cool.

The point at which is takes excessive voltage for minimal gains for 24/7 use is the point that you will degrade varies chip to chip but 1.475 is preety much safe for all chips.

Best way to find out is to remove the buffer (IHS) run on cool water at a max prime stable speed.

Let water get room temps then boot. It will instapop the chip think that happened at 1.5v IIRC but not 1.475 with quite a few test candidates.

If you can instapop, then it's fairly safe to assume you can degrade with the buffer in place.

[BeepBeep2]

1.475 is max i would bother with most chips for 24/7.

interesting...

Updated chart, don't know if it's much better.

I started this thread to help avoid people going "I put 1.70v on my Phenom and it still fails prime whats wrong"

[Particle]

interesting...

Updated chart, don't know if it's much better.

I started this thread to help avoid people going "I put 1.70v on my Phenom and it still fails prime whats wrong"

Then they need more vcore, clearly.



Looking good

[Tomasis]

great!

[panzerchaos47]

Nice to hear thubans can do 1.5 vcore for 24/7. I was reading there website so I tried stay under 1.4 vcore. I was going post question the CPU-NB core voltage. Does this apply for CPU-NB core? NB voltage start at 1.15 vcore I was wondering increasing them to 1.4 is too much?

[BeepBeep2]

Nice to hear thubans can do 1.5 vcore for 24/7. I was reading there website so I tried stay under 1.4 vcore. I was going post question the CPU-NB core voltage. Does this apply for CPU-NB core? NB voltage start at 1.15 vcore I was wondering increasing them to 1.4 is too much?

I'd stay under 1.4v for 24/7, I could later update this thread to include CPU/NB voltages but that can wait.

I shoved 1.55v into my Deneb's CPU_NB once for SuperPi 1M and it doesn't appear degraded but thats stongly ill-advised.

[slaveondope]

Id like to hear more from people who have actually degraded any of these chips under "normal" overclocking.
Ive had a 7 year old Athlon64 running at 1.65v for 1 year full load 24/7 on stock cooler with no issues. I know not same processes.
Along with a Athlon 620 at 1.55v and 955c3 at 1.5v 24/7 full load for 1.5 months and ran those volts for 6 months prior both under 55c.
Then I have a Thuban at 1.52v at ~45c full load 24/7 last 1.5 months. As of yet I havent seen any I'll affects on any cpu Ive ever owned.
So lets hear some horror storys

Added: btw I like what you got going on here Beep sub'd and I'll be looking forward to hearing more about the NB voltages

[cdawall]

i ran my 945BE (RB-C1) at 1.67-1.72v 24/7 on water for months and never ran into degradation had it buzzing at 4ghz way before most people could get that high (chew* excluded obviously)

my current 1090T is pretty happy at 1.5625v (1013CPMW stepping) but with vdroop thats only like 1.475ish from meter readings on the CH3 benchmarking over that i haven't seen improvements on tec/air with phase 1.6v is about the max i saw scaling on

[afireinside]

I ran 1.63v for months on water and previously on air for months. Recently I started crashing, but I was on the ragged edge of 24/7 stability and ambient temps have increased greatly. I don't want to say it HASN'T degraded, but I doubt it.

Best way to find out is to remove the buffer (IHS) run on cool water at a max prime stable speed.

Let water get room temps then boot. It will instapop the chip think that happened at 1.5v IIRC but not 1.475 with quite a few test candidates.

If you can instapop, then it's fairly safe to assume you can degrade with the buffer in place.

What do you mean by "instapop"?

[BeepBeep2]

I ran 1.63v for months on water and previously on air for months. Recently I started crashing, but I was on the ragged edge of 24/7 stability and ambient temps have increased greatly. I don't want to say it HASN'T degraded, but I doubt it.



What do you mean by "instapop"?

The chip goes "pop" and dies...metaphorically speaking...

[afireinside]

Well as much as I didn't want to admit it... 1.6v+ for extended periods of time (5 months on water, no idea how many on air when I first got it) has damaged my 940BE. Needs increased vNB for stability at stock speeds.

[BeepBeep2]

Well as much as I didn't want to admit it... 1.6v+ for extended periods of time (5 months on water, no idea how many on air when I first got it) has damaged my 940BE. Needs increased vNB for stability at stock speeds.

vNB? isn't that a motherboard voltage? Even if that's CPU_NB it shouldn't have anything to do with core voltage should it? That's kinda scary...

[DEVIL K-ce]

If cpu is scale with voltage then I push it to max 1.55v for 24/7 on air ofc. This is on new am3 cpu's, in era of am2/am2+ cpu smth like 1.6v with good air cooling isnt problem ;] Good = Mugen/Ninja ... push/pull setup and you got some MHz.

