Want to help a beginner troubleshoot Fx8350 OC?

[Spaghetti_Bender]

Hello!

I have a FX8350 in a Gigabyte GA-990FXA-UD3 Rev. 1.0 motherboard.
Coolermaster hyper 212 Evo (upgrading to Noctua NH-D15)
Seasonic FOCUS Plus 850W power supply.
32 gb ballistic tactical 1866 ram 1. 5v
RX 580 GPU
Fractal meshify C case
I have been researching and trying to overclock my cpu for the last 1.5 years haha. I have never succeeded. This is a project I'll pick up every few months read some new material or watch some new videos and play around with my fx for a few days before I inevitably fail.

This time things are going to be different because I am going to ask for help. Previously I was using a cooler master hyper 212 evo but now my cooling solution is a Noctua NH-D15. I used CPU-Z to monitor temps and clocks but in the past I got weird numbers. I read on another forum that cpu-z might not fully support my cpu.

So to begin with in bios I will turn off all the auto power saving and overclocking settings and rase the frequency to 4.2ghz or so but leave the voltage to stock but manually typed in. I'll set the ram to manual and type in the timings and try booting. Sometimes it works for a little while but after I try to run unigine heaven or play a game the computer freezes or crashes and I will have to clear the bois with the jumper on the motherboard and start all over. Temps would be high around 80 to 86C.

1. What are your suggestions for hardware monitoring software?
2. I have not tried the noctua cooler yet but I did try once more with my hyper 212 and ended up with the same results as before. Is there anything in the bios I should do first to increase the odds of success?
3. Is there any other information you need from me? I read though this forum for a few hours before I posted this first thread. So please help me out if I don't have everything in order for you all.

Thanks in advance!

 

[Mussels]

You're going about this the wrong way, and it WILL cause you problems

"weird numbers" says nothing useful. Elaborate?

A.Do not disable power saving functions. Just. Dont. Do. It.
B. Your VRM's are going to overheat on that board, especially with those style coolers. Doesnt matter how, but get a fan blowing air directly onto them.
C. Crashes like that simply say the system was unstable, and if it wont even POST afterwards, that leans heavily towards a component overheating severely - VRM's being most likely.

1. Find some demanding stress tester and temp software, and test it at stock clocks. I bet you'll find you've got hot temps and overheating VRM's at stock as well (software cant read them, you'll find out by burning your fingers on them, or by seeing your clocks go DOWN at load)

2. See my repeated advice over the VRMs

3. "weird numbers" needs an explanation, if it was seeing your clocks drop down lower when OC'd, thats definitely due to the CPU or VRMs overheating

 

[Bones]

+1 with the above for sure.
You're running your system for milder useage unlike what I do so as said above keep the power saving stuff going.

Personally I do disable all the power saving crap but that's due to how I run my system, going for max clocks is something such things will hold you back on but also it's because I've been running FX that way since it first came out and means I have the experience to run it however I want.

With that said......
DO NOT set your Digi (VRM) settings to 130% like I've seen some say to do, that places a crapload of heat and stress on the board's components, not to mention the CPU itself being made to run hot when it doesn't have to.
"Auto" is fine for most anything you'll be doing period since you're not running anything beyond aircooling anyway from what you've described. Those extended range settings are for the crazies (Like ME ) and nothing else to be honest. And also, I never run mine at 130% anyway, never needed to and the max I've ever had mine when running things with sub-zero cooling was 110% so 130% on air is just nuts.

BTW if you would say what settings you are using for things like CPU voltage as an example, we can get a better idea of what's going on and advise from there.

 

[Spaghetti_Bender]

Thanks for the replays that is extremely insightful. I will set everything up like I normally do minus the power saving and and the VRM. I will post all of this tomorrow morning because of my work schedule though.

I didn't know about the VRM stuff at all nor to leave the power saving bios settings in auto. Is the VRM % the same thing as VCORE load line calibration? I think that's the only setting that is increased by percentages like you describe.

The CPU-z and HW monitor weird numbers are as follows all in stock at idle
CPU VCORE = 1.45
CPU temp = 42 to 45 C
Amd Overdrive shows cpu thermal margins bounce between 42C and 52C

I guess what I mean by weird is they seem high to me so I'm not sure if the software is reporting correctly.

But I will post all my bois settings and sock idle/load parameters tomorrow. Thanks for helping me get srarted!

 

[Shambles1980]

idle temps at 42-45 with 1.45v does not seem that high to me.
Would i want to get them lower?? yes i would, but its a 8350 running at 1.45v its 1.38v stock and slight voltage increase = quite a bit more heat.

