How to Eliminate Throttle on OC'ed FX-8350?

clutchc

Titan
Ambassador
Using an Asrock 990FX Extreme6 motherboard http://www.asrock.com/mb/AMD/990FX%20Extreme6/

Even at a mild OC it throttles. Thermal Margin never drops into the single digits. So I don't think it's heat related.
I have increased the CPU multiplier to 23.5 and eased the core voltage up to 1.425 and it passes the stress test. I've even had it up to 4.8 GHz w/1.475V and it passes. So it is a stable OC, it just keeps throttling during the stress test.
I have disabled in BIOS:
.Turbo Mode
.APM (Power Mgmt.)
.CPU Thermal Throttle
.C&Q
.C1E
.C6

I have set Win7 to High Performance and made sure that the min CPU state was 100%. Yet when I run any stress test, all cores will drop back to 1400 MHz briefly several times a minute along with the core volts dropping down as well. This is the 2nd FX-8350 I've had do this and the 2nd board it did it on. I'm open to suggestions from the AMD aficionados out there.
 
Have you played with loadlinecalibration and increased that. Also, during the stress test, is the voltage dropping at all? Might need a bit of boost to the NB voltage as well to handle the OC. Lots of things to adjust when it comes to ocing these beasts. Took me weeks to hit 4.86. lol.
 
As st3v30 mentioned, maybe more cooling over the vrm? It's a heatsinked vrm design but I've seen multiple people complain of the vrm heat issue on that particular board. One person on newegg mentioned the same exact throttling problem with a 9590 and said they were having the same issues with an 8350. Asrock's response was better cooling/ventilation for the vrm.

"Other Thoughts: have asrock 990fx extreme6 with 9590 . cpu jumps from 4,7 to 1.3 to 4.4 4.5 5.0 mhz. it not the cpu .I have 8350 it does the same thing .I put the 9590 in my asrock performance fatality it work fine.

Manufacturer Response:
Dear rollandthedrifter,

We sincerely apologize for any inconvenience this may have caused you.
Please allow better air circulation inside the chassis and VRM are to avoid overheat.
Once system is overheated the CPU throttling (dropping multiplier) will occur to avoid damage the processor chip.
If using FX series with 125w and 220w CPU, please install a CPU cooler with a top-down blowing design."

Another user wrote "Cons: The components under the 990FX part caught on fire after ~ 6 months.
perhaps a defective mobo ?" - The vrm being under the '990FX part'.

Yet another - "Cons: After two weeks of usage, it started to have a heating issue with the north bridge, ranging from 144f to toppin out at 205f . causing random restarts and random video drivers to delete them selfs. and random unstable issue`s"
 
Yes, I did have to add a fan over my VRM's ( stock AMD heatsink) with a wood screw through 1 hole into the VRM heatsink fins, and it fit perfect and stays in place. The wood screw threads just thread through the fins enough to stay firm and not bend them.
 

clutchc

Titan
Ambassador
Ah! The VRM overheating issue. You fellows might be onto something. I noticed one of the HWinfo temp readings for the board getting into the upper 70s, but I didn't know what component it was. I will do some experimenting with cooling the VRMs. Thanks for the tips.

I have a 212 EVO on at the moment. The only other cooler I have available right now is another Corsair H60. But the CPU temps are fine even with the 212. So I'll work with the VRM cooling and see if that makes any difference.

@ getochkn
How much did you end up boosting your LLC and NB by? And why would that cause throttling?
 
I agree it can take real detective work sometimes with the abstract labeling of sensor data on monitoring programs. Different programs list them under various 'names' and you're left playing sherlock holmes to figure out what actual component the temp sensor is reading. Or if it's even reading a proper component and not just an unused sensor field reporting some arbitrary number.
 

clutchc

Titan
Ambassador
It doesn't appear an extra fan on the VRMs did any good. In fact, I doubt it is necessary. The VRM heat sink is right between the 'pull' fan on my 212 EVO and the case exhaust fan. There is gobs of air flow across the heat sink.
2lDY4BM.jpg


That and the fact that the only warm temp on the board was @ mid 70s max when throttle occurred. Hardly too hot for a the VRMs if that's what it was. The other temps were all in the 30s.

@ getochkn
I will try the LLC and NB increases next. And yes, the core voltage drops along with the multi during the stress test.
 

clutchc

Titan
Ambassador


Well, that 70+C is the hottest MB item on HWInfo's list. If that isn't the VRMs, then they would be one of the cooler temps. Btw, I tried taking the CPU NB up in increments to 1.4V (a little higher than I like), but no change in throttle. And I can't find any reference to load line calibration in the BIOS.

Hopefully it won't throttle in everyday usage. I ran the Firestrike benchmark using my GTX 970 and got a 9210 score. Seems like it would have been lower if throttle was rearing its ugly head during the run, no?
 
Ya, I go for real world and have backed off my 4.86 to I think 4.2 with 200 or 300 turbo boost available. Maybe for the winter if it gets chilly. lol.

For more general performance, dialing back and using the turbo clock is better since not everything can take advantage of 8 cores, I rather have a couple cores that can hit higher speeds when needed and less on all 8.
 

TRENDING THREADS