Hey all, I'm back to OCing my FX8350 again as I have time to kill while I'm job hunting and I'm running into some difficulties.
Currently I've been trying to hit 4.8 GHz unsuccessfully at 1.450v as my vcore.
Now, I've cleared p95 blend test for almost 3.5 hours with no issues and yes I probably should let it run for the recommended 7+ to know for sure but I've been impatient 😅 so I also suplemented with a some time in league of legends, cs:go and fallout 4 on top of some IBT linepack.
The only time my system freezes up now is during champion select for league about 50% of the time. After I have no issues and never an issue in any of the games. I have bumped up the voltage 3 times so far from when I first froze but I've had no luck. Is it possible its not the core that go unstable but the HT or NB at this pointn
Long story short I was wondering if someone had some recomendation as to a range of fft's to run in P95 to further test the stability in a way that would best mimic the champion select process.
Long shot, yeah I know. Figured I'd ask the community if anyone has seem similar issues or suggestions.
1.45v is too low try 1.47 or 1.50v would also help if you're on a custom loop, an AIO or air cooling. As well as motherboard model. Also if you have access to infrared gun or socket temps etc..
Thanks for the replies guys, sorry forgot to call out the mobo and cooler, I have the Asus M5A99FX Pro r2.0 with nothing additional on the vrms and then the EK L240 kit on the cpu keeping it cool. The VRM heatsink isnt too hot to touch I can easily hold my fingers on it for 2 seconds before it starts to get close to uncomfortable, don't know how that relates since i think only once i had heard if its too hot to touch then the VRMs are deffinitely too hot.
I do have 4 case fans blowing as well: one rear exhaust, two front intake and one of the Noctua ippc 2k rpm fans blowing on the socket from the back.
@LordXenu I really wish i had an IR gun to check temps but i dont and i dont know of friends that have one. Might search around and find the cost or renting one for the day. Just cleared 4.75 @1.45 for 9 hour P95 custom for 100-4000 fft's with 12 gb ram used (out of my 16 total) and I was getting real close to my thermal limit at 4.8 and 1.45v so I might need more rad space if im pushing volts higher, nbd if thats the case itll just be put off for a month or so.
@ Zibob did not realize P95 wasn't recommended, i use Aida instead from here on out. I haven't noticed any overvolting showing up on HWiNFO64 as ive been testing but i guess could still be over stressing it other ways.
@Giulianno_D I guess it could be but the crashes have been before any gaming starts and not during. Going from the desktop to connecting to the league of legends champion select screen seems to be where it freezes on me. I do have a handful of Chrome tabs open with Discord open as well but cpu usage stays well below 20% for most of it.
My 8320e would be fine on prime95 or Aida....but it would kick me out of games or freeze...upping the voltage on the NB fixed it. I don't have any stuff on the SB on my cheap board or I would have went their first. Look for little subtle signs during boot up (like num lock not coming on, sb or nb) or mem error on memtest (cpu/nb). I would back down the cpu volts to what you had them first and make sure it is as stable as it was. Also try the HT voltage if that fails. Another way to do this is just drop down the SB NB and HT speed or multiplier and turn them up one at a time. Your board is probably better then mine....so you probably have a ton of options that I don't have...like LLC. Also did you up the dram by .10v? I only use prime 95 for 5 min on blend then 5 min (or less) on small. This has on occasion pointed to a dram voltage adjustment instead of a CPU adjustment. I would also recommended testing with open GL. My 8320e with HT at 2600 a 4.0 blows away 4.4 when the HT is low (90fps vs 78fps). Hope this helps
Your OC isn't stable, period. It's not just about Prime95, that app only test one aspect of instruction set in the CPU. Your OC should be stable under multitudes of instruction sets. Aida64, Cinebench, and Asus Real Bench will test these for you. Also, there are two different bins of FX8350, before the FX9000 series was released, there were more FX 8350s that could easily hit 5.0Ghz 24/7 stabe, however, AMD reserved the high bins and sold them as a separate SKU, hence FX 9000s. Most FX at best these days can hit only 4.6-4.7, 4.8 if you're lucky. Also, 1.4-1.5v isn't too low, that's the max ranges you'd wanna push the chips. And if you're pushing those types of voltages, you're sure as heck your VRMs are getting hot, 2 seconds hand test is meaningless. Make sure you're getting airflow to them.
That is a really good point. The paths through a CPU are different. Test one path is not a tell for all the paths. You have to run it through all the instruction set as much as possible then dial it back 100Mhz or so.
@Giulianno_D Thanks for the advice, makes sense that I should be using multiple bench's to check stability. I have got the 8350 24/7 stable at 4.6 and 1.406v from when I had my NH-D14 cooling it so I've made the poor assumption that putting it under water with just a 240mm rad would be able to jump it another 200 MHz i think.
Ill go back to incremental OC'ing and start at 4.65 or 4.7 again find the stable voltage and keep it for a week or two before jumping up again. Guess I've gotten impatient lol.
Ill also grab the fan from the stock cooler (or ~80mm one) and get it blowing on the main vrm heatsink.
Nothing i do really needs the extra compute power so no need for a 9370 and im looking to potentially upgrade to one of the skylake hex cores when they drop next year sometime, funds permitting and assuming rumors on their release is true.
Ok, so I have backed my clock off to 4.72 and then tested and passed 9 hrs of Aida, 20 normal runs of IBT, 5 consecutive runs of Cinebench multithreaded test and then froze up on chrome 40 mins into browsing -_-
So just for grins I bumped the CPU NB and NB voltage up one notch (0.00625) and haven't had a lock up in the last 3 hours....
Fingers crossed thats the issue
Altho, I had never touched those to begin with but if it fixes it who cares lol
Edit: forgot to mention its 4.72 GHz at 1.4625v, NB clock is 2147 GHz at 1.10625v, and cpu nb is 1.25625v
So I haven't had a lock up yet but its only been a few hours of use since I dropped back down to my 4.6 GHz settings. Is it possible that the NB is just marginally stable and doesn't get stressed enough at 4.6 but gets pushed just to the brink of instability when i bump the core clock up?
@anon85933304 My HT is clocked just short of 2.600 GHz and NB HT voltage is the stock 1.2v
Have you allready tried to keep the multiplier at stock, but instead of that just overclock on the Northbridge? I reached 4.8GHz stable out of my FX8350 + Asus M5A97 EVO board, which has the same vrm implementation.