AMD FX 8320 throttling on a low temperature and fluctuating under load

Toggle sidebar Toggle sidebar
W

WerneHR

Honorable
Nov 15, 2013
17
0
10,520
Well, I have an overclocked AMD FX 8320 (4.72GHz, 1.428V) on a Gigabyte GA-970A-UD3 Revision 3 motherboard, and I'm experiencing CPU throttling even though the CPU temperature in CoreTemp never exceeds 47C during a 5h Prime95 run (it jumps between 46C and 47C).

I also have a problem where the CPU frequency fluctuates under load, multiplier drops down from 23.5 to 14.5 on random cores during the Prime95 run. Doesn't happen when HPC mode is active, but that thing gets on my nerves since Prime tends to fail for no good reason, same happens if I leave SpeedFan open while running Prime95, one core shuts down within a minute even though I can stress the CPU with Prime95 for more than 12h if I use CoreTemp instead of SpeedFan. Stupid software...

I've also tried benchmarking it on Debian Wheezy using mprime (Prime95 for Linux) and the CPU frequency doesn't fluctuate at all, it stays stable at 4.72GHz until it drops to x7 (throttles), and the temperature never exceeds 46.4C.

Full hardware specifications are:

CPU: AMD FX 8320 eight-core processor (4.72GHz, 1.428V, all cores enabled)
Mobo: Gigabyte GA-970A-UD3
RAM: 4x2GB DDR3-1333 Transcend JetRAM CL9
Graphics: XFX AMD Radeon HD 7770 Core Edition
PSU: XFX Core Edition 550 Pro
HDD: 1TB Hitachi Deskstar HDS721010CLA330
SSD: 120GB Kingston V300 SSDNow SV300S37A120G

Cooling, I'd say I have a well ventilated case. it's an ATX midi tower with 3x92mm 2000RPM fans on the front that push air in, 2x92mm fans on the back and 1x92mm on top pushing air out (2000RPM each), 1x60mm 3800RPM fan for Nortbridge and 1x60mm 4400RPM fan for VRM cooling. CPU cooler is an Arctic Cooling Freezer 13 with disabled PWM that spins at 2300RPM (I didn't have money for better, though this one is apparently doing a good job too).

All power-saving features (C6, Cool & Quiet, C1E) are disabled, even some virtual machine thing as well since I don't use VMs, I dual-boot instead.

I honestly have no idea why this happens, all should theoretically be good and the CPU is running 15-16C below 62C max on the same voltage it uses for Turbo. So, does anyone know what the problem might be? I'm kinda running out of options to try here.:(
 
Sort by date Sort by votes
N

nihilispc

Honorable
Nov 16, 2013
2
0
10,510


 
Upvote 0 Downvote
N

nihilispc

Honorable
Nov 16, 2013
2
0
10,510
matter 1 : for instability, its probably because of low vcore. try to increase it. But before doing that check matter 2
matter 2 : VRMs on your mainboard getting too hot. Even tho your cpu temps are fine, VRMs can also cause cpu to throttle.

if you increase vocre to help with matter 1, its gunna hurt matter 2 even more. Before increasing vcore, check your VRM heatsinks if they are really hot with your current OC setup.


As a result, you might have to lower clock and vcore to solve VRM problem and instability during tests.
 
Upvote 0 Downvote
W

WerneHR

Honorable
Nov 15, 2013
17
0
10,520
@nihilispc VRMs are only warm on touch, not hot. I'd say some 45C, and that's after a few hours of running Prime. As for VCore, wouldn't low VCore shut down cores during Prime due to wrong calculations? At least I think that's what should happen...

@WaSquids I've searched through the thread, unfortunately, no mention of a solution.

Ah well, I'll try enabling HPC mode along with power-saving stuff and see where that takes me, it's the last thing on my list of "what my or may not work".
 
Upvote 0 Downvote
W

WerneHR

Honorable
Nov 15, 2013
17
0
10,520
@WaSquids Already have a fan blowing on VRMs, an AMD stock CPU fan from a socket 939 cooler, installed it the moment I got my components and it reports cca 4400RPM at the moment. I also replaced the thermal paste for all heatsinks on my components with Arctic Cooling MX-4 the second I removed them from the case, I had bad experience with stock paste on those things (I still have a dead ASUS A8N-SLI Deluxe around with burned-out VRMs on stock CPU voltage/frequency).

And it seems that my last attempt is getting me somewhere. I enabled C6, Cool & Quiet, C1E and HPC Mode, running P95 for about 40 minutes now, frequency doesn't fluctuate at all like it did. But since my temps got raised a bit because it now works at 4.72GHz and 1.428V constantly (as opposed to x23.5-14.5 and fluctuating voltage), the CPU got warmer, reaching 51C at which it drops to x7 a bit to cool down. It can work constantly at 50C (Test 2, pass 1 for example, it got the CPU to 50C constant and no trottling) but as soon as it jumps to 51C - bam, x7 multiplier.

