I7 8700k overclock

[SpiralTTGL]

Seems a bit odd my core voltage at Idle vs cine bench load with my i7 8700k at 4.8ghz

 

 

 

 

Cpu is at 1% usage (Idle) at 1.36v

 

 

Cpu's running cinebench core voltage at 1.2V

 

 

Overclock is manual

 

All core  = 48

offset voltage at 10mv

MCE = off

 

Its odd my temps are at 35-40 idle and 75 max under full synthetic load

[.Apex.]

@SpiralTTGL
holy shit, you either won the Silicon lottery or that is a false reading...

[SpiralTTGL]

I download aida64 and ye its reporting 1.2v under cine bench load aswell and 1.36v at idle

This is under load cinebench load ^

 

 

 

idle ^

[.Apex.]

Just now, SpiralTTGL said:

I download aida64 and ye its reporting 1.2v under cine bench load aswell and 1.36v at idle

This is under load cinebench load ^

[.Apex.]

@SpiralTTGL
which motherboard are you using?

[SpiralTTGL]

asrock extreme 4

[.Apex.]

4 minutes ago, SpiralTTGL said:

asrock extreme 4

i think it's maybe because you're using Offset mode instead of Fixed mode, but anyway 4.8ghz at 1.2v is insanely good.. you can probably even go lower than that  

[SpiralTTGL]

2 minutes ago, syn2112 said:

i think it's maybe because you're using Offset mode instead of Fixed mode, but anyway 4.8ghz at 1.2v is insanely good.. you can probably even go lower than that  

Thing im worried about is the 1.36v at idle dont understand why its so high compared to load

[.Apex.]

Just now, SpiralTTGL said:

Thing im worried about is the 1.36v at idle done understand why

because offset is just offsetting the stock voltage, if you set it to Fixed you'll get the voltage that you want.

[AvocadosGuac]

7 minutes ago, SpiralTTGL said:

Thing im worried about is the 1.36v at idle done understand why

Temps are low enough. Don't worry about it in that case.

 

What i would do is put a custom loop on that bitch and see how far she'll OC.

[SpiralTTGL]

For those wondering 1584 was my cine bench score @4.8ghz

[SpiralTTGL]

3 minutes ago, Jrock said:

Temps are low enough. Don't worry about it in that case.

 

What i would do is put a custom look on that bitch and see how far she'll OC.

Thanks Im currently using a mugen 5 rev b I might change to water or a bigger air cooler in the future

[_Aontaigh_]

Can you run Prime95 and see if it's stable and what the voltage is?

 

v26.6: http://www.mersenneforum.org/showthread.php?t=15504

 

[You can download the latest version instead of 26.6 if you to light the chip on fire....]

[Amaranth]

3 hours ago, _Aontaigh_ said:

Can you run Prime95 and see if it's stable and what the voltage is?

No need to use Prime, Realbench or AIDA64 would be better since neither of them are power viruses and you're still getting an AVX load.

[SpiralTTGL]

Just ran aida64 voltage went up to 1.248 under full load

 

temps were higher bouncing between 70 and 90 avg 86

[_Aontaigh_]

8 hours ago, Amaranth said:

No need to use Prime, Realbench or AIDA64 would be better since neither of them are power viruses and you're still getting an AVX load.

I've had overclocks that could run AIDA64 and ASUS RealBench like it was nothing for a day but Prime95 v26.6 (Non-AVX) will die after less than five minutes....

 

If you want true stability: Prime95 v26.6 (Non-AVX)

If you want to burn your house down: Prime95 (Latest Version - w/ AVX)

If you want to squeeze every last drop out of your overclock with dodgy stability for your e-peen: AIDA64 and RealBench

[SpiralTTGL]

prime95 seems like an unrealistic load

[DarkSmith2]

24 minutes ago, SpiralTTGL said:

prime95 seems like an unrealistic load

actually not. 

 

This is what misinformed people spread. Infact prime95 26,6 load is pretty similar to Cinebench which is actually the same kind of load you would have in a real world rendering applications. Rumors started from people that where mad that they couldnt achieve stability, and further led into false impressions after the introduction of AVX on top of it.

 

Stable remains objectivly. There isnt anything like 100% stability on Desktop plattforms anyway, so people just opted for the "least resistance" quite normal human behavior. If you really would want to gain highest stability possible you for sure would use prime95 or Linx.

 

But i disagree that benching 26,6 with 1344 means "enough stable". The real stability test to be made is a full prime95 run. (takes ~22hrs). If you are willing to spare some time you could rather try to test with 8FFT's or 80FFT's for a more reliable hint. Just because 1344 doesnt generate enough heat to come close to some real world applications. And heat matters for stability.

