Overclock.net banner
1 - 17 of 17 Posts

Thedarklord

· Registered
Joined
·
145 Posts
Discussion starter · #1 ·
So, I am having trouble getting a stable overclock and am wondering what I am missing on this AMD motherboard...

Specs;
CPU: Phenom II X4 960T (3.0 Ghz Stock)
Mem: Corsair 16GB DDR3 1600 Mhz
Mobo: Asus M5A99FX Pro R2.0

The first thing I did originally was I set the memory from the stock 1333Mhz and set it manually at 1600Mhz, this was rock stable and had no issues.

I then starting pushing the multiplier up in 200Mhz increments, all the way to 3.6 Ghz, on stock voltage. This is 100% stable.

When I set the base to 3.8 Ghz and the boost to 4.0 Ghz, the machine started to freeze very early on, most of the time it wouldn't even boot to Windows. I started bumping the voltage offset, but got up to about 0.085V increase, and it did get more stable, but would still have errors on games, and would eventually freeze/BSOD.

Any ideas on this?
 
Speaking from experience.

1st. Disable Spread Spectrum
2nd. Disable Turbo Boost.

Is your CPU unlocked to 1600/1605T yet?

If you are going to overclock it only as a quad, 1.4V is a must for 3.8Ghz.
If you intend to unlock it, 1.4V for 3.4.Ghz is the minimum, 1.425 for 3.6Ghz (i run it at 1.4125V 3.6).

First, set the CPU voltage to 1.4V, set the Offset to +50mV, set the multiplier to 3.8Ghz, try to boot. If you boot, launch IBT and try running 5 times on Maximum to see if its stable and if it will handle the temps.
 
Quote:
Originally Posted by knifemind View Post

What is your current CPU volts? Most Thubans take at least 1.4v-1.45v to run at 3.8GHz+
This. In my experience, when an Overclock stabilizes with more vcore, it generally needs more juice. In my opinion, starting from the ground up is a bad route to yo take. Raise your vcore to as high as acceptable for your cooling, run it stock and see the temps. Then up your multi and bus speed as you see fit (i think the phenom 2 responds more to bus speed increase as apposed to multiplier) until your Max temps and that will be your best Overclock.
 
Aside from what have been suggested . . . you might get ideas starting with Post #6 . . .

http://www.overclock.net/t/1229458/hi-960t-d

I disable Turbo on mine as well as C1E, and other power saving features in bios. Make sure your load temp stays below 50C if possible.
 
Discussion starter · #7 ·
Current CPU Voltage is at stock in the BIOS, CPU-Z is showing it as between 1.368V - 1.500 voltage, as I was putting the CPU under load.

I cannot unlock this CPU because this motherboard does not support, I sadly did not find that out until after the fact, turns out Asus took that ability out in the R2.0 version, though the original still has it.

I was hoping to be able to keep Cool N Quiet and Turbo, though I could prob live without Turbo as it never really works with the Phenom II lineup anyway.

Cpu cooling is a Zalman 9500, as I write this, at 3.6 Ghz, its about 31C

Also this system is in my signature as "Phoenix"
 
Quote:
Originally Posted by Thedarklord View Post

Current CPU Voltage is at stock in the BIOS, CPU-Z is showing it as between 1.368V - 1.500 voltage, as I was putting the CPU under load.

I cannot unlock this CPU because this motherboard does not support, I sadly did not find that out until after the fact, turns out Asus took that ability out in the R2.0 version, though the original still has it.

I was hoping to be able to keep Cool N Quiet and Turbo, though I could prob live without Turbo as it never really works with the Phenom II lineup anyway.

Cpu cooling is a Zalman 9500, as I write this, at 3.6 Ghz, its about 31C

Also this system is in my signature as "Phoenix"
31 at idle is actually good. you want below 35, so it won't go above 55 at load from my experience. can't run Prime if i oc and the idle is higher than 35. Turbo messes up the oc when it kicks in and the voltage is not enuf. i think at stock of 3.0GHz it turbos to 3.4GHz. 3.6GHz is a decent oc. I keep mine at 3.8GHz with a CM 212 cpu cooler and it handles a HD 7950 fine. I unlock to Six core for games like C3 or BF4.

Keep your HT close to 2000MHz and the NB to around 2600MHz or higher.

EDIT: with the 480, i don't think you need a six core.

Right now i got it at 4GHz actually . . .



that's my vcore for six core. i just leave it even if i lock it to quad. Check out the HT link and the NB freq.