And that behavior got me thinking, shouldn't the CPU throttle at 60C instead of 50C? My VRMs and NB are pretty cool no matter what I do, so is it possible that my CPU sensor reports an incorrect temperature? That might be why I'm having issues, though 10C off doesn't seem like a small mistake to me.
 
Upvote 0 Downvote
W

WerneHR

Honorable
Nov 15, 2013
17
0
10,520
@WaSquids Hmm, solution in that thread only made it worse. Enabling CPB only results in even more fluctuating and in different frequencies too (x7, x14.5, x18.5, etc) across random cores. Not to mention it overvolts the CPU, I've seen the voltage hitting 1.478V and higher while all cores were kept at 3.7GHz or lower. Only good thing is that mobo doesn't shut down and start again upon reboot/power on, but that doesn't bother me anyway. Maybe the overvoltage has something to do with me using the Extreme LLC but the Regular overvolts like crazy on idle instead of load (over 1.5V on idle, 1.428V under load with Regular).

And seems I did manage to get the CPU to partially behave, I found it when following random links from the threads you linked. I enabled all power-saving features and HPC mode, voltage is the only thing that fluctuates based on load (1.416V-1.440V) and the CPU is completely stable, passed an 8h Prime95 test without a problem. I still get thermal throttling once Core 0 reaches 51C but other than that, it runs like a charm.

There's also another way to bypass the fluctuating issue - BCLK overclocking. I tried setting the multi to x20 and only raising BCLK (keeping HT and NB multipliers down of course). Seems that's the most effective way, with power-saving features on, HPC off, multi at x20 and BCLK set to 235, it experiences no fluctuating. A dedicated NB fan along with a decent thermal compound comes in handy then, to reach 235 BCLK I had to raise NB voltage by +0.125V but my fan did a great job keeping it cool. CPU experiences thermal throttling at 51C even then, I can't seem to understand why 51C if CPU can handle up to 62C, unless my sensor is off or the board is messing with me.

So I solved the fluctuating under load but thermal throttling is still there even though there shouldn't be any. Maybe a newer BIOS would solve that, I'll go flash the thing and see if that helps. If not, I'll just reset it through backup BIOS, that's why I like Gigabyte's dual-bios design (I hate the CPU voltage control though).
 
Upvote 0 Downvote
WaSquids

WaSquids

Honorable
Oct 23, 2013
175
0
10,710
Dam man, complicated :S i hope you managed to get it stable. Just read bout many problems with some of giga boards(im not an oc expert) just trying to chime in with any help if i could (was just looking you have 8+2 phases aswell... should be good:S) Let us know if you find a fix. Hope those links kinda helped :s.
 
Upvote 0 Downvote
W

WerneHR

Honorable
Nov 15, 2013
17
0
10,520
@WaSquid The links did help, one of them lead me to believe the temperature report is incorrect so I found the reason for the thermal throttling, the temperature sensor is indeed off. I ripped the mobo out of the PC and tested it on the table, the laser thermometer reported 58C on the cooler base while the Core 0 temperature sensor reported 51C on the die at the same time. It's impossible for the cooler base to be hotter than the CPU, which means the temperature sensor is off by cca 10C and the actual temperature on the sensor is ~60-62C, which explains the throttling.

So basically, my CPU is lying to me and I have to add 10C when looking at temps. Also need a better cooler. Bummer, I was hoping this thing will hold when I saw rather low temps, but it seems the CPU cooler is indeed inadequate. Ah well, back to 4.2GHz and stock voltage until I can afford to buy a better one, best air cooling I can get around here is Scythe Mugen 4 PCGH which should keep it adequately cooled, water is too expensive for me and I have a bad feeling keeping water around electronics.

Anyway, thanks for the help man, I wouldn't have though about measuring temps on the cooler base if it weren't for that one thread I've read while going through one of those links.
 
Upvote 0 Downvote
WaSquids

WaSquids

Honorable
Oct 23, 2013
175
0
10,710
Damn...Although at least you managed to single it down to the heat. No worries, glad it could help (from my random reading&bookmarking of amd mobo's :p)

I'm not sure if this would help for a cooler (price/availability etc...) But this Noctua looked interesting for the lower middle fan (poss help the vrms too?) You prob know of it, effective, but pricy.(It's huuuuge though :p)

http://www.amazon.co.uk/Noctua-D14-2011-Dual-Radiator-Cooler/dp/B0062K6BIC/ref=sr_1_2?s=computers&ie=UTF8&qid=1384793585&sr=1-2&keywords=noctua+d14

Anyways glad i could help as best i could :)
 
Upvote 0 Downvote
G

ggggg1

Honorable
Dec 29, 2013
1
0
10,510



AMD FX CPU Throttling issue may appear if MB north bridge is overheated (use of liquid CPU cooler) - try to add fan on MB North Bridge.

 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom