[ssateneth]

vdroop is normal. don't try to remove it. it's there to help smooth out spikes of voltage (spikes you can't see in software sensors; you have to use an oscilliscope) because the VRM can't react as fast as a cpu load can fluctuate. If you're pushing dangerous voltages, it may help to flatten the curve a little in order to run less idle voltage, but having a flatter curve usually means you are setting a lower voltage, which means even lower voltage during sudden load transients which can freeze the system. if you are running high voltage with a flat curve, then you send even higher voltage transients when a load disappears.

you are saying you are still getting crashes. have you established that you get no crashes when running fully stock? you might not have read my last post. You can't establish what is causing the problem because you changed umpteen setting willy nilly and now you don't know why it's crashing. If you run default settings and establish it's fully stable, then change 1 setting at a time and establish it is still stable, then you repeat until it's running at the speed you want while stable.

your crashes are random and intermittent, so you can't expect this process to be fast. 30 minute tests between each change may be fine for benchmark-levels of overclock but not for 24/7 operation. you need to establish that the system is, without a doubt, stable. this will take a long time to do properly, otherwise you'll either have to just run a mild and safe overclock (or none at all), or deal with crashes/freezes with an untested overclock.

 

[Senotajs]

Unless you're using memtest86 7.2.0 and Modulo very long run - use HCI Memtest.

Ran HCI Memtest for 4 hours, no errors whatsoever. Will attach an image of the results.

vdroop is normal. don't try to remove it. it's there to help smooth out spikes of voltage (spikes you can't see in software sensors; you have to use an oscilliscope) because the VRM can't react as fast as a cpu load can fluctuate. If you're pushing dangerous voltages, it may help to flatten the curve a little in order to run less idle voltage, but having a flatter curve usually means you are setting a lower voltage, which means even lower voltage during sudden load transients which can freeze the system. if you are running high voltage with a flat curve, then you send even higher voltage transients when a load disappears.

you are saying you are still getting crashes. have you established that you get no crashes when running fully stock? you might not have read my last post. You can't establish what is causing the problem because you changed umpteen setting willy nilly and now you don't know why it's crashing. If you run default settings and establish it's fully stable, then change 1 setting at a time and establish it is still stable, then you repeat until it's running at the speed you want while stable.

your crashes are random and intermittent, so you can't expect this process to be fast. 30 minute tests between each change may be fine for benchmark-levels of overclock but not for 24/7 operation. you need to establish that the system is, without a doubt, stable. this will take a long time to do properly, otherwise you'll either have to just run a mild and safe overclock (or none at all), or deal with crashes/freezes with an untested overclock.

I am taking your advice and started literally from 0. Just reset BIOS to factory settings and didn't change a thing and let run prime95 for 30minutes on stock, no errors nor freezes. Also deliberately started and stopped prime95 5 times to make sure that it doesn't crash when the load eases(it always froze when I did this). Just now increased Clock rate to 4.8GHz and Vcore to 1.360, ran it for 30minutes, everything went smooth, also did start and stop prime95 5 times. Now I will do some gaming to see whether everything is ok under realistic loads as previously prime95 passed but froze when gaming.

To make it clear, the only BIOS settings I have changed so far are;
OC Explore Mode > Expert
CPU Ratio > 48
CPU Ratio Mode > Fixed
CPU Core Voltage > 1.350 (HWInfo reports 1.360)
Intel C-State > Disabled

I'm not planning to run my system 24/7 and never have done so.

Crap i almost forgot as you should be aware that the GPU you have wont work with early models of the Seasonic Focus Plus so if you get one that is old stock you are going to have a ton of problems with the GPU since the Asus GTX 970 Strix specifically had problems with the Seasonic Focus Plus

Seasonic addressed the issue and newer models should not have that problem so you should check when it was made before installing it
Or you could replace the GPU since its old anyway

How do I check when it was made? Although haven't experienced a GPU failure so far.

 

[shilka]

There should be a date somewhere on it

 

[Senotajs]

Thank you all for your input! Appreciate everyone's contribution!

The problem was one of my ram sticks. Shortly after I restored my bios to default settings I changed one thing at a time until it came to selecting XMP profile. As soon as I selected XMP profile 1, the crashing started again. Then I ran the Windows Memory Diagnostic for like 5th time and for the first time ever, errors were reported. Then I ran the same Windows Memory Diagnostic on each stick separately and discovered that one of my ram sticks kept reporting problems. I sent back the flawed ram stick and received a replacement a few days later. Now I am rocking my 4.8GHz overclock with no problems at all, runs smooth!

All of these point to corruption of data in RAM. CPU issues would arise more likely with clock watchdog timeouts, WHEA errors or L1/L2 cache errors (found in hwinfo64), or in the case of unstable cache (though not always), system freezes.

Testing ram stability while maintaining high speeds can take a very long time. "stable enough" comes from situations where you only run short stress tests or stresses not good enough that leads to problems where your system is inexplicably crashing now where it wasn't before. Some people say "oh, i degraded my system somehow". No, it just wasn't stable to begin with. Unless you are shoving dangerous volts for long periods of time without appropriate cooling (I draw the line at 1.55 vcore, 1.7 vdimm, 1.35v sa and 1.35v io), I can't see any "degrading" happening unless the part was destined to die to begin with.

I prefer GSAT and HCI Memtest (At least 8 hours, sometimes 16 as even the most intermittent/rare error means a fail in my book. "stable enough" is not stable), and sometimes P95 with AVX and/or linpack stress tests if I want to test cooling (At least 8 hours). It'll take about 1.5 to 3 weeks to dial in a proper 24/7 overclock and requires you not using the PC at the same time.

Could go into the process of dialing in a stable overclock or troubleshooting the issues you are getting, but to do it properly requires a lot of detail. You'd more or less have to start at stock speeds / JEDEC non-XMP settings and go from there and change one thing at a time between each stress test. If you change a bunch of things and then you start getting errors, you won't know which setting exactly is causing the error and you're back to square one.

You, sir, were right in regards to information you provided on ram. I guess I got lucky with capturing the ram issue with Windows diagnostic as I did it for around 5 times in total and ran several memory tests for hours with no luck of finding anything.

 

[misoonigiri]

Ran HCI Memtest for 4 hours, no errors whatsoever. Will attach an image of the results.

FYI your HCI memtest screenshot showed 3.2MB x5 = 16MB tested only (in megabytes not gigabytes)

 

[wingman99]

Glad that you found the problem with the memory and solved it.