What is holding my overclocking back???

[Niteshade]

Hello
I'm an average overclocker (no newbie, but not really hardcore as many others here) from Sweden in need of some help with my current system.
My problem is that I don't seem to be able to push my system further and I can't really tell what is holding me back: is it my PSU, the motherboard or bad chips? (GPU/CPU)

Current system is:
AMD Phenom II X6 1090t OC @ 3.9GHz with Noctua NH-U9B SE2 -air cooler
Gigabyte GA-970A-D3
Corsair XMS3 Vengeance 8GB 1600MHz
Asus HD6970 2GB DirectCU II OC 930MHz clock/1425MHz mem - OCZ ModXStream 700W
My chassis has 4x 120mm fans (2 out/2in) and 2x 140mm fans (1 out/1 in) so my airflow is pretty good.
I also have fan controller which tells me the temps in various areas of my chassis and I'm pretty happy with how it is right now.

The CPU overclock:
Currently I'm running at a 19,5x multiplier with my reference clock set at stock speed of 200MHz, which gives me a little over 3.9 GHz (stock speed is 3.2GHz). My core voltage is set to 1.5 volts - default value there is 1.325 (I think, can't remember).
My CPU/NB is on a 14x multiplier giving me 2800MHz. Stock there is 2000MHz @ 10x multi.
CPU/NB voltage is at 1.25v
HT is at factory settings (10x multi which equals 2000MHz)

System is stable with the above setup.

CPU OC problem:
I don't seem to be able to even touch 4GHz - even though I know that this is a realistic goals (I've seen PLENTY of overclocks reaching 4GHz at or around 1.4v). I've tried a 20x multiplier with more voltage but that won't do the trick, and I don't want to go over 1.55 volts as AMD recommends no more than 1.55v in vcore. I've also tried combining a higher reference clock speed with a lower multiplier - but that haven't worked either. 4GHz is simply not possible @ 1.55 volts despite how I overclock (refclock, multiplier or both).
My CPU/NB also doesn't seem to be able to climb higher than 2900 (gives instability near 2950) despite increases on the voltage.
Temperature is not holding me back at all - at least not on the CPU. However, I do not have any custom cooler on the NB chip itself on the motherboard.

The gcard overclock:
Stock voltage: 1.175 volts, upped to .025 to 1.2v
Factory OC is 890MHz core and 1375MHz memory, which I've succesfully OC'd (stable) to 930MHz/1425MHz.

Gcard OC problem: I can't up the core any more. I can reach 950MHz and pass my benchmarks/stress tests - but once I play Battlefield 3 at that speeds and my entire system (CPU, GPU and memories) gets workload, my display driver crashes along with the game and I have to cut back to 930.
The same goes for the VRAM. I can up it independently to 1500MHz and pass my benchmarks/stress tests - but once it comes to actual gameplay (BF3) the display driver crashes and drags the game along with it.

My thoughts:
I started by OC:ing the CPU by first finding out my max values independently (keeping everything at stock and just OC:ing one part of the system at a time). I got frustrated hearing about how people could reach 4GHz with worse coolers than I have, and with much lower voltages (best chips seem to be able to do this at 1.375v - most at 1.4 to 1.45). At this time I thought that I simply had a bad chip and I just shrugged it off and moved on to OC other parts of my system.

When OCing the graphics card I had a bit of a problem because some overclocks worked fine in Furmark while failing in 3d mark 11. I could push my graphics to I think it was 975MHz core and still pass Furmark but at some not-so-great GPU temperatures of around 85 degrees C. So yeah, at the start I played around a lot at 960-975 - even trying to push it to 1GHz. But with time I cut down more and more as I saw that not even an overclock of 60MHz (950MHz) would pass in 3d mark 11 or they would crash my display driver when playing BF3.

Now after OCing the CPU, CPU/NB, GPU and graphics memory, I'm a bit confused because I'm unsure of what exactly is holding me back. It could be my PSU? The gcard I have recommends a minimum of a 700W PSU according to the manual. Or it could be that my budget mobo just doesn't cut it for overclocking. Or it could be that I've been unlucky with both the CPU and GPU chips. But which one of these is the most probable cause do you think?

Another possibility is that my VRM is getting overheated. But I haven't really been able to find any info about that in regards to my mobo: I don't really know what a high temperature is, and don't even know which of the TMPIN0 to TMPIN2 -meters is the VRM in HWmonitor (as these meters/labels are different from mobo to mobo).