Check out my temp with only a 212 but my ambient is pretty low . . .



23C using MX-2.
 
Quote:
Originally Posted by rdr09 View Post

31 at idle is actually good. you want below 35, so it won't go above 55 at load from my experience.
Idle temperatures have absolutely nothing to do with temperatures under load.
When I used stock cooler my idle temperatures was actually lower than using a 212 evo but yet under full load it was running much cooler.
 
There's been no mention of which memory kit/setup your running (2x8GB or 4x4GB?). Can make a huge difference!
Current Dram setup, CPU/NB voltage and Dram voltage?

Freezing is usually a sign of a lack of Dram voltage (check first) than go for additional CPU/NB voltage. Many times it's a combination of the two.
Running all four dimms is almost a guaranteed increase to CPU/NB voltage.

It's not uncommon to need an additional .05v to .1v Dram voltage (above specs on the sticker on the Dram itself) on the AMD side as the factory "sticker" spec is rated for Intel.
As you raise the CPU freq this usually also goes up again after a point/s.

You mentioned that setup "was" rock stable before increasing Dram freq. How was stability tested, which test and how long? Prime95, IBT AVX?
 
Quote:
Originally Posted by boxman222 View Post

Idle temperatures have absolutely nothing to do with temperatures under load.
When I used stock cooler my idle temperatures was actually lower than using a 212 evo but yet under full load it was running much cooler.
i said based on my experience using a particular cooler - the 212. op's cooler might result to a much better outcome. But what i do suggest is keeping the cpu temp below 50.

i go by the cpu temp and not the core. unlocked the core temps can't be read anyways.
 
Discussion starter · #12 ·
To rdr09, I do have other video cards just kinda chilling, I have a GTX 480 in there now more to push old hardware.

Quote:
Originally Posted by The Sandman View Post

There's been no mention of which memory kit/setup your running (2x8GB or 4x4GB?). Can make a huge difference!
Current Dram setup, CPU/NB voltage and Dram voltage?

Freezing is usually a sign of a lack of Dram voltage (check first) than go for additional CPU/NB voltage. Many times it's a combination of the two.
Running all four dimms is almost a guaranteed increase to CPU/NB voltage.

It's not uncommon to need an additional .05v to .1v Dram voltage (above specs on the sticker on the Dram itself) on the AMD side as the factory "sticker" spec is rated for Intel.
As you raise the CPU freq this usually also goes up again after a point/s.

You mentioned that setup "was" rock stable before increasing Dram freq. How was stability tested, which test and how long? Prime95, IBT AVX?
The RAM setup is 4x 4GB Corsair DDR3 1600 Mhz 9-9-9-24, the default voltage for these are 1.5v, which is what everything is running at.

As far as stable, I consider stable to be, running various benchmarks (3DMark's mainly) and being able to play various games without any issues for hours on end. I usually also run Prime95 for about 4 hours if I really want to push it.

I plan to try to push the OC higher tonight/tomorrow, but this time I am thinking I will disable turbo at the very least. I don't think I would need to increase the voltage on the RAM itself since its not being overclocked. Now I have thought about increasing the voltage to the uncore (CPU-NB) which may help stabilize the memory controller if that is the issue.
 
op, to give you an example, here are screenshots of my bios using one of my 4GHz settings (i have like 2 or 3 with different values) . . .





Just to give you an idea for a 4GHz setting. My rams (1600 MHz) use 1.65v as default. Take note of the vcore for the oc'ed CPU-NB. The Turbo is disabled along with C1E, which is not in the screenies.

Also take note of the LLCs (Load Line Calibration), which are enabled. My mobo is an Asus, so it should be similar to yours.
 
Discussion starter · #14 ·
Thanks everyone thus far for the help on this, really appreciate it.

Now to the update;

So I got the CPU past the previous limit of 3.6 to 3.7, to do so I did the following;

1) Disabled turbo mode
2) Increased voltage offset to +v0.10000
3) Increased CPU Load Line Calibration from Auto to Extreme

This made it stableish, it stopped blue screening and crashing, and it made it through all 3DMarks except for the latest 3DMark version, where it stops because "a unknown error as occured", also it looks like my Prime95 runs stop after about 20-25min because of a "rounding error".

So then I;

1) Increased voltage offset to +0.11250
2) Increased CPU/NB voltage offset to +0.01250
3) Increased CPU/NB Load Line Calibration from Auto to Extreme

Afterwards, I did not run 3DMark, I jumped straight to Prime95, which ran without issues for an hour, before I left it and went to sleep to let it run overnight. This morning however it stopped one of the worker threads because "a rounding error" has occurred.

