3700X - Overclocking Stability question

[SFFDesigns]

For a while, I've been using P95 to test my 24/7 OC (not used for video encoding or anything, just gaming and "regular" usage) but I just found this:

https://www.reddit.com/r/overclocking/wiki/faq#wiki_...amd_ryzen_3000_series_.22matisse.22_.287nm_non-g.29.3F

 

On this page, it is suggested to use x264 Stress Test if you are okay with a ~99% stability OC. I am for sure, as I'll take every bit of performance I can get (FS2020 hammers the CPU especially in single thread). 

 

Is this an okay measure of stability. Like will only testing with this result in constant game crashes, or is it perfectly fine for non essential use (im not rendering for 24 hours or anything)?  I might be able to push further than 4.2 GHz if so.

[Bombastinator]

13 minutes ago, Hymenopus_Coronatus said:

For a while, I've been using P95 to test my 24/7 OC (not used for video encoding or anything, just gaming and "regular" usage) but I just found this:

https://www.reddit.com/r/overclocking/wiki/faq#wiki_...amd_ryzen_3000_series_.22matisse.22_.287nm_non-g.29.3F

 

On this page, it is suggested to use x264 Stress Test if you are okay with a ~99% stability OC. I am for sure, as I'll take every bit of performance I can get (FS2020 hammers the CPU especially in single thread). 

 

Is this an okay measure of stability. Like will only testing with this result in constant game crashes, or is it perfectly fine for non essential use (im not rendering for 24 hours or anything)?  I might be able to push further than 4.2 GHz if so.

With my old 4770k cpu I could get 4.2 or 4.3 using prime95 testing.  I could do that for an hour no problem.   Irl though if I wanted actual fo’realz stability it was 4.0. After a couple weeks after each crash, 4.3 went to 4.2, went to 4.1, went to 4.0, where it actually stayed. 

[SFFDesigns]

2 minutes ago, Bombastinator said:

Irl though if I wanted actual fo’realz stability it was 4.0.

For me it's been the opposite. I've comfortably used my 3700X at 4.3 GHz and 4.275GHz, it only crashed in P95. The reason I ask about this test is that if it's actually a good indication of day to day stability with gaming and just usual web browsing, then I can 100% get much higher clocks

[Bombastinator]

2 minutes ago, Hymenopus_Coronatus said:

For me it's been the opposite. I've comfortably used my 3700X at 4.3 GHz and 4.275GHz, it only crashed in P95. The reason I ask about this test is that if it's actually a good indication of day to day stability with gaming and just usual web browsing, then I can 100% get much higher clocks

Oh I comfortably used it too... till it crashed.

[SFFDesigns]

1 minute ago, Bombastinator said:

till it crashed.

Yeah that would really suck if it crashed during a long FS2020 flight.

 

If this actually tests stability tho, then i could probably get a decent uplift in perf

[Bombastinator]

1 minute ago, Hymenopus_Coronatus said:

Yeah that would really suck if it crashed during a long FS2020 flight.

 

If this actually tests stability tho, then i could probably get a decent uplift in perf

When I had it at 4.1 it went for nearly a week of solid daily gaming before I got a crash.

[SFFDesigns]

8 minutes ago, Bombastinator said:

When I had it at 4.1 it went for nearly a week of solid daily gaming before I got a crash.

Okay, that's probably fine. I'll test stability with this tool and see how far I get, then I'll try it for a while and see if its stable in normal use

[Demonic Donut]

There is no one test that you can run to promise stability. I recommend using multiple stress tests to verify an OC. Only your use case will prove stability in that program. For example my aggressive GPU overclock is stable in everything I do/use except RDR2. CPU OCs can be the same.

 

Different instructions and programs put different loads on the system

 

Sometimes an OC can also seem perfectly stable until you have something that causes the load on the cpu to go up and down a lot.

 

Personally, I use Prime95, cinebench, blender and the programs I use day to day to test stability. There are a ton of stress tests you can do, and the best practice is to use all of them.

 

*Usually* if something is prime95 stable for over an hour, it'll be stable in 99% of situations.

 

Buildzoid (Actually Hardcore Overclocking) has a video about cpu stability and stress tests where he goes over it as well.