***official amd 83x0 overclocking thread***

[Panos]

Thanks a lot guys, that's the most comprehensive help I've been offered and I will put it to good use. Had a bit of a nightmare today but will come back strung tomorrow, not good when you have a high end potential rig running at stock speeds!

Thanks again chaps, have a good weekend!

Thank you, you too.

And you are welcome.

FYI my mate using the CCC Overdrive from the latest Catalyst 13.11 with everything stock settings in BIOS, got it to overclock the FX6100 @ 4400

 

[petertr2000]

Could anyone help me with a wall that I seem to have hit?


Just set up a new FX8320 system. H80 cooler, push/pull. Idle socket temps are around 27C ("Package", i.e. Core temps are wierd at idle, from around 10C to 20C - but I hear the sensors don't work well at idle).

2 x 4Gb 1867 RAM, using stock profile in the BIOS (which matches the timings exactly).

Ran P95 overnight, no issues at all. Socket temps maxed at 55C, Core temps were much lower (sorry, can't remember).

So I'm happy that everything is set up correctly and that my RAM is working fine with no errors.



I've been trying to overclock using the multi all morning, with strange results.
I've found the limit at stock voltage is 4.2Ghz (Load Core Temp 40C, Socket 60C). But when I try to raise the voltage by one single increment (around 0.005V), the system simply won't boot. If I put it back to Auto, its fine.

However, if I reset the BIOS back to stock, I can raise the voltage just fine and Prime has no issue. The temps don't rise much either. I ran it at 1.425v to test it (seeing as that is the P-Stage VCore for the turbo setting, I figured it was as good a place as any to start).


Anyone know why I can't raise the VCore when the CPU is being overclocked?



One thing, the VRM heatsink gets really hot. So I got an old 90mm fan and it cooled right down - didn't make any difference, though.

Spec: Asus M5A97 EVO R2.0 / AMD FX-8320 / 2x4GB 1867 Patriot Red / MSI 2GB 7850 / BeQuiet Pure Quiet L8 500W

 

[teppic]

When you use auto, it may be raising the voltage automatically along with clock speed. Try setting the voltage to 1.45v, make sure turbo is disabled, and try setting the multiplier to get 4GHz, then go up 100MHz at a time until you stop being stable. It's an easy way to start off.

The VRM heatsink will get hot, but it shouldn't cause a problem on your board unless you're going to very high clocks.

 

[ALXAndy]

I relished getting my M5A97 set up. Mostly because I knew what a challenge it was going to be, being so heavily VRM limited.

After two days of clocking the absolute best I could get out of it was 4.3ghz. Note - with that board you can set the voltage as high as you like but you're going to end up with around 1.34v max to work with. Even though you can set higher (I set 1.4v) the most the VRMs will give out is around 1.3-1.4.

4.2ghz is usually the limit on a 4+2 board tbh. I guess I was lucky that my second 8320 would do 4.3 using 1.3v. On my old Asrock I had around 1.31v (4+1) and the best I could do was 4.2ghz. So I was pretty pleased at the extra 100mhz I got out of the Asus.

Something else I noticed was that cranking on the multi needs more voltage and generates more heat on the CPU (and thus power use) than using the FSB. So ideally you want to work on the FSB alone, then hit the multi at the last moment.

 

[BurntProc]

Could anyone help me with a wall that I seem to have hit?


Just set up a new FX8320 system. H80 cooler, push/pull. Idle socket temps are around 27C ("Package", i.e. Core temps are wierd at idle, from around 10C to 20C - but I hear the sensors don't work well at idle).

2 x 4Gb 1867 RAM, using stock profile in the BIOS (which matches the timings exactly).

Ran P95 overnight, no issues at all. Socket temps maxed at 55C, Core temps were much lower (sorry, can't remember).

So I'm happy that everything is set up correctly and that my RAM is working fine with no errors.



I've been trying to overclock using the multi all morning, with strange results.
I've found the limit at stock voltage is 4.2Ghz (Load Core Temp 40C, Socket 60C). But when I try to raise the voltage by one single increment (around 0.005V), the system simply won't boot. If I put it back to Auto, its fine.

However, if I reset the BIOS back to stock, I can raise the voltage just fine and Prime has no issue. The temps don't rise much either. I ran it at 1.425v to test it (seeing as that is the P-Stage VCore for the turbo setting, I figured it was as good a place as any to start).


