BEST AND SAFE OVERCLOCKING STABILTIY TEST ?

[INDIANAJUNE2]

hey guys , a serious thanks for any advice. I'm about to do my first over clock on my i7700k and id be lying if I wasn't nervous. I have the asus apex ix mother board that has a profile to OC to 5ghz for you so I figured id try that but I'm wondering what the best benchmark is to (as safe as possible) test for stability ? I hear prime 95 thrown around a lot but I heard that's REALLY hard on your cpu. was thinking about OCCT instead. not looking to break any records here just gaming and everyday use. I own 3d mark but I heard that's more for gpu but if it will work I know that well. also how long would you realistically run it for 24 hours seems extreme and unsafe. thanks again !

[Flying Sausages]

1) Run any of these for 6 hours min: Prime95/OCCT/Aida64/AsusRealBench/IntelBurnTest. Let say you ran Prime95 and Aida64, and you passed Prime95 and failed Aida64 doesn't mean your system is unstable.

 

2) Unigine Heaven and Valley Benchmark and 3dmark 11 mark 11. Cinebench R15 and R20.

 

3) Play all your games. If you have BF1 then this game can be use to test your oc because it is very sensitive to oc systems.

 

4) Use your computer normally

 

6) Put your computer to sleep for more than 2 hours and idle your system

 

 

 

[Labeled]

Don't be nervous, it's a lot harder to kill components than you'd think, I've pushed 1.52v(even got an error trying to go higher than my motherboard disagreed to boot) through my 7700k trying to get to 5.3Ghz, but I run daily at 5.125 with 1.465v.

As long as you have the cooling and don't set voltages to the red I'd say learn to do it yourself because the auto version might have a voltage higher than you need to be stable.

I prefer to do all my testing with gaming, realbench, and aida64. while prime95 and occt show you max heat, in my experience they're too unreal of a workload to bother.

I'll usually aim for a quick 10-20 minute aida run to see if it is just plain unstable, if it passes without crashing I'll play some csgo or other games, and if I don't crash while playing I'll run a realbench test while I sleep.

[Flying Sausages]

Just now, Labeled said:

Don't be nervous, it's a lot harder to kill components than you'd think, I've pushed 1.52v(even got an error trying to go higher than my motherboard disagreed to boot) through my 7700k trying to get to 5.3Ghz, but I run daily at 5.125 with 1.465v.

As long as you have the cooling and don't set voltages to the red I'd say learn to do it yourself because the auto version might have a voltage higher than you need to be stable.

I prefer to do all my testing with gaming, realbench, and aida64. while prime95 and occt show you max heat, in my experience they're too unreal of a workload to bother.

I'll usually aim for a quick 10-20 minute aida run to see if it is just plain unstable, if it passes without crashing I'll play some csgo or other games, and if I don't crash while playing I'll run a realbench test while I sleep.

Aida64 is unrealistic too. The only stress test is realistic I know is AsusRealbench that uses Handbrake encoding to test your oc stability.

[Falkentyne]

22 minutes ago, INDIANAJUNE2 said:

hey guys , a serious thanks for any advice. I'm about to do my first over clock on my i7700k and id be lying if I wasn't nervous. I have the asus apex ix mother board that has a profile to OC to 5ghz for you so I figured id try that but I'm wondering what the best benchmark is to (as safe as possible) test for stability ? I hear prime 95 thrown around a lot but I heard that's REALLY hard on your cpu. was thinking about OCCT instead. not looking to break any records here just gaming and everyday use. I own 3d mark but I heard that's more for gpu but if it will work I know that well. also how long would you realistically run it for 24 hours seems extreme and unsafe. thanks again !

Best and safe?

1) Cinebench R20, 7200 seconds loop.

2) Realbench 2.56, four hour stress test.

3) Prime95 29.8 build 6, small FFT (AVX disabled option) <--SSE2 stability.

 

Keep Hwinfo64 sensors open and look for CPU Cache WHEA errors to appear.

[Labeled]

58 minutes ago, OlympicAssEater said:

Aida64 is unrealistic too. The only stress test is realistic I know is AsusRealbench that uses Handbrake encoding to test your oc stability.

it doesn't cause un-realistic temperatures in my experience usually 3-5c higher than my temps in csgo. That's also why I only run it for 10-20 minutes to get any obvious instability out if the way before moving on.