What is your thoughts on this?
Also, for clearifications: TEMPERATURES is not my problem. The only time I get alarming temps is when I drag the GPU voltage up too much and run Furmark (which is like an angry army drill sergeant on your GPU: nothing in the world will push it harder).

 

[I.M.O.G.]

Hmm, we can't really leave out temps - you are running a fairly high vcore for air cooling. What does coretemp or hwmonitor read at idle and load?

Also sometimes less is more, you may get further at 1.45V than you do with 1.5V+, depending on how well your cooler is performing.

 

[Niteshade]

At my current voltage, the vcore temp never rises above 58 degrees @ load - and that's when running Prime95. In games and 3D Mark 11 the temp is lower. And @ idle the temp is 30C.

Graphic card temp never goes above 62 C when playing games/running 3D Mark 11. Only Furmark and a higher voltage can get it higher.

It's a good point you have about "less is more" though. I've tried some lower voltages for 4 GHz but nothing below 1.4v really - so I'll check that out when I can.

 

[EarthDog]

58C load is too warm for AMD CPU's. Around 55C or so one can start to see some instability. So its best to keep it where you are at or lower. Since you are having issues. That is where I would start first.

 

[Niteshade]

58C load is too warm for AMD CPU's. Around 55C or so one can start to see some instability. So its best to keep it where you are at or lower. Since you are having issues. That is where I would start first.

No you are wrong about that.

"As long as your core temperature has not exceeded the high side of the 60 degree mark for extended periods of time you should be ok.
62 degrees holds a generous safety net to begin with." - Alex Cromwell, Senior Technology Director. Advanced Micro Devices (AMD)

source: http://forum.overclock3d.net/index.php?/topic/33173-1090t-temps/page__view__findpost__p__402292

And considered that I ONLY reach 58C if I stress my CPU ****loads with Prime95, I don't think the temperature is the problem here.
I've read some other forum posts about the 1090T max temps as well and they all say the same: 62C is max but not really a problem.

 

[EarthDog]

We are talking/I mentioned stability issues. He is talking about processor shutdown/damage.

 

[Niteshade]

I know, and I said that my CPU is stable at that temp
24 hours of Prime95 running all 6 cores at 100%, max temp 58C = stable.
So I really don't understand why an application (BF3) that doesn't stress my CPU anywhere near as much, would cause instability caused by a temperature that is much lower. I don't remember exactly what my CPU max temp is after 4-5 hours of BF3 play but I'm sure it's not above 50C.

I'm thinking it's more likely it's a PSU/motherboard problem because my display driver only seem to crash when both parts (CPU and graphics card) are stressed at the same time. I'm gonna read some more about the effects of an overheated VRM.

 

[EarthDog]

Almost sounds like a power issue if thats the case. Have you tried to run prime95 and loop vantage/3d11 at the same time? If its not power, SOMETHING is not stable in your overclock... try stock to see if the problem goes away.

It was also common to see P95 run and have something crap out on you. Any stability test isnt the end all for stability, though it is generally a great judge.

If you have tried the newest, including beta, GPU drivers then I would look towards power.

 

[Niteshade]

Almost sounds like a power issue if thats the case. Have you tried to run prime95 and loop vantage/3d11 at the same time? If its not power, SOMETHING is not stable in your overclock... try stock to see if the problem goes away.

It was also common to see P95 run and have something crap out on you. Any stability test isnt the end all for stability, though it is generally a great judge.

Yeah, power issue seems the most logical. But I really can't tell because my only experience from weak PSU's is immediate shutdowns (not even BSOD but power outs).

Nah I haven't tried that combination yet - good suggestion. I'll try that when I'm back home. I could both Prime95 and Furmark at the same time even, seeing that Furmark is even harder on the gcard than 3d11 is.

 

[Daleon]

It sounds like you've hit a wall with the cpu or mb. Additional NB cooling could possibly help if its getting pretty hot. Not like going to 4ghz vs 3.9 is going to be some huge difference. I'd say your pretty much already there at the high end of what most 1090t's OC to on air. Often its just luck of the draw in what your chip will do.

You don't saw what happens when you try and do 4ghz. Does it boot at all?
What is your memory at?

The ModXStream is an ok psu, not the most efficient and probably has some ripple issues when pushed so replacing it wouldn't be a terrible choice but not sure its gonna get you any higher.