[afireinside]

vNB? isn't that a motherboard voltage? Even if that's CPU_NB it shouldn't have anything to do with core voltage should it? That's kinda scary...

CPU_NB is what I meant. I guess it's possible I damaged the IMC? Was running that at 1.425 I believe. I remember seeing an official paper from AMD (that I can't find, go figure) and I want to say it said 1.55 was on for CPU_NB, but maybe not. Either way my CPU is NOT happy.

[gOtVoltage]

CPU_NB is what I meant. I guess it's possible I damaged the IMC? Was running that at 1.425 I believe. I remember seeing an official paper from AMD (that I can't find, go figure) and I want to say it said 1.55 was on for CPU_NB, but maybe not. Either way my CPU is NOT happy.

CPU_NB or NBvid or NBIMC ive always kept below 1.55v as per AMD for the IMC DDR2/DDR3.. I see Thurban being slightly lower due to more cores and stress on IMC with higher Freqency Rams.

Just think of the IMC(NB) as another CPU... 1.45volts@3100mhz .

Hence the memory alone coupled with High NB is like having a Supercharger in your CPU .

The Extra Volts up to 1.45v will indeed get things cooking across 6cores and 1800mhz+ ram..

Adding to much Volts or using the Max Volts usually spells Misfire. Voltage is like your Boost.. To much Boost (28psi on the street (and poof there goes the intake/Lower end To0 little volts and your NB limited( like a stock supercharger with 6psi).

Keeping a happy Voltage medium on the NB is priceless..Ive maxed lots of CPU's and ive found that 1.4volts or lower is all ive ever needed on C2/C3..

Thurban is an animal , just remeber More NB(IMC) volts will add heat faster than X4cores .Two more cores pushing the IMC(= more Stress).

just my 2cents

If i had a Thurban id VoltPush it to the moon for Ya

[BeepBeep2]

I've pushed my 965BE to 1.55v on the CPU_NB for bench sessions.

I try to keep under 1.45v for 24/7. Maybe that will be 1.40v now.

[haylui]

my Phenom II x3 720BE quad-core runs 3.43GHz @ 1.48V 24/7, most of the time lightly loaded.

[haylui]

I've pushed my 965BE to 1.55v on the CPU_NB for bench sessions.

I try to keep under 1.45v for 24/7. Maybe that will be 1.40v now.

4.1GHz @ 1.40v?
what a nice chip you have

[buff]

4.1GHz @ 1.40v?
what a nice chip you have

I believe that was his IMC voltage, not cpu vcore

[JumpingJack]

This is lovely!

If my Deneb need 1,48 volts @ 4ghz but I read 1,45 volts in Cpu-Z (Voltage drop).

Do I consider the real voltage as 1.48 or 1,45?

1.48?

CPU-Z reports the voltage that is monitored by the support chipsets such as Winbond or ITE based sensor chips. These are IO chips specifically made to provide montioring, sensing, and feedback (reference) for VRMs, etc.

The quality of the voltage reported by CPUID is directly related to the quality of these chips. They are typically pretty good, in the Asus boards I have looked at in detail, the voltage reported is almost always dead on, to the 3rd sig fig, with a measured voltage from a calibrated Fluke.

Having said that, voltage droop is unavoidable (though some MB makers try load line calibrations to correct for it, none are perfect), VRMs under load will provide less voltage (lesser than the requested) than at idle. It is much like a pressure regulator in which you allow air to flow constantly, the measured pressure past the regulator is always less than the set pressure without air flowing.

Thus it is correct to assume the voltage on the CPU is the reported voltage as a drooped VRM is actuall providing that voltage, not the requested.

Jack

[digitalfrost]

I should have read this thread sooner. I just killed a X2 240. Hell of a chip, did 4.0Ghz without any additional voltage, 4.1 needed some extra volts. While trying to get it stable at 4.2 I set the voltage to 1.6v in BIOS (Board overvolts about 0.025v under load). I ran IntelBurnTest for some time, then I smelled burned electronics and the chip was dead.

2nd Chip is now stable at 4.0Ghz 1.475v. I'm gonna leave it at that.

[Ccaution]

I should have read this thread sooner. I just killed a X2 240. Hell of a chip, did 4.0Ghz without any additional voltage, 4.1 needed some extra volts. While trying to get it stable at 4.2 I set the voltage to 1.6v in BIOS (Board overvolts about 0.025v under load). I ran IntelBurnTest for some time, then I smelled burned electronics and the chip was dead.

Are you sure it not the board that bit the bullet though? It seems - it's rather easier to kill your board that you cpu nowadays...

People with Thubans & phase changing @ 24/7, should know what I mean

[pipoo]

Those voltages are under water right? What about them with a good (high-end) air cooling?

[Manicdan]

if its just 2 cores it might not be that bad with temps