Overclocked FX-8150

[Vexillarius]

Hello everybody!

I'm a first time overclocker. I just spent a good two days studying and tweaking attempting to overclock my FX-8150.
After a whole let of experimenting I decided to just stick to CPU multiplier and voltage adjustments as I simply could not get anything else stable, and hey, it's my first time overclocking so let's not make things too difficult.
I have:
GA-990FXA-UD7 rev 1.1
FX-8150
2x Geforce GTX 660
2x 4GB Corsair ValueSelect 1333Mhz RAM
Corsair H60
Corsair GS800 PSU
In a HAF X case

I managed to get the CPU multiplier up to 22, so it's running at 4,4Ghz right now. CPU voltage is at 1,4875V. I turned off Turbo Boost and al the power management settings and put Load Line Control (as far as I can tell this is just Load Line Calibration) up to its highest setting. Stresstesting (Prime95), CPU temperature gets up to about 65 degrees Celsius, which, since I don't render/edit videos or anything I'm okay with.
Raising the multiplier further seems to require a ridiculously high voltage and since my cooling isn't that great I decided to stop here.
Messing with anything else just seems to result in crashes or errors as soon as I put any king of load on the CPU. Right now it seems stable!

Quick question, should I turn the power management settings back on now that I got it stable? All I do with this PC is some gaming and the occasional benchmark.
And what do you guys think? Respectable numbers for a first time overclocker?
Any tips are also much appreciated!

 

[Mandrake4565]

Vexillarius, welcome to the forum. A picture is worth a thousand words and will help us take a peek under the hood so to speak. When overclocking most of us only consider a rig stable if we can run 2 hours of prime blend without a failure, just an FYI. Anyhow, if you could download Cpu-Z and Hwmonitor "non pro" and do the following we can see if anything stands out to us. Open Hwmonitor and run prime blend 20 minutes, then post individual screen shots of Hwmonitor and the following tabs in Cpu-Z, Cpu, Spd and Memory.

As far as the power management you can turn it down if it is stable, I run my rig with the power saving features on so when I'm just browsing it will clock down the Cpu and voltage.

Here is how to host the screen shots:
View attachment 145462

 

[Vexillarius]

Thanks for the warm welcome!
Its quite late over here right now (I live in the Netherlands) so I'll do all that tomorrow and I'll get back to you. After that I'll try turning the power management options back on and see if it's still stable.

 

[caddi daddi]

manny is spot on, we nedd to see all those settings.

 

[Vexillarius]

Well I guess it's not as stable as I thought. Small FFT testing works perfectly fine, but blend testing crapped out one of the cores after about 15 minutes. No errors, 100 warnings.
Forgot to take screenshots, so I tried again. 1 error after 5 minutes.

 

[Johan45]

Wowzers your V_Core is hitting up to 1.584v. That's way too much for an H60 to look after. You must have a lot of things left on auto. If you haven't already you really should give this guide a read through http://www.overclockers.com/forums/showthread.php?t=688663. This will go through the procedure for a good OC on an FX CPU

 

[Vexillarius]

It gets even higher in small FFTs testing. No idea why, voltage just seems to be all over the place.

I read that guide but when I get to step #4 I just can't get my system remotely stable. It'll boot just fine, but any load on the CPU just freezes or crashes my system and the guide tells me nothing about what to do when something doesn't work. My system only became somewhat more stable when I put the CPU voltage up to about 1.45v.

 

[Johan45]

My first guess would be you need to set your CPU LLC, I'm not too familiar with Giga boards so set it somewhere in the middle. That should stabilize your voltage and yes at that point you may need to up your V_Core since your board isn't going to compensate any longer.

 

[Vexillarius]

I put LLC to 'High' which confusingly is its middle setting
Boots up fine, as soon as I start Prime95 I get errors in multiple cores and my system freezes within a second.
I'm including photos I just took of my BIOS settings. I'm a little scared to turn up the voltage further since it's already at 1.4875v and I'm starting to think I'm just completely missing something here.

EDIT: It also freezes up if I don't run a stresstest, it just takes half a minute longer...

EDIT2: Well I put the voltage up one notch, still the same result. HWMonitor reports a Vcore of only 1.264v even though I set it in BIOS to 1.5125v. I assume this is due to droop, but which value should I go by? 1.5125v which I think is quite high, or the 1.264v that the software reports? If it's the latter I guess I still have a lot of space voltage wise.