 

[Mussels]

if you run something line linpack xtreme/intel burn test for a long time (20 passes at least) you'll possibly see your CPU temps nice and stable, but your actual scores drop right down - thats how you can tell the CPU has been throttled by the VRMs

 

[Bones]

Thanks for the replays that is extremely insightful. I will set everything up like I normally do minus the power saving and and the VRM. I will post all of this tomorrow morning because of my work schedule though.

I didn't know about the VRM stuff at all nor to leave the power saving bios settings in auto. Is the VRM % the same thing as VCORE load line calibration? I think that's the only setting that is increased by percentages like you describe.

The CPU-z and HW monitor weird numbers are as follows all in stock at idle
CPU VCORE = 1.45
CPU temp = 42 to 45 C
Amd Overdrive shows cpu thermal margins bounce between 42C and 52C

I guess what I mean by weird is they seem high to me so I'm not sure if the software is reporting correctly.

But I will post all my bois settings and sock idle/load parameters tomorrow. Thanks for helping me get srarted!

The VCORE calibration you speak of is probrably the same thing (I've never owned a Giga board to really know) but regardless there is no need to have it set for anything else but "Auto".
To get 4.2GHz from your chip 1.45v's sound a little high as in it can do that with less voltage used, probrably could get that with just 1.40v's instead making it run a little cooler.

I typically run my CPU-NB voltage around 1.100v's if at stock (2000MHz), since a Vishera doesn't like high CPU-NB speeds anyway, I just leave it at that unless I'm running it hard. You'd normally get about 2400MHz from it reliably for everyday use, anything beyond that shoudn't be expected but some can do it.
The lid with that is around the 2600MHz mark or so and thats after volting it up but in your case thats something you don't want to do, 2400 is doable even on air so that would be the max I'd go for, you can use about 1.20 - 1.25v's and it should work there.
Thing to remember here is with CPU-NB voltage, that can make a chip run a little warmer too.

As for temp fluctuations that's actually normal to see between idle and load states of the chip and your temps according to what you posted are fine, the max you'd want to see a Vishera hit UNDER LOAD is 62c or so.
Idle temps are largely useless as reference except in extreme cases of cooling problems, always base what temps it's hitting under load.
Since I do have an AM3+ setup in my benching area ATM I could try and grab a few pics of it's BIOS to illustrate what I see/use in the BIOS here.

 

[Spaghetti_Bender]

I am running a few tests to get a base line. I am also currently trying to set up the fans so move more air at lower Temps.

What I have determined is that my load Temps are stable in the mid 50s. I ran unigine Valley benchmark and Linpack Xtreme each 3 times with my system completely stock.

The stock CPU voltage is almost always at 1.45v at 4100 MHz, the Northbridge voltage is 1.1625v at 2200 MHz, and the HT Link Speed 2600 MHz. When I try to increase the CPU frequency I'll should I start at a lower voltage?

I will turn off Core Performance Boost but leave enabled Cool & Quiet, Core C6 State, and Amd Application Power Management (enabled =default = dynamically monitors the power consumption of the CPU cores and automatically optimizes the CPU to its best performance level).

There is a setting High Performance Computing Mode that is disabled by default. According to the manual HPC mode default = disabled "enabled prevents the CPU frequency from being lowered during system halt state. I will leave this in default disabled for now.

 

[TheMadDutchDude]

You should know that those CPUs become massively unstable after 60c... mine couldn't even survive past 63c on the cores.

 

[eidairaman1]

You should know that those CPUs become massively unstable after 60c... mine couldn't even survive past 63c on the cores.

Hmm how odd, mine passed Ryzen Blender at 75 at 5.0GHz.

https://www.techpowerup.com/forums/...ks-at-200-samples.228686/page-13#post-3582441

I have power saving functions off and I think HPET enabled, No LLC, NB voltage set to what auto initially said,

NB is at 2400 due to ram being 2400.

200x25 1.476Vcore- it creeps to 1.524 under stress load.

 

[Spaghetti_Bender]

I did not know about HPET but I checked my Bios and it is enable by default. I tried to run some benchmarks (linpack Xtreme and unigine Valley) after changing my bios settings to manual. The unigine Valley ran normal but the Linpack Xtreme stress test failed. The Vcore voltage changed to 1.33 with everything in manual as well. Below is what I changed in bios:

Manual Frequencies
CPU Core 4100
NB 2200
HT 2600
Bus 200
The only multiplier I changed was setting the CPU Clock Ratio to 20.5

voltages
I only manually entered CPU Vcore to 1.41250V which is what my default settings are.
Every thing else was left in auto (I did not manually enter values)
NB core
DRAM
HT Link
NB/PCIe/PLL
NB Vcore Loadline Callibration