 

[_Aontaigh_]

29 minutes ago, SpiralTTGL said:

prime95 seems like an unrealistic load

 

It is when compared to most day to day tasks, meaning you can disregard the temperatures -kind of-. Though, of course, that doesn't give one a license to let the cores sit at 95°C.

 

It's purpose is to test stability and nothing else can come close to beating it.

[Amaranth]

4 hours ago, _Aontaigh_ said:

I've had overclocks that could run AIDA64 and ASUS RealBench like it was nothing for a day but Prime95 v26.6 (Non-AVX) will die after less than five minutes....

 

And I've seen overclocks pass Prime and fail under other loads, no program is going to catch everything and you should never use just one for testing. Personally I've found that there isn't enough of an advantage to Prime to make it worthwhile, especially given how it was known to damage hardware in the past, and that Realbench and AIDA64 are enough to achieve a stable overclock and are closer to real-world loads.

[ViBEE]

4 hours ago, DarkSmith2 said:

actually not. 

 

This is what misinformed people spread. Infact prime95 26,6 load is pretty similar to Cinebench which is actually the same kind of load you would have in a real world rendering applications. Rumors started from people that where mad that they couldnt achieve stability, and further led into false impressions after the introduction of AVX on top of it.

 

Stable remains objectivly. There isnt anything like 100% stability on Desktop plattforms anyway, so people just opted for the "least resistance" quite normal human behavior. If you really would want to gain highest stability possible you for sure would use prime95 or Linx.

 

But i disagree that benching 26,6 with 1344 means "enough stable". The real stability test to be made is a full prime95 run. (takes ~22hrs). If you are willing to spare some time you could rather try to test with 8FFT's or 80FFT's for a more reliable hint. Just because 1344 doesnt generate enough heat to come close to some real world applications. And heat matters for stability.

 

I usually run Prime95 29.4 - Maximum Heat and Blend for like 15 minutes each and I dont have any problems, after that I do some video rendering 1440p/60fps/130k mbps. 

[Amaranth]

4 hours ago, DarkSmith2 said:

 Rumors started from people that where mad that they couldnt achieve stability...

Actually it was more that Prime95 damaged hardware, for some reason people don't like it when a program kills or degrades their stuff. That's the same reason Furmark isn't generally recommended outside of thermal testing.

[DarkSmith2]

16 hours ago, Amaranth said:

Actually it was more that Prime95 damaged hardware, for some reason people don't like it when a program kills or degrades their stuff. That's the same reason Furmark isn't generally recommended outside of thermal testing.

the software itself cant and couldnt ever damage hardware, bad cooling solutions do.

[Amaranth]

7 hours ago, DarkSmith2 said:

the software itself cant and couldnt ever damage hardware...

Of course it's not the software's fault despite Prime being the only program that regularly had issues.

 

But it's clear you've made up your mind and think that Prime95 is great to the exclusion of any other line of thought, thus making any more conversation on the subject pointless.

[DarkSmith2]

4 hours ago, Amaranth said:

Of course it's not the software's fault despite Prime being the only program that regularly had issues.

 

But it's clear you've made up your mind and think that Prime95 is great to the exclusion of any other line of thought, thus making any more conversation on the subject pointless.

Na, i just think its not worse than any other stability testing tool. Seeing professional overclockers preferring it to save some time, so im not seeing any reason to not personally use it to test stability on 24/7 overclocks. And i really havnt seen any so called "issues" with prime95. Only thing ive seen was people saying to be careful with it in relation to CPU's with FIVR for example, and ofc poeple not knowing how to use it properly. 

 

But to be honest when i play and stream at the same time 720p60 x264 slow preset my CPU hits 20°C hotter temperatures than testing stability in prime 26,6 1344k's... If you compare Aida, Realbench or alot of others they do less stress on the CPU than alot of actual real world applications thermal wise. And with the latest chips thermals matter alot for stability. You can stress test with those when you only play extremely GPU-bound games or having alot of time to really do your stresstesting many many hours.

 

If i try to find a sweetspot overclock for a CPU, i start testing with cinebench, you know how it goes, f.e. you set a VCore and increase core clocks until cinebench crashes and then do finetuning, in first real stability tests prime95 26.6 1344k's comes in pretty handy and takes only about 1hour to do.

After getting rid of any errors including WHEA errors you can simply put another customrun with said 80k/80k small FFT's to test your stability and cooling solution once and for all in another hour. If you really want to push things you can use 128k/128k though, (would maybe use it for custom loops).