Some questions about overclocking

john01dav · December 25, 2018

Hello,

I would like to overclock my Fx-8350, and I have some questions:

I am able to have a stable system under synthetic load at 4.3GHz with no voltage boost, but the temperatures are nowhere near 90°C, so I want to try to push it further. When I bring it up to 4.4GHz, however, it seems stable for awhile, but, after a minute and a half or so under synthetic load, the screen freezes and stays frozen for ~10 seconds and then the system reboots. This is on Linux, with CPU frequency settings being changed in the bios. What does this mean? Did I hit a thermal limit (The last recorded CPU temperature before my screen froze is 42°C, which seems a bit low for this, although the temperature was still rising 10-20°C per minute)?
I suspect that the previous issue is due to a low voltage (I am interested in knowing why voltage can cause that though), suggesting that the solution is to raise the voltage. I am very hesitant to do this though because I have no reference frame for what voltages are safe and what voltages aren't. I prefer some idea of why some voltages are safe and some aren't, as opposed to the plethora of contradictaory and unsubstantiated numbers I can find all over.
JayzTwoCents' video for this CPU recommends changing a bunch of settings (some of which don't exist in my comparatively crappy bios). What are these settings and what do they do? Why does he recommend changing them?
One particular recommendation of interest in the aforementioned video is northbridge overclocking. What is this and why does he recommend it?

Thank you

TheBean · December 25, 2018

could be a limitation in silicon? (silicon lottery)

could be voltage protection in bios?

low quality mobo?

could be many things. but if it doesnt work, dont try. you could damange your cpu permanently

LIGISTX · December 25, 2018

48 minutes ago, john01dav said:

Hello,

I would like to overclock my Fx-8350, and I have some questions:

I am able to have a stable system under synthetic load at 4.3GHz with no voltage boost, but the temperatures are nowhere near 90°C, so I want to try to push it further. When I bring it up to 4.4GHz, however, it seems stable for awhile, but, after a minute and a half or so under synthetic load, the screen freezes and stays frozen for ~10 seconds and then the system reboots. This is on Linux, with CPU frequency settings being changed in the bios. What does this mean? Did I hit a thermal limit (The last recorded CPU temperature before my screen froze is 42°C, which seems a bit low for this, although the temperature was still rising 10-20°C per minute)?

I suspect that the previous issue is due to a low voltage (I am interested in knowing why voltage can cause that though), suggesting that the solution is to raise the voltage. I am very hesitant to do this though because I have no reference frame for what voltages are safe and what voltages aren't. I prefer some idea of why some voltages are safe and some aren't, as opposed to the plethora of contradictaory and unsubstantiated numbers I can find all over.

JayzTwoCents' video for this CPU recommends changing a bunch of settings (some of which don't exist in my comparatively crappy bios). What are these settings and what do they do? Why does he recommend changing them?

One particular recommendation of interest in the aforementioned video is northbridge overclocking. What is this and why does he recommend it?

Thank you

Wow, thats an old video from him lol. Way before I started watching him.

But, to address the questions, and to put some info out there to start, I have never overclocked AMD (theory is the same, but I have no practice with their settings and lingo for settings), and same with Linux. I am a windows user for desktop applications, and I don't overclock my servers

So, that being said, lets see what we can do here. So system being stable at 4.3 means either stock volts is enough to support this, or whatever auto setting the board is using is within the range its comfortable providing. Every chip is different and the voltage it asks for is not the same chip to chip, so the board has some leeway in what it will do out of the box. 4.3 may need a voltage within that range. To know for sure, I use CPUz in windows as it shows live time vcore. That with Coretemp and I can tell if things are running "within reason" prior and during stress testing. Within reason is different to everyone, but usually you can google recommended voltages for every chip, or at least series of chips. I have no first hand knowledge of your chip's voltage, but I guarantee google can tell you what is normal, safe for 24/7, safe for handful of time record setting and then LN2. I usually try and stay a bit bellow "the 24/7" recommendation as a few hundred more Mhz really won't make a huge difference day to day when that few Mhz takes legit percentages more voltage to attain. For instance, I am sure my 6700k in my sig can do more than 4.7,especially if I delid it (i delid my 4770k with AMAZING results) but it gets to a point where it will likely need a lot more volts for a minimal gain, IMO not worth it.

42c seems VERY low, is that under load or at idle? If that is idle, then its hot, if its load, its cold. lol.

So, yes, low voltage will cause it to crash. You can try 4.5-4.6 Ghz with same voltage you are at now and I bet it won't even make it to windows, and will BSOD during boot (or, well, whatever linux would do). It just can't operate with the volts being provided and crashes. Once again, I will google what people say is acceptable, set that vcore and go from there. Or, there are a few mentalities, you can set the max vcore your brain has accepted as "ok" and keep pushing clock speed until its not stable and then back off ~100-150 mhz to be safe, or go up incrementally, getting each speed stable and eventually settling on a speed you are ok with. As to why some are safe and some are not. Well, that's "simple". If your cooling solution isn't adequate your chip will get to hot, which goes hand in hand with the temp limits people say (usually stay below 90c under max load, personally I don't let mine get over 80c under max synthetic load). But, even with temps under control, voltage can kill components. Giving it more volts can just cook things, I am not an EE so beyond saying that, I can't give you much.

Once again, not huge on AMD, but intel northbridge would be what controls memory controller and other motherboard devices. This can also control underlying board settings if I remember correct from Intel, so that may be able to help the board keep up with CPU speed changes in a manner of speaking. Honestly, the best way to really understand it is to just read A LOT of articles from respected sites such as toms hardware, gamers nexus, ars technica etc that explain what things do. It takes time, but to really understand it that is what it takes. Back on my core 2 duo which was the first CPU I ever overclocked back in ~2007 I spent WEEKS reading and learning before I ever tried. I know people who grab a K sku Intel chip, watch one video, put some random voltage in there and GHz and just hope it works, no tweaking to get lowest possible voltage for a given speed, or trying to really dive into it. Do their PC's work, sorta mostly lol. But personally, even now after I do have a good handle on it, when I get a new CPU I spend about 4-5 days reading and seeing what settings are new since my last platform, what voltages and temps I should be looking for, and start testing. It can take me a week to get a solid stable overclock with all the trial and error. Bump vcore here, push MHz there. Rinse and repeat until I find a happy medium.