AMD "Thuban" Core (Phenom II X6) XS Overclocking Charts

[almighty15]

A load core temp of 55 or so is typically where Phenom II's, including the X6, start getting unstable.

Is that true?

[RussC]

Thats a good ball park. My 555 and 955 all started to get bad in the 55-60c range. All my best over clocks are definatley under 50c loaded. There are exceptions, some here have gotten air OC's in the 60s and 4+GHz. Those are very special parts and are not the norm. Out of 5 PII Ive had none would do anything over 3.6GHz stable on air&load. All would run to 80c @ stock speeds just fine.

RussC

Is that true?

[wez]

If you can keep it under 50C, you'll get very good clocks

[almighty15]

Mine will do 50 run runs of Intel burn test at 4.1Ghz with a load voltage of 1.55v with a CPU temp of 70 degrees C.

If I drop down to 4Ghz stable Max temp is between 55-60 degrees C.

My 1075T has a Thermochill PA 120.3 all to itself so how am I going to improve my cooling?

[BeepBeep2]

Particle there are issues I have with this list...

For example a post like Daveburts...(not to pick on him, it was just there)
He has 1.45v set, idles at 1.452, and in second screen 1.464, chances are load is at least 1.476-1.488v in CPU-Z and upwards of 1.49v on the DMM.

I can post 4.1 stable set at 1.43-1.44v, and 4 Ghz stable at "1.39v" and CPUNB at 3 Ghz at 1.33v if we all went that way.

My CPU doesnt even do 4.2 stable AFAIK most of the time I crash around an hour.

I'm not a believer in long stability tests especially at high volts considering my M4A89TD only has a 6+2 (doubled 3+1) phase setup. My mosfets are up in 95-100c range with just 4 Ghz/1.4v.

Typcally anything over 4 hours is overkill as well, all you will have past 4 hours is a CPU failing because of heat (unrealistic unless you run WCG) or a board failure.

I'd gladly run 4 hours, maybe 6-7 if I start it when I go to bed and turn it off or see what went up when I wake up...but I normally will not go for a 4.2 Ghz stable run at "1.51v" (1.54 real) load and not watch it.

[almighty15]

I'm Intel burn test stable at 4Ghz with a load voltage of 1.44v and Intel burn test stable at 4.1Ghz with a load Vcore of 1.5v

Max temp at 4Ghz according the CPU socket temp was 57c

Max temp at 4.1Ghz according to the socket temp was 62c

[felix_w]

Particle there are issues I have with this list...

For example a post like Daveburts...(not to pick on him, it was just there)
He has 1.45v set, idles at 1.452, and in second screen 1.464, chances are load is at least 1.476-1.488v in CPU-Z and upwards of 1.49v on the DMM.

I can post 4.1 stable set at 1.43-1.44v, and 4 Ghz stable at "1.39v" and CPUNB at 3 Ghz at 1.33v if we all went that way.

My CPU doesnt even do 4.2 stable AFAIK most of the time I crash around an hour.

I'm not a believer in long stability tests especially at high volts considering my M4A89TD only has a 6+2 (doubled 3+1) phase setup. My mosfets are up in 95-100c range with just 4 Ghz/1.4v.

Typcally anything over 4 hours is overkill as well, all you will have past 4 hours is a CPU failing because of heat (unrealistic unless you run WCG) or a board failure.

I'd gladly run 4 hours, maybe 6-7 if I start it when I go to bed and turn it off or see what went up when I wake up...but I normally will not go for a 4.2 Ghz stable run at "1.51v" (1.54 real) load and not watch it.

You have a good point on Vcore...my last submission here was with Bios Value and if you look at ss, voltage is 0.04v lower...

As for stability, my last is also 8hr prime, which finally seemed not to be rock stable....something wrong with my IMC maybe, i don't know, but it could not be revealed even with 8hr prime stress...

[BeepBeep2]

You have a good point on Vcore...my last submission here was with Bios Value and if you look at ss, voltage is 0.04v lower...

