Worst 8700k ever?

[ThreePeenSoup]

I thought I had 4.7 stable at 1.26v but I was wrong. In fact, my 8700k can't do 4.5ghz at 1.26v....so wtf? Is this normal? I have it at 4.4ghz again 1.26v. Pretty annoying.

 

8700k 4.4ghz All core, 1.26v (1.280 full load) LLC 6. Backed off from 4.7 all core 1.26v LLC6 because it almost crashed my PC in GTAV going around a corner, major cpu hitch. I had tested with OCCT large, medium data and aida no FPU for 2 hours each with no issues. Can't run any stress tests like aida with everything checked, prime, occt linpack, or realbench due to temps. Realbench crashes even with the stock settings 4.4ghz 1.25v auto everything else, so I don't know.

Maximus X Formula (latest BIOS 2102)
EVGA GQ 1000watt 80plus gold
3000mhz tridentz 16-18-18-38 1.35v
VCCIO 1.1
VCCSA 1.1
SVID Disabled
BCLK Adaptive Voltage Disabled
Ring Down Bin Auto (x41)
VRM Spectrum Disabled
BCLK Spectrum Disabled (Just disabled it. I think I notice a slight increase in performance, could be placebo)
C-States Auto
Speedstep Disabled
Long duration power limit 4095
Short duration power limit 4095
Something I can't remember 255.75d
Samsung 970 Evo 500gb
Toshiba 7200 RPM + 16gb intel optane (couldn't afford another NVMe M.2 1Tb =/)

Right now I'm looking for PCIE spread spectrum, MCP PCIE spread spectrum, SATA spread spectrum, and LDT spread spectrum. I found this guide from 2008 so I'm not sure if these things even apply anymore, but the details of what they all do are informative. I might be missing some setting that's relevant up there. All the power increase options are still on auto and the VRMS temp throttling option is enabled. They never go above 63c or so in hwinfo. I'm either missing some crucial setting or I have a really bad chip. Can't even do 4.5 at 1.26v and a LLC that puts it at 1.28v under full load. Crashes in rebel galaxy outlaw of all games.

[TheKDub]

Looks like you lost the lottery. I'm doing 4.8GHz on 1.2v with my 8700K right now, and can go as high as 5.2GHz on all cores with only a little over 1.3v. Even managed 5.3GHz on a single core.

[PopsicleHustler]

You got unlucky with silicon lottery.  That sucks.

[ThreePeenSoup]

Great.  Well, so far after some testing I think I have it stable at 4.5ghz 1.26v after disabling some other BIOS power saving options.  I had speed shift still enabled as well as cstates.  I'll find out for sure after some gaming.  

Thanks for replies.

[CPUguy101]

44 minutes ago, TheNoid said:

Great.  Well, so far after some testing I think I have it stable at 4.5ghz 1.26v after disabling some other BIOS power saving options.  I had speed shift still enabled as well as cstates.  I'll find out for sure after some gaming.  

Thanks for replies.

How long have you had it? Perhaps exchange it 

[ThreePeenSoup]

1 hour ago, SMBGUY said:

How long have you had it? Perhaps exchange it 

About a year.  I have the intel tuning protection so if it gets all borked I can get a new one for free. Any reason FPS would drop in game after an overclock?

[Boyohan]

Dude... Crank up the voltage. Anything below 1.4V is fine. If you have having heat issues do a delid. I dropped almost 20C on my 8600k after the delid. Delidding also helps with overclocking as it becomes more stable with lower temps. I am also a loser in the silicon lottery with my 8600k. I could not get anything over 4,7ghz at 1.37V. After delid I am running 5.0ghz @ 1,36V. I tested with Prime95 8h and Realbench 4h and max temp was 67C. Before the delid i was at 80-85C at 4.7ghz. 

[Boyohan]

18 minutes ago, TheNoid said:

About a year.  I have the intel tuning protection so if it gets all borked I can get a new one for free. Any reason FPS would drop in game after an overclock?

Sudden FPS drops are a sign of an unstable OC if either CPU or GPU depending on what you have done to your system.

[ThreePeenSoup]

1 hour ago, Boyohan said:

Sudden FPS drops are a sign of an unstable OC if either CPU or GPU depending on what you have done to your system.

Well after a 2 hour OCCT large data run the voltage only averages about 1.274v I never run at %100 for that long.  I don't even think I have a game or a program that does other than the stress tests.  The FPS stutter I was seeing went away after I turned off Cstates, intel speedshift, and disabled the BCLK spectrum.  The lower FPS was just on certain BFV maps that obviously have more buildings and clutter that gets collapsed or otherwise exploded.  My settings are all on Ultra except the post processing effects and TAA is low.  It actually runs above ~150 ish unless something huge happens.  My GPU isn't overclocked further than the stock GPU boost 3.0 speeds.  I have the voltage of my 1080FTW2 at %100 and power %120 priority in precision XOC.  It runs 2025mhz 212watts max, around ~1900mhz 143watts average, rarely ever hits 68c when I'm gaming.  I checked the FPS and more importantly Frametime with XOC HW monitor and it's right on par with where it was before the core clock/voltage changes.