Anyone know why I can't raise the VCore when the CPU is being overclocked?



One thing, the VRM heatsink gets really hot. So I got an old 90mm fan and it cooled right down - didn't make any difference, though.

Spec: Asus M5A97 EVO R2.0 / AMD FX-8320 / 2x4GB 1867 Patriot Red / MSI 2GB 7850 / BeQuiet Pure Quiet L8 500W

As teppic said, on auto the board is automatically cranking up your volts to a high level.

You need to start from stock and do some more initial setup, disable power saving features (cool n quiet, C1E) and set your voltages to manual.

You can leave most voltages on auto but manually set CPU VCore and CPU/NB Voltage to be begin with. Then set CPU load line calibration to medium / high and CPU autocurrent protection to 120% or 130%.

Then start your clocking as you did previously and you'll have better results (and be in total control as you do so).

 

[teppic]

I relished getting my M5A97 set up. Mostly because I knew what a challenge it was going to be, being so heavily VRM limited.

After two days of clocking the absolute best I could get out of it was 4.3ghz. Note - with that board you can set the voltage as high as you like but you're going to end up with around 1.34v max to work with. Even though you can set higher (I set 1.4v) the most the VRMs will give out is around 1.3-1.4.

4.2ghz is usually the limit on a 4+2 board tbh. I guess I was lucky that my second 8320 would do 4.3 using 1.3v. On my old Asrock I had around 1.31v (4+1) and the best I could do was 4.2ghz. So I was pretty pleased at the extra 100mhz I got out of the Asus.

Something else I noticed was that cranking on the multi needs more voltage and generates more heat on the CPU (and thus power use) than using the FSB. So ideally you want to work on the FSB alone, then hit the multi at the last moment.

He's got the EVO R2.0 though, which people have got up to 4.8GHz or so. It has a 6+2 power design and good quality heatsinks.

 

[petertr2000]

Thanks v much for the tips - Never crossed my mind that "auto" was increasing the voltage each time.
I should have spotted that in CPU-Z.

Currently reading, Idle : 1.380V / Full Load :1.308V @4.2Ghz

Will look into those other tips too - must get some sleep though, as on night shifts and I was up most of last night with issues.

& Yes, its the EVO motherboard, it has fairly chunky blue heatsink/fins for the VRM and Northbridge. I *thought* it was a decent board for overclocking. But no?

 

[teppic]

Thanks v much for the tips - Never crossed my mind that "auto" was increasing the voltage each time.
I should have spotted that in CPU-Z.

Currently reading, Idle : 1.380V / Full Load :1.308V @4.2Ghz

Will look into those other tips too - must get some sleep though, as on night shifts and I was up most of last night with issues.

& Yes, its the EVO motherboard, it has fairly chunky blue heatsink/fins for the VRM and Northbridge. I *thought* it was a decent board for overclocking. But no?

It is. The next step up is the Asus 990FX Pro, but that's about £120.

And that's a big vdroop there. You want to set the LLC up.

 

[petertr2000]

Sorry for the delay in posting, been asleep for a couple of hours.

Tried to put the Vcore up to 1.4 - no boot. Tried 1.42, no boot & had to remove the bios battery and reset. (also went back and tried 1.45V as suggested above, no difference - no boot to windows)

Booted back with stock settings, and the idle VCore is THE SAME as it was on Auto when I had it overclocked to 4.2Ghz. Its currently running at 3.5Ghz (load VCore isn't as bad at 1.325)


So that rules out "the bios was automatically raising the voltage for you".

So why can't I raise the voltage without it having a fit?





Also, running Prime again at 4.12Ghz, and the loaded Vcore is averaging at ~1.315V

 

[Hagar]

Do you have voltage offset?

Make sure turbo mode is disabled, set CPU LLC to ultra high, CPU/NB LLC to high. Set CPU over-current protection to 130%, set CPU/NB over-current protection to 120%

Set CPU PWM phase control to extreme. CPU PWM mode to T-probe, CPU voltage frequency to auto, CPU spread spectrum to off.

Set CPU voltage mode to offset, set offset to +0.100V, leave CPU/NB voltage offset to auto.

You may find by trial and error, a lower offset works.

Hope this helps, andy.

 

[petertr2000]

Do you have voltage offset?