As for stability, my last is also 8hr prime, which finally seemed not to be rock stable....something wrong with my IMC maybe, i don't know, but it could not be revealed even with 8hr prime stress...

There are a lot of factors that come into play...

[felix_w]

Yeap, that's why the 8hr Prime doesn't seem to me to belong to my stability criteria anymore...just an intermediate step...

[BeepBeep2]

Yeap, that's why the 8hr Prime doesn't seem to me to belong to my stability criteria anymore...just an intermediate step...

Its not even that...

I run 1 1/2 hour OCCT Small for CPU only, 30 pass LinX for IMC and 4 Hour prime 95 for memory. Most of the time LinX will fail for IMC within 15 passes when Prime will take hours...

Then I game extensively like I usually do and within the next week if I dont crash, have programs crash or BSOD at any point that is when I will label it stable.

[almighty15]

Prime95 Small FFT's = CPU only stress?

Prime95 Large FFT = ?

Prime95 Blend = Whole system?

[wez]

prime95 small fft's = mostly high core load

prime95 large fft = high imc/ram load

prime95 blend = both small+large fft's, so a mix of both

...

[almighty15]

...

So in Theory then If I can do say 8 hours small and large fft's then I should be able to do 8hrs blend?

[wez]

So in Theory then If I can do say 8 hours small and large fft's then I should be able to do 8hrs blend?

Yea, in theory

[Particle]

If I mandated people use calibrated DMMs to measure vCore under load, we'd have virtually no entries where people could be classified as "stable". It would be even less useful than what we've got now, which I believe is a reasonable compromise. Since we don't exist as members of a laboratory, I just use the highest number I'm presented with. When people say "xV BIOS / xV load" or whatever, I pick whichever is higher so that the chart errors as much as possible on the side of worst case scenario. As for heat, if you're going to fail at due to thermals, you'll probably be caught within 8 hours. It's not uncommon for a CPU to fail only after extended durations even way beyond 8 hours. My personal standard is 24 hours with 48 being better, and I've seen failures at 25-26 hours for example. It doesn't take that long for things to heat up and saturate thermal solutions. Maybe your chip is on the fence. If it is, you're not stable regardless of the cause. Your definition of stable, beep, is more of a definition of stable-ish, and that's not good enough for me.

Blend is better than small or large FFTs imo. Who cares if your CPU is stable if the memory system isn't. It's an unusable system either way, you know?

No, the system isn't perfect, but it would be unrealistic to expect members of a forum like this to contribute better information on average. This is an informal setting.

[almighty15]

Currently running prime95 on blend at 4.1ghz with a load Vcore of 1.488v

Memory is at 1826Mhz @ 9-11-9-27-1T with 1.65v

NB is at 2739Mhz

HT is at 2191Mhz

Max temp is 59c and it's been going for 30 mins with no problems at all, even with me using the computer

On the old BIOS I need a load Vcore of 1.55v to handle 4.1Ghz

Will find my max CPU overclock then work on tweaking the RAM speed and timings

[BeepBeep2]

If I mandated people use calibrated DMMs to measure vCore under load, we'd have virtually no entries where people could be classified as "stable". It would be even less useful than what we've got now, which I believe is a reasonable compromise. Since we don't exist as members of a laboratory, I just use the highest number I'm presented with. When people say "xV BIOS / xV load" or whatever, I pick whichever is higher so that the chart errors as much as possible on the side of worst case scenario. As for heat, if you're going to fail at due to thermals, you'll probably be caught within 8 hours. It's not uncommon for a CPU to fail only after extended durations even way beyond 8 hours. My personal standard is 24 hours with 48 being better, and I've seen failures at 25-26 hours for example. It doesn't take that long for things to heat up and saturate thermal solutions. Maybe your chip is on the fence. If it is, you're not stable regardless of the cause. Your definition of stable, beep, is more of a definition of stable-ish, and that's not good enough for me.

Blend is better than small or large FFTs imo. Who cares if your CPU is stable if the memory system isn't. It's an unusable system either way, you know?

