Odd Underclock When Stress Testing

docov06

Reputable
Feb 21, 2014
44
0
4,530
Hey guys! Just recently bought a H100i watercooler (freakin' love it so far) and decided to do some overclocking on my AMD FX-8320. Before I bought the cooler, I had it OC'ed to 4.2 Ghz and had no problems whatsoever (except for temps). But now, I have a higher thermal threshold, so I decided to OC it gradually to 4.6 Ghz. It runs stable at this frequency, I've played several hours straight of BF4 and had no problems with framerate or BSOD. But when I stress test it with Prime95, CPU-Z and AMD Overdrive indicate that after about 5 min of testing, the multiplier drops from x20.0 to x7.0 for like a split second then goes back up. Also, what's weird about specifically dropping it to x7.0, is that CPU-Z shows my current multiplier and then it has "(7-20)" after it (shown in the screenshot). So what does that mean and why is my processor being underclocked? Ive also noticed that it only does this under Prime95. I tested it with CINEBENCH and had no issues whatsoever. Thanks!

TlYkJ1.jpg
 
Solution
The socket temps will be close enough to the vrm temps. I put my 80 mm fan literally half an inch above the VRMs. Its squeezed in between my hyper 212 evo and my back exhaust fan (a make shift solution with twisty ties lol). The closer the fan can get to the vrms the better but the side panel fan should help top, just potentially not enough (since its further away from the vrms but still does give some airflow, assuming your cases side intake sits across from the vrms nicely )

This pic is similar to my set up, except i have a 212evo not a liquid cooler.

http://cdn.overclock.net/8/8c/500x1000px-LL-8cd3a0fa_Picture193.jpeg
Disable cool n quiet while testing . prime 95 puts all 8 cores under 100% load. Bf4 doesn't even come close to the stress p95 causes. Could be vrm throttling..sometimes with liquid coolers he CPU gets better cooling but the vrms have no fan blowing on them (like air coolers often provide).

What motherboard and what power supply areu using ?
 


When the VRMs (voltage regulating modules) are giving power to the CPU and get too hot, they clock the CPU down to prevent heat damage to the motherboard. The power demands for the FX at that GHz is simply a lot for the VRMs to handle, them being at full capacity makes them hot.

I have the same motherboard. 6 power phases for the FX isnt so great. I have to add a small 80mm fan over the VRMs.

Hyper 212 EVO 4.3 Ghz with borderline temps in P95
 

docov06

Reputable
Feb 21, 2014
44
0
4,530
Got it. Is the VRM temp the same as the socket temp? And is there a way to turn that throttling off or should I just look into some better cooling options? Or maybe just turn the voltage down and hope the clock remains stable?
 
In the digi+ options of our motherboard is a setting for vrm with t.probe and extreme. Set it to extreme. It will get hotter though so be careful. Probably a good idea to look into active cooling fixed directly over the vrms or the back of the socket your case allows it
 

docov06

Reputable
Feb 21, 2014
44
0
4,530
Ok, I'll look into that! By any chance, do you know how to monitor the vrm temps? I use HWMonitor to see my Cpu core temps, but I'm not sure if it has a vrm monitor. And did you just mount that small 80mm fan to the inside of the side panel on the case?
 
The socket temps will be close enough to the vrm temps. I put my 80 mm fan literally half an inch above the VRMs. Its squeezed in between my hyper 212 evo and my back exhaust fan (a make shift solution with twisty ties lol). The closer the fan can get to the vrms the better but the side panel fan should help top, just potentially not enough (since its further away from the vrms but still does give some airflow, assuming your cases side intake sits across from the vrms nicely )

This pic is similar to my set up, except i have a 212evo not a liquid cooler.

http://cdn.overclock.net/8/8c/500x1000px-LL-8cd3a0fa_Picture193.jpeg
 
Solution

docov06

Reputable
Feb 21, 2014
44
0
4,530
Thank you SO MUCH. Just installed a 90mm fan over the VRM's with zip-ties and had NO throttling for the whole 30 min that I ran P95. My socket temps got up to 85C though, but that was that way ever since I OC'ed the CPU. Thanks again!! :D
 


85 is way too hot for a socket, sometimes i wonder if our board is just flat out misreading the socket temps. ive had weird anomalies where my actual CPU sits around 55 but the socket will keep climbing steadily with no end, i shut it down before it climbed past 78, but at the rate it was going it would keep "reading" until it hit the hundreds i bet. Ive always been confused with this board in that regard.

With my last case i had a spot for a fan on the side panel exactly behind the socket on the back of the motherboard, i had slipped a very thin 60-80mm fan behind there and turning it to full blast (with my fan controller i was using at the time) would keep the socket temp from climbing. With my new case (SPEC-01) i cant get a fan behind the socket, but i settled for the fan over the VRMs and that helps too.

glad to hear is working well for you :]
 

docov06

Reputable
Feb 21, 2014
44
0
4,530
Ya I thought there was something weird with my socket temp, because it does keep climbing after the core reaches peak at 63C. I actually have a leftover fan from my stock AMD cpu cooler that came with the processor, but I'm not sure if i can fit it behind the socket without it actually touching the socket (NZXT H440 case). There's alot of sound proofing foam behind the panel, so I could just remove that. Also, quick random question, do you use the small FFT test on P95 to stress test or blend?? Both of them max out the cpu usage, but the blend test pushes my core only to 54C rather than the 63C from the small FFT test (same amount of time for testing).
 


i do blend and small fft
 

TRENDING THREADS