 

[Mandrake4565]

You probably want to dial back the Load Line Calibration so it doesn't overshoot the set voltage by too much and also doesn't droop below the set voltage. That said you may need to dial back the overclock because even with the LLC pushing the voltage to 1.58 it's not stable. Try keeping the voltage set where it is drop the LLC back 1 and set it to 4.2 then run prime. Watch the temps 70c on the socket and 62 on the package are recommended max temps. Also watch the Voltage we're trying to get it stable near the set voltage. If iy's not stable at 4.2 drop it down again I'd like to get it stable with the set voltage and then try pushing it higher.

 

[Vexillarius]

I dialed it back to 4.2Ghz and put the voltage at 1.4875v and LLC at its middle setting.
First thing I noticed is that HWMonitor actually reported 1.48v this time. Also, idle temps are suddenly much higher (from low twenties into 30-40 degrees Celsius).
Then, about 15 seconds into running Prime95 the voltage shot up for a fraction of a second to more than 3v! The system immediately froze.
At least it doesn't seem to be fried... I'll count my blessings and try 4.0Ghz next I guess.

 

[caddi daddi]

I see tmpin1 caddi style, dialing the clock back sounds like a very good idea.

 

[Vexillarius]

4.0Ghz at 1.4875v seems stable so far, at least much more stable than previous attempts. 5 minutes of blend test and no problems, now trying small FFts for a while to see how hot it gets. So far temps seem fine, around 44 on the cores and 55-60 on the socket.

On a somewhat related note: since I'm obviously going to need better cooling (it's been on my wishlist for a while now) which cooler would you guys recommend? I have my eye on the Swiftech H220, the Corsair H100i and the Cooler Master Nepton 280L. Any other recommendations are also appreciated. As long as it's a complete package, I'm not going to try to build a custom cooler or anything.

 

[Mandrake4565]

4.0Ghz at 1.4875v seems stable so far, at least much more stable than previous attempts. 5 minutes of blend test and no problems, now trying small FFts for a while to see how hot it gets. So far temps seem fine, around 44 on the cores and 55-60 on the socket.

Idle temps don't really mean much but, they will be increased because you raised the set voltage from 1.3xxx to 1.4xxx

First off I doubt the 3v was a correct reading, it was probably just a errant reading that was caused by the system crashing. 3v would have killed the chip in an instant!

As far as cooling goes I have never used any of those coolers so I can't be of any help on them. My 8350 has been under a full custom loop since day 1.

 

[RGone]

On a somewhat related note: since I'm obviously going to need better cooling (it's been on my wishlist for a while now) which cooler would you guys recommend? I have my eye on the Swiftech H220, the Corsair H100i and the Cooler Master Nepton 280L. Any other recommendations are also appreciated. As long as it's a complete package, I'm not going to try to build a custom cooler or anything.

Vexillarius, we have had a problem of sorts in that we have TOO often had to operate as what seems a Mobo Police around here with far too too many brands and model of motherboards. Some so cheap they did n0t have nearly enough VRM circuitry to handle an overclock of 6 and 8 core FX processors and then many from Gigabyte that just flat acted like crap with huge Vdroop and when users across the net talked with Giga Tech Support, all they were told is that the board was following the specs set forth by AMD. This has left too many Giga AM3+ board users in a lurch for sure.

I have spent nearly 2.5 hours reading a huge 1000 plus page thread at OCN where they start and continue long arse threads on specific brands/models of mother boards. Here is a small snippet about the UD7 board and probably most appropriate for your own situtaion.

COPIED:
4/17/12 at 2:17pm
The UD7 from what I've sen can be a "finicky" board. The LLC seems to not be consistent between boards and furthermore processors used in the boards. So when exploring your OC with the UD7, I would suggest finding the individual limits of each system first. By this I mean, run everything at stock speed and increase one thing at a time. (I would also take thefallen's advice and disable CnQ, Turbo and C1e while finding said limits.
END COPY.

http://www.overclock.net/t/1023100/official-gigabyte-ga-990fxa-series-owners-thread-club
[Official] Gigabyte GA-990FXA-Series Owners Thread/Club << My browser shows that thread to be 1002 pages long.

What I copied above and pasted was from around page 390-ish. Circumstances were not actually much better or worse from earlier posts to later posts. When they got to posting about Rev 3.0 of the various models of AM3+ the Rev 3.0 boards took people far into hair-pulling and that we have seen here in the OCF forums for sure.