May get some better suggestions from the amd cpu forum if you haven't tried there yet.

 

[Niteshade]

It sounds like you've hit a wall with the cpu or mb. Additional NB cooling could possibly help if its getting pretty hot. Not like going to 4ghz vs 3.9 is going to be some huge difference. I'd say your pretty much already there at the high end of what most 1090t's OC to on air. Often its just luck of the draw in what your chip will do.

You don't saw what happens when you try and do 4ghz. Does it boot at all?
What is your memory at?

The ModXStream is an ok psu, not the most efficient and probably has some ripple issues when pushed so replacing it wouldn't be a terrible choice but not sure its gonna get you any higher.

May get some better suggestions from the amd cpu forum if you haven't tried there yet.

Thanks for your reply Daleon.
Yea it could be as simple as that I guess. But I just can't stop thinking about pushing it further - especially after reading about other 4GHz+ setups on air
I can set my multiplier to 20x (4GHz) just fine. Booting works - haven't tried playing a game at that speed though. But once I start Prime95 I get an error after 3 minutes. My memory is set at 1600MHz (200MHz ref clock). I don't think it's a memory issue because afaik memory problems always result in BSOD - but maybe I'm wrong about that?

 

[EarthDog]

Lower your memory speed to see if that is the culprit. You are removing variables from your problem.

4Ghz(ish) were the limits of PhenomII really. I would imagine cooler temps would allow for better stability (regardless of what that link says - there has been anecdotal reports after anecdotal reports of instability of overclocks happening after 55C regardless of CPU temp limits in AMD land).

 

[Niteshade]

Lower your memory speed to see if that is the culprit. You are removing variables from your problem.

4Ghz(ish) were the limits of PhenomII really. I would imagine cooler temps would allow for better stability (regardless of what that link says - there has been anecdotal reports after anecdotal reports of instability of overclocks happening after 55C regardless of CPU temp limits in AMD land).

Yea I'll try that.
I'm also considering buying this one: http://www.enzotechnology.com/mst_88.htm
Costs about 20 USD in Sweden

Why?
Because my TMPIN2 monitors in HWmonitor goes to 90C when I run Prime95 at a high voltage/cpu freq. And considered that my mobo doesn't even have a heatsink on the MOSFET: http://content.etilize.com/900/1020379243.jpg
...it's very possible that this is the problem. So I looked this up and I found this interesting thread regarding 125W TDP processors and overclocking - and the problems that might occur with the VRM if you OC too much:
http://www.overclock.net/t/943109/about-vrms-mosfets-motherboard-safety-with-125w-tdp-processors

From the thread:
"Heat causes a lot of VRM problems including unstable power delivery and even fire hazard ... ideally the temperature should be the same as the CPU load"

...which in my case it's CLEARLY NOT.

As it is now, step 1 for me is to get a MOSFET cooler.

 

[EarthDog]

Assuming TMPIN2 are the VRM's thats a good move.

 

[Niteshade]

Assuming TMPIN2 are the VRM's thats a good move.

That thread to the VRM guide says that most Gigabyte mobos list the VRM as TMPIN2 in HWmonitor. I checked around a bit more and found the same information here:
http://www.overclock.net/t/1048912/official-gigabyte-ga-990xa-970a-series-owners-club-help-thread

So I'm pretty confident that is the case.
Damn I can't wait to see what kind of change that heatsink will do to my system How can one NOT like overclocking?

 

[EarthDog]

Dont expect miracles. Like I said 4Ghz = sweet spot for PhII. 1.5v is brave already I believe. I wouldnt push it more...but thats just me.

 

[Niteshade]

Dont expect miracles. Like I said 4Ghz = sweet spot for PhII. 1.5v is brave already I believe. I wouldnt push it more...but thats just me.

Yea I know what you are saying.
I just want to know if I can get to 4GHz or not, or if I can lower my current voltage @ 3.9 GHz. Because 1.525 for 3.9GHz isn't normal unless the chip is really crappy.

One thing I've been thinking about though:
Is it possible that the overclocking of the graphics card gets unstable due to overheating motherboard VRM's/MOSFET's?
I'm no hardware guru but I can't really see how that would effect my motherboard. But maybe those higher voltages "eat up" more wattage from the PSU? Like I said, I'm no hardware guru.

Because if it's unrelated then I really don't understand why I can't OC my card to at least 950MHz when I've heard of people doing just that with the stock voltage or just an +.025v increase.