Make sure turbo mode is disabled, set CPU LLC to ultra high, CPU/NB LLC to high. Set CPU over-current protection to 130%, set CPU/NB over-current protection to 120%

Set CPU PWM phase control to extreme. CPU PWM mode to T-probe, CPU voltage frequency to auto, CPU spread spectrum to off.

Set CPU voltage mode to offset, set offset to +0.100V, leave CPU/NB voltage offset to auto.

You may find by trial and error, a lower offset works.

Hope this helps, andy.

I've tried using offset and manual voltages. Made no difference.

Turbo mode is off.

I'll try the rest tomorrow - on night shift at the moment and didn't bring the PC with me.

 

[BurntProc]

So why can't I raise the voltage without it having a fit?

It may be the power saving features or CPU overcurrent protection.

 

[petertr2000]

Do you have voltage offset?

Make sure turbo mode is disabled, set CPU LLC to ultra high, CPU/NB LLC to high. Set CPU over-current protection to 130%, set CPU/NB over-current protection to 120%

Set CPU PWM phase control to extreme. CPU PWM mode to T-probe, CPU voltage frequency to auto, CPU spread spectrum to off.

Set CPU voltage mode to offset, set offset to +0.100V, leave CPU/NB voltage offset to auto.

You may find by trial and error, a lower offset works.

Hope this helps, andy.

Followed your instructions - except for adding an increment to the voltage - wanted to see if all of the other settings made any difference. They certainly did. VCore was reporting just over 1.4v, and under load, it actually went up by 0.01v - instead of dropping by nearly 0.08.
Did a 5 min P95 and all ok.

So rebooted and set the increment as you said. VCore now reporting as 1.368 - a massive drop - looks like "Auto" was doing something after all!

Will have a play now, hopefully I have control of the voltage now.
Thanks!

 

[petertr2000]

So I'm currently running 4.4Mhz @ 1.404V, Prime is running and my temps are
Socket 64C / Core 43C

CPU-Z is showing occasional downclocking though, from 22X to 7X. Any idea why that would be?

I don't think those temps are high enough for CPU throttling, are they?

 

[gpuerrilla]

CPU-Z is showing occasional downclocking though, from 22X to 7X. Any idea why that would be?

Have you still got any power saving features enabled?

Throttling apparently starts into the 60's but the gauges aren't that accurate to be fair.

 

[ALXAndy]

He's got the EVO R2.0 though, which people have got up to 4.8GHz or so. It has a 6+2 power design and good quality heatsinks.

Oops my bad.

 

[ALXAndy]

So I'm currently running 4.4Mhz @ 1.404V, Prime is running and my temps are
Socket 64C / Core 43C

CPU-Z is showing occasional downclocking though, from 22X to 7X. Any idea why that would be?

I don't think those temps are high enough for CPU throttling, are they?

It's usually the VRMs that throttle more than the CPU itself. Try putting a fan over your VRMs and north bridge.

And FFS for the last time. DO NOT RUN PRIME 95 ON AN FX. Run 5 passes of LinX.

 

[teppic]

So I'm currently running 4.4Mhz @ 1.404V, Prime is running and my temps are
Socket 64C / Core 43C

CPU-Z is showing occasional downclocking though, from 22X to 7X. Any idea why that would be?

I don't think those temps are high enough for CPU throttling, are they?

You're on an H80 right? Temps seem ok, certainly the core is nothing to worry about. 22x to 7x sounds like power saving rather than throttling - does it happen when you're stress testing? You shouldn't be getting VRM throttling unless your case has no air flow.

And FFS for the last time. DO NOT RUN PRIME 95 ON AN FX. Run 5 passes of LinX.

I think Andy means, 'I wouldn't recommend Prime95, try LinX instead'

 

[petertr2000]

Have you still got any power saving features enabled?

Throttling apparently starts into the 60's but the gauges aren't that accurate to be fair.

No they've been switched off the whole time.

After 7 hours in Prime, at 4.4Ghz / 1.416V, the Core temp peaked at 60C - Socket peaked at 80C.
Although as I sit in front of it now, its 52C/72C

 

[petertr2000]

It's usually the VRMs that throttle more than the CPU itself. Try putting a fan over your VRMs and north bridge.

And FFS for the last time. DO NOT RUN PRIME 95 ON AN FX. Run 5 passes of LinX.

Any link that explains why not to use Prime? I've seen it recommended by people for years?