No, the system isn't perfect, but it would be unrealistic to expect members of a forum like this to contribute better information on average. This is an informal setting.

I didnt ask for DMM numbers. I asked for one standard.

That is either:
1. Post what you set and mention you have LLC on or off
2. Post the voltage CPU-Z tells you on load, not the voltage you have on idle or the voltage you set.
People are currently doing all three.

Daveburt's CPU is listed at 1.45v when he idles at 1.464 and to my knowledge of LLC he's up around 1.48 on load.
Particle I'm pretty sure I can pass 8-12 hours blend and fail LinX...I can also make it pass LinX and fail Blend...you tell me why.

Whats the point in failing at 8 to 12 hours because of heat?

On water we have equilibrium, it will even out after about 8-9 hours. ...on air you can get constantly rising temps if it is incapable of handling the heatload. Even by your standards you will never find out what stable really is. Also what you call stable and what I call stable can be different. It's more a matter of opinion. I can run my rig at 4.3 Ghz and game all day and night on it, encode video on it, make a sandwich with it but it will fail Prime95 within 10 minutes.

[Manicdan]

how reliable is cpuz with voltages? even if they are all off by a small amount, is it the same on all boards?

[almighty15]

On water we have equilibrium, it will even out after about 8-9 hours. ...on air you can get constantly rising temps if it is incapable of handling the heatload.

I'm water cooled and running Prime95 right now and my temps slowly climbed to 59c and that's were they have stayed for the last ~30 mins or so.

I'm running Blend for an Hour after that I'm calling it a day, I'm not going to waste time running Prime95 for hours when I should be enjoying my computer

[almighty15]

how reliable is cpuz with voltages? even if they are all off by a small amount, is it the same on all boards?

No idea, I do have voltage read out points on my Crosshair 4 Extreme but I don't have a multi meter to test accuracy.

[felix_w]

I measured on my M4A79 Bios 3702 and for my 1100T the following:

Bios Set:

1.5375

Idle DMM:

1.513

Load DMM:

1.523

Idle Cpu-Z:

1.534

Load Cpu-Z:

1.493

Other setting may have other δV. Cpu-Z is Version 1.57 x64

My DMM is a V&A VA18B, with fresh batteries.

Measuring point is on the back of the board, choke leg (as shown by CAL930 in this thread)

[almighty15]

I measured on my M4A79 Bios 3702 and for my 1100T the following:

Bios Set:

1.5375

Idle DMM:

1.513

Load DMM:

1.523

Idle Cpu-Z:

1.534

Load Cpu-Z:

1.493

Other setting may have other δV. Cpu-Z is Version 1.57 x64

My DMM is a V&A VA18B, with fresh batteries.

Measuring point is on the back of the board, choke leg (as shown by CAL930 in this thread)

Quite a difference if you ask me, I'll have to try and get a DMM from the girl friends dad and see how my Extreme compares

[BeepBeep2]

ASUS M4A89TD Pro/USB3 - LLC on

BIOS Set
1.40v

Idle -
CPU-Z 1.416v

DMM 1.419v

Load -
CPU-Z 1.428-1.440v

DMM 1.457v

DMM is an Extech MN35, measuring points are capacitor leads on 3rd phase from bottom...

Essentially if I went with the same method Daveburt went I could claim 1.4v when I'm actually close to 1.46 or even 1.44 according to CPU-Z (again, not to pick on him, just an example)

[almighty15]

Yes! Prime95 Blend for 1hr at 4.1Ghz with 1830Mhz RAM

Load Vcore of 1.488v

On previous BIOS it was 4.1 with 155v load Vcore with max temp of 70c

Will aim for 4.2Ghz tomorrow

[BeepBeep2]

Oh particle btw:
I've had P95 Blend fail at 9 hours (went to sleep) before due to using box cooler and ram heating up from CPU heat
2400 NB and 3600 CPU was obviously not the cause for a rounding error...and sticks do not fail much after 4 hour range, if it fails after 5 its because of heat build up.

There are way too many factors to define stable...