There is a pretty long thread about the UD3/UD5/UD7 boards at Xstreme systems and again all over the map and hair pulling. I uncovered a hard-mod that helps stop the Vdroop on the UD boards but it is different for each model be it UD/3/5/7 so am not going to do other than mention it here since it IS hard-mod and requires changing out a resistor.

At this point I am going to just shoot straight from the hip after nearly 2.5 years of dealing with users and FX processors and motherboards >> Get a Sabertooth 990FX R2.0 and get some reliable adjustability in your system. THEN and only then worry about more cooling since cooling is not going to fix your ability to accurately clock that Rev 1.1 UD7 motherboard. The crux of the matter is that the variance from board to board and then cpu to cpu is all over the map and not fun for someone who wishes to become a "user" of an overclocked FX processor system in daily 24/7 usage.
RGone...

EDIT:
By the way that problem you have in following the Tutorial linked at the top of the forum section was done with a UD7 board and your inability to follow it is indicative of how the boards are all over the map setting and ability-wise.
END EDIT.

 

[Vexillarius]

Yeah I'm noticing some ridiculously bad vDroop on this board - bad as in 0.25v bad.
I ordered a H220 anyway, it won't hurt and I'll probably be using it for a long time, past this motherboards lifespan/usefulnesstomespan. I'll see how far I can overclock, but it won't be particularly high. I'll tweak the voltage, multiplier and LLC setting and keep it at that. Definitely not going to town on my board with a soldering iron!

Long story short, the problem I'm having seems to be ridiculous vDroop and there doesn't seem to be much I can do about it.

Thanks for all the help guys!

EDIT: Actually there's one thing I don't get yet. Isn't vDroop supposed to lower the voltage under load? HWMonitor reports a 0.25v loss regardless of load - it's that low even when idling.
Also, setting LLC to auto seems to improve things quite a bit, voltage loss is reduced to about 0.15v, still regardless of load.

 

[Mandrake4565]

The LLC on that board seems to be a big issue, in it's inability to keep the set voltage. Finding the best setting where it doesn't droop and overshoot by too much will be the key.

 

[RGone]

Vdrop is voltage "drop" under little load.

Vdr00p is voltage "dr00p" under load.

So there are two types of voltage "lowering/fall-off" some boards have it bad on both on drop and droop.

Truthfully on a motherboard with g00d voltage regulation it would stand to reason to have the voltage to the cpu "go" lower under next to no load or when just about doing no work. Heck that makes real sense to me.

I, as have many, dealt with Vdroop for years when we ran Intel rigs. It is/was/and will be an Intel spec and now from FX processors also an AMD spec.

I know the author of this article and he is that knowledgeable. I hesitated to post the link since it is very detailed in description and without a little EE experience it gets too deep for many. Suffice it to say Vdroop has a real value, but it can also be very off-putting to many and if the boards are not very good in their power regulation to the cpu, then overclocking can be a real pain. One minute not enough voltage/power for stability and the next way too much voltage/power that causes increased heat and even in some cases the quick voltage rise overstresses other circuits and the board thinks it needs to throttle back for safety sake.

http://www.thetechrepository.com/showthread.php?t=126
Vdroop Explained

After a period of time many years ago; users were up in arms about Vdroop and many and now most motherboard makers developed CPU_LLC to try and compensate for the cpu Vdroop specs. But if the board does not have effective circuitry to handle the demands of CPU_LLC then the whole overclocking experience tends to suffer.

With a good board and well working CPU_LLC you can do as I do on my first edition Asus CHV mobo and have less than 1.0Vcore at idle and also have approximately 1400Mhz but when loaded jump to 1.48Vcore and 4800Mhz cpu speed that is stable for over 2 hours Prime95 Blend mode at the 4.8Ghz and this on air cooling. First the power supply for the system and then the motherboard as foundations for computer tweaking fun.

 

[Vexillarius]

Okay, I officially have no idea what's going on anymore. I can raise vCore as much as I want, it doesn't go above 1.36v according to HWMonitor and CPU-Z. Going higher than 1.36v doesn't seem to increase stability either. The offset between the BIOS value for vCore and the value HWMonitor gives me also becomes much smaller, almost negligible, when I put vCore under 1.36v.

So except for the odd fraction-of-a-second spiked in voltage, vCore completely refuses to go above 1.36v, if HWMonitor is to be believed (and I'm inclined to do so, seeing as how it doesn't seem to make a difference for stability either).

 

[Mandrake4565]

What BIOS are you using?