 

1 hour ago, INDIANAJUNE2 said:

how long would you realistically run it for 24 hours seems extreme and unsafe

Not particularly. I've been running my 7700k at 1.465v for the past 2ish years and haven't experienced any degradation yet. for the most part your pc will just crash and reboot before any damage is done.

 

In the end it comes down to what you're comfortable with, if a 2 hour stability test is enough to give you confidence in daily driving the OC then that's fine, you might blue screen along the way, but you can always just nudge the voltage up .005 any time you blue screen. The people that test for 24hrs+ are mostly just the types that want to have proof behind their OC claims. If you don't mind some random blue screens on the way to stability you can just get it to a point where it passes a quick stability test and just use it like normal until you bsod and up the voltage ever so slightly until it stops happening.

 

 

There is no RIGHT way to OC, how ever you want to do it is fine. Everyone is going to tell you something slightly different.

[Flying Sausages]

Just now, Labeled said:

it doesn't cause un-realistic temperatures in my experience usually 3-5c higher than my temps in csgo. That's also why I only run it for 10-20 

It does on my end with Ryzen. Aida64 is the new Prime95. 

[Zberg]

While I admittedly do use Aida64 for an hour to test stability, it is unrealistic for what most people do.  For a more real world quick-load stress test, I will do multiple CB runs back to back while opening other things (many browsers, games, random programs) to see what happens.  Sometimes while CB is running in the background, sometimes in between runs.

 

The reason I do this is I have found there is a difference between when something is running full bore for a long time, and when continual blips of high stress are being put on it.  I have had weird set-ups in the past that didnt like certain things in games, certain programs simultaneously being opened, etc (even with higher ends of the LLC settings), but that always passed prime and aida64 for 2 hour runs without any issue.  Someone smarter than I probably knows why that is.  

 

So now I just run aida for an hour.  If it hasnt gotten above dangerous temps by then, literally nothing I do is going to load the cores up to 100% for an hour, I just dont have that much intensive tasks for my CPU to do.  Then I do some random testing by opening multiple of the programs that I use on the daily, with CB running now and then to see what happens.  Got me a very stable 5.1 Ghz on my 9900k with that method.

[Labeled]

2 minutes ago, Zberg said:

While I admittedly do use Aida64 for an hour to test stability, it is unrealistic for what most people do.  For a more real world quick-load stress test, I will do multiple CB runs back to back while opening other things (many browsers, games, random programs) to see what happens.  Sometimes while CB is running in the background, sometimes in between runs.

 

The reason I do this is I have found there is a difference between when something is running full bore for a long time, and when continual blips of high stress are being put on it.  I have had weird set-ups in the past that didnt like certain things in games, certain programs simultaneously being opened, etc (even with higher ends of the LLC settings), but that always passed prime and aida64 for 2 hour runs without any issue.  Someone smarter than I probably knows why that is.  

 

So now I just run aida for an hour.  If it hasnt gotten above dangerous temps by then, literally nothing I do is going to load the cores up to 100% for an hour, I just dont have that much intensive tasks for my CPU to do.  Then I do some random testing by opening multiple of the programs that I use on the daily, with CB running now and then to see what happens.  Got me a very stable 5.1 Ghz on my 9900k with that method.

This is why I like testing with csgo, while you do risk getting a competitive cooldown if it crashes too much, playing competitive always gets me a crash if it is unstable.

[INDIANAJUNE2]

3 hours ago, Caroline said:

Yup, I'd really avoid Auto OC unless you want to pump 1.55V to the CPU while playing a game and make it throttle like mad.

Overclocking **is** still a manual procedure no matter how much mobo manufacturers try to impose those, pardon my french, shitty tools with presets and nice looking graphics.

 

What about the rest of your components? Graphics? PSU?