 

Also the menu lag and frame skips I was seeing in GTAV disappeared with the aforementioned settings turned off.  Everything runs smooth now so i'll see how it goes.  Also, my AIO is middle tier.  1.4v would melt my CPU in half a skosh.  Not even one full skosh.  Delid is on the list of things.  Voids warranty though so i'm iffy.  If a new AIO doesn't fix the temperature issue then I will.  I'll probably get the corsairh100i or whatever the 240mm version is like i should've done.

Thank you for your reply 

[Mister Woof]

Any reason you need to stick to such low voltage?

 

1.35-1.4v is generally the minimum for a 5ghz OC, and 4.7 generally with MCE will start you off at 1.35v anyway

[ThreePeenSoup]

21 minutes ago, Plutosaurus said:

Any reason you need to stick to such low voltage?

 

1.35-1.4v is generally the minimum for a 5ghz OC, and 4.7 generally with MCE will start you off at 1.35v anyway

Well, my AIO isn't great, thermal take floe riing 240mm rgb.  Fan's spin at ~1600rpm pump is at ~1800rpm even though it's advertised at ~3600rpm.  It's ventilated fine, just reapplied MX4 and saw a big drop in temps because there was a bald spot right in the middle of the IHS.  Previously, if I run 1.27v the temps skyrocket because the LLC 6 Hits 1.316v.  LLC5 undervolts to where even 4.6ghz isn't stable so there's now way 4.7 will.  I basically can't use any stress tests thermal testing.  Prime95 instant tjmax practically, aida64 CPU+FPU+CACHE+RAM almost does the same thing.  OCCT Linpack 2019 %90memory goes for about 45seconds then something happens and it heats up to my 87c shut down point.  I can run benchmarks, OCCT medium/large data sets on auto, AIda64 CPU+RAM+CACHE, and realbench stress for about 3 minutes until I shut it down at 88c.

It wouldn't even run all 16gb of my ram until I upped my pagefile to 8gb min 16gb max.  

I read all these posts, "Yea man I'm running 98ghz all core on 1.15v, 37c under load"  Really man.  Where are these chips and how do I get one.  Honestly, I see 5.1ghz on 1.27v delidded and I cry a bit.  I lost the lottery I think, just like the first few replies said =/

EDIT:  I just upped to 4.6ghz 1.27v.  What I don't understand is why the manual vcore setting can be 1.26v or 1.27v and it still only draws 1.28v max.  It's the same thing for 4.4ghz, pulls the same max voltage 1.28v.

[Mister Woof]

Serious question: what do you actually use your computer for?

 

If you use your computer for the sole purpose of stress testing then yes it is important to make sure it can handle that load, but if you only play games or do specific tasks, it just needs basic stability and to be stable in the tasks you actually perform.

 

I personally feel a lot of these "stress test milestones" are totally arbitrary and over used.

 

It's like people who do torture tests on cars and guns....yes let's throw cement into the chamber or run the engine without oil! If it fails it's trash! Things we'll never ever do in real life.

 

If it survives an hour or two of RealBench within acceptable limits and is stable for my actual uses, it's good to go imo.

[ThreePeenSoup]

7 minutes ago, Plutosaurus said:

Serious question: what do you actually use your computer for?

 

If you use your computer for the sole purpose of stress testing then yes it is important to make sure it can handle that load, but if you only play games or do specific tasks, it just needs basic stability and to be stable in the tasks you actually perform.

 

I personally feel a lot of these "stress test milestones" are totally arbitrary and over used.

 

It's like people who do torture tests on cars and guns....yes let's throw cement into the chamber or run the engine without oil! If it fails it's trash! Things we'll never ever do in real life.

 

If it survives an hour or two of RealBench within acceptable limits and is stable for my actual uses, it's good to go imo.

Yea I use it for gaming and netflix/youtube.  Nothing really else other than these forums and others trying to figure out my OC.  Do you generally want your average voltage to be under what the LLC is providing?  4.5ghz 1.264v pushes 1.28v sometimes, averages 1.271v.  The 4.6ghz 1.27v pulls a constant 1.271average despite the current vcore reading 1.264v, same max 1.280v.  As I test, the average voltage rises to 1.279v  I shut it off at that point thinking it's going to need more vcore.  Is this a safe assumption?  If I put 1.28v it's going to hit 1.312v with this LLC and my aio can't handle that during a stress test, probably can handle during everyday use though and that's what i'm trying to figure out.  These major stress tests don't seem to apply to my everyday use.  The only time I run AVX is during stress testing.

[Mister Woof]

13 minutes ago, TheNoid said:

Yea I use it for gaming and netflix/youtube.  Nothing really else other than these forums and others trying to figure out my OC.  Do you generally want your average voltage to be under what the LLC is providing?  4.5ghz 1.264v pushes 1.28v sometimes, averages 1.271v.  The 4.6ghz 1.27v pulls a constant 1.271average despite the current vcore reading 1.264v, same max 1.280v.  As I test, the average voltage rises to 1.279v  I shut it off at that point thinking it's going to need more vcore.  Is this a safe assumption?  If I put 1.28v it's going to hit 1.312v with this LLC and my aio can't handle that during a stress test, probably can handle during everyday use though and that's what i'm trying to figure out.  These major stress tests don't seem to apply to my everyday use.  The only time I run AVX is during stress testing.

I presume LLC 6 is the maximum LLC setting?

 

LLC "guesses" how much vdroop you will have under load and compensate; this sometimes results in boosting above your specified target and generally speaking if you can maintain