Advanced Frequency Settings (power savings and amd technologies)
Core Performance Boost = Disabled (enabled by default)
Cool N Quiet = Enabled (default)
C1E = Enabled (default)
SVM = Disabled (enabled by default)
CPU Core Control = Auto (default)
Core C6 State = Enabled (default)
HPC mode = Enabled (disabled by default)
APM = Enabled (default)

The memory settings were left in XMP auto

Like I said the Linpack Xtreme stress test failed with these settings. If I can't pass the stress test with these settings I must be doing something wrong. Any suggestions to make my system more stable so I can finally try and increase the CPU Clock Ratio to overclock this thing? I will also install amd blender to add that to my stress testing. When I ran linpack Xtreme and Valley the temps were normal ~ 55C. But like I said above the CPU voltage changed to 1.3v and stayed stable with the clock stable at 4100 as well. This is the same problem I had in the past. I couldn't figure out what settings I needed to even get the fx8350 to run at stock speads outside of the mobo default settings.

 

[TheMadDutchDude]

Those CPUs came with a massive overvolting when they were at stock, FYI. Mine was 1.425v and ran stock speeds at 1.2v without issue. It's all about tweaking. Welcome to the wonderful world of crashing your head into the wall when things don't go right.

 

[Batou1986]

sync your NB and HT to 2200mhz and try 1.350v ~ 1.375v with LLC on extreme, sometimes more voltage just makes things worse on the FX chips in my experience
I easily got my 8320 to 4.2ghz on a cheaper 970 version of that board just by setting multiplier and Vcore it was AVX linpac stable.

 

[Bones]

Those CPUs came with a massive overvolting when they were at stock, FYI. Mine was 1.425v and ran stock speeds at 1.2v without issue. It's all about tweaking. Welcome to the wonderful world of crashing your head into the wall when things don't go right.

You got that right - The 9590 is a great example of this. It's stock 1.50v's is just too damn high, the chip can run 4.7 with far less used making it run cooler too.

@ the OP:
If it did fail running Linpack you can bump the CPU voltage up a little.
In fact I'd do that incrementally until it does pass but do be aware your system will hardly ever see such harsh conditions that Linpack induces in real life useage. I'd expect your chip to need at least 1.42v's or so to pass minimum, in fact your previous voltage setting could be it. I know I said 1.45v's might be a tad high earlier but each chip acts differently than another would, yours could indeed be hungry for voltage to run.

 

[eidairaman1]

SVM is for vistual machines,

HPC enable that.

 

[Spaghetti_Bender]

Thank for replaying everyone. I will enable HPC before I change my HT and NB frequencies to 2000 MHz. I will start with a lower voltage 1.35v and see if I can get it stable at 4.0 GHz.

I did not know that Linpack was so harsh. However I would still like to see it pass at 4.0 GHz so I will give it a few tries. It passed the Unigene benchmarks though so maybe I can let it slide?

Should I enable Core Performance Boost? The AMD power saving and performance technologies are the hardest for me to understand. I have seen many posts were what should be enabled and disabled has been debated.

Thanks again for all the info. I am already learning more then when I tried to do this on my own. I hope soon I will not be the only soul that owns an FX processor without an OC. LOL

 

[eidairaman1]

Thank for replaying everyone. I will enable HPC before I change my HT and NB frequencies to 2000 MHz. I will start with a lower voltage 1.35v and see if I can get it stable at 4.0 GHz.

I did not know that Linpack was so harsh. However I would still like to see it pass at 4.0 GHz so I will give it a few tries. It passed the Unigene benchmarks though so maybe I can let it slide?

Should I enable Core Performance Boost? The AMD power saving and performance technologies are the hardest for me to understand. I have seen many posts were what should be enabled and disabled has been debated.

Thanks again for all the info. I am already learning more then when I tried to do this on my own. I hope soon I will not be the only soul that owns an FX processor without an OC. LOL

Iirc i turned off any turbo core functions and c1e power states, I ran my chip unthrottled.

 

[eidairaman1]

Would i want to get them lower?? yes i would, but its a 8350 running at 1.45v its 1.38v stock and slight voltage increase = quite a bit more heat.

Mine stock was 1.36.

Hey I run 5.0GHz at 1.476, 55 Gaming.

Under extreme worst case it creeps to 1.524 and it's at 70-75.

 

[Ruru]

Personally I do disable all the power saving crap but that's due to how I run my system

Same here, Ryzen 5 2600 @ 4.15GHz 1.4V even with idle.

You should know that those CPUs become massively unstable after 60c... mine couldn't even survive past 63c on the cores.

Didn't Phenoms have the same thing? With FXs it's funny since they generate pretty damn much heat.