 

Not really, if the cooling is good enough running P95 or OCCT for 24 hours isn't unsafe, the tests by themselves aren't unsafe, inappropiate settings and poor cooling is, if the CPU skyrockets to 99C it's not the test's fault, it's probably because of cooling or excessive voltage.

this was something I was thinking about as well. I really dont understand voltages though as this is my first time. I found this good how too vid with my cpu and mother board that looks legit enough but id still be taking this guys word for it. in the end I thought go with the asus profile not just becuse its easy , but I thought they would be more proven and stable becuse its a sold product but I have no info to back that up either. here's the vid I mentioned  

 

[INDIANAJUNE2]

3 hours ago, Labeled said:

it doesn't cause un-realistic temperatures in my experience usually 3-5c higher than my temps in csgo. That's also why I only run it for 10-20 minutes to get any obvious instability out if the way before moving on.

 

Not particularly. I've been running my 7700k at 1.465v for the past 2ish years and haven't experienced any degradation yet. for the most part your pc will just crash and reboot before any damage is done.

 

In the end it comes down to what you're comfortable with, if a 2 hour stability test is enough to give you confidence in daily driving the OC then that's fine, you might blue screen along the way, but you can always just nudge the voltage up .005 any time you blue screen. The people that test for 24hrs+ are mostly just the types that want to have proof behind their OC claims. If you don't mind some random blue screens on the way to stability you can just get it to a point where it passes a quick stability test and just use it like normal until you bsod and up the voltage ever so slightly until it stops happening.

 

 

There is no RIGHT way to OC, how ever you want to do it is fine. Everyone is going to tell you something slightly different.

one nice thing about my mainboard is it has a button that will reset the OC. so I'm waaaaay more worried about permanently damaging the cpu then just resetting and dropping it to 4.8ghz or whatever. I cant afforded to replace it rn, why im trying to be as safe as possible. I do understand its inherently unsafe though

[INDIANAJUNE2]

On 11/18/2019 at 12:25 PM, Caroline said:

der8auer is legit

 

Not all CPUs are the same though, some might work with 1.43V others with 1.53V and others might not even hit 5GHz, it's the lottery, not all CPUs have the same silicon quality and that's OK as long as they run properly at their default frequency.

The problem with presets and profiles is exactly that, you can't use the same settings for every chip out there so what they do is usually pump more vcore while maintaining low-medium LLC to deal with that.

In most cases you'll need the opposite, less vcore and a more aggresive LLC to keep it stable, it's an Intel so you'll probably need to tune IO and SA voltages as well. In the end it's all about trial and error, find the sweet spot of your particular chip. 5GHz vs 4.8GHz won't make any noticeable difference in real programs, if you need 1.47V to hit 5GHz and 1.39V to hit 4.8 I'd definitely go for those 4.8.

 

Don't worry about damaging the CPU as long as you use standard settings, if it fails to boot you just clear CMOS and try again.

You can only damage it by disabling its protections on purpose while overvolting it for extended periods (months of 24/7 usage probably, I've never tried tbh)

 

I'm more familiar with AMD chips so I won't be very helpful here.

 

nah , just this little bit of advice is helpful thank you. I wasn't aware there were still protections in place, thought overclocking was by definition 's circumventing those protections. that makes me feel better 

[Otto_iii]

On 11/17/2019 at 4:56 PM, OlympicAssEater said:

It does on my end with Ryzen. Aida64 is the new Prime95. 

I'm on Ryzen 3000, and running a (i suppose somewhat modest) 1.325V at 4.2ghz on 3600 (4175ccx0 4225ccx1) through ryzen master, with it minimized when in use (otherwise my voltages spike coming out of sleep states to 1.5V, even if i use BIOS) i have far lower temps, about 4-5C.  Power consumption seems the same tho

I get a very tiny single-core improvement in benchmarks this way, but multi-core is about 6-9% better, its no insane oldschool overclock but i suppose because the voltages are modest its a improvement without temps going through roof, somehow being lower  , i could push it harder but this was what i ended up with for stability.  Also if i had LLC control on this board i figure i could go another 50hz higher, no biggie but i do see mild voltage swings in HWMonitor that probably restrict things a little so far as how high i can go with frequency,

I'm gonna guess you are on ryzen 1000 or 2000? where all core voltages could safely be pushed further?  Notably in my case, im fairly sure i would see singlecore at same speed i have with this overclock if i could get a working 1.0.0.4 ABBA BIOS but its not out for every board yet, that said id be surprised if said overclocks like what im using still wouldnt be useful for bringing temps down