I believe I am on the right track to resolving the Prime95 rounding errors, increasing voltages slowly?

Thoughts?
 
Quote:
Originally Posted by Thedarklord View Post

Thanks everyone thus far for the help on this, really appreciate it.

Now to the update;

So I got the CPU past the previous limit of 3.6 to 3.7, to do so I did the following;

1) Disabled turbo mode
2) Increased voltage offset to +v0.10000
3) Increased CPU Load Line Calibration from Auto to Extreme

This made it stableish, it stopped blue screening and crashing, and it made it through all 3DMarks except for the latest 3DMark version, where it stops because "a unknown error as occured", also it looks like my Prime95 runs stop after about 20-25min because of a "rounding error".

So then I;

1) Increased voltage offset to +0.11250
2) Increased CPU/NB voltage offset to +0.01250
3) Increased CPU/NB Load Line Calibration from Auto to Extreme

Afterwards, I did not run 3DMark, I jumped straight to Prime95, which ran without issues for an hour, before I left it and went to sleep to let it run overnight. This morning however it stopped one of the worker threads because "a rounding error" has occurred.

I believe I am on the right track to resolving the Prime95 rounding errors, increasing voltages slowly?

Thoughts?
IMHO yes, you are on track. Small increases and retest.

Be sure to keep good written notes of any changes/increases made and include the time it took to failure.
You can use the failure time to help determine whether or not the latest changes made an improvement.

Over time you'll see patterns develop. Results increase to a point than start going backward meaning you may need another adjustment to a different area before you can proceed.

Extreme LLC setting (if it's like my CHV-Z highest setting) is usually for extreme cooling.
Most run "very high" (1 from top setting) to "high" for those that work better with a little Vdroop.

Which version of P95 are you running?
Version 27.9 works real well on the Phenom II.
 
Discussion starter · #16 ·
Quote:
Originally Posted by The Sandman View Post

IMHO yes, you are on track. Small increases and retest.

Be sure to keep good written notes of any changes/increases made and include the time it took to failure.
You can use the failure time to help determine whether or not the latest changes made an improvement.

Over time you'll see patterns develop. Results increase to a point than start going backward meaning you may need another adjustment to a different area before you can proceed.

Extreme LLC setting (if it's like my CHV-Z highest setting) is usually for extreme cooling.
Most run "very high" (1 from top setting) to "high" for those that work better with a little Vdroop.

Which version of P95 are you running?
Version 27.9 works real well on the Phenom II.
I am running Prime95 v27.9, I haven't upgraded my local server copy in a while, which I should probably do as well.

I will try dropping the LLC setting back to High, and see if its still stable there.

I think the issue that was holding back my overclock was that I was trying to keep turbo mode enabled, and no matter how much extra voltage offset I gave it, I couldn't get it stable. As soon as the CPU would try to turbo it would lock up.

I have it running at 3.7Ghz still with turbo disabled and CPU cool and quiet enabled, using CPU voltage offset for the overclocking (I really don't want to lose power saving features).

Next thing I want to do is upgrade P95 version, and up clock to 3.8Ghz.
 
Quote:
Originally Posted by Thedarklord View Post

I will try dropping the LLC setting back to High, and see if its still stable there.

I think the issue that was holding back my overclock was that I was trying to keep turbo mode enabled, and no matter how much extra voltage offset I gave it, I couldn't get it stable. As soon as the CPU would try to turbo it would lock up.

I have it running at 3.7Ghz still with turbo disabled and CPU cool and quiet enabled, using CPU voltage offset for the overclocking (I really don't want to lose power saving features).

Next thing I want to do is upgrade P95 version, and up clock to 3.8Ghz.
You'll have to play a little with CPU LLC and CPU/NB LLC one at a time. I was thinking if Extreme is the highest setting try one setting lower.
If you lower LLC too much at one time you very well might have to just bump Vcore a bit. Shoot to keep Vcore fairly flat and/or just slightly higher than desired value when under load.

Turbo and OCing just don't play well together as you can see. Definitely should be disabled.
As for C&Q I run it on my setup at 5117GHz (during summer months) but only after I stabilize my OC.
When testing always disable all power saving features. Once you're stable than retest with C&Q enabled if you like.

As for upgrading P95 I'd stay with 27.9 for a Phenom II.
 
1 - 17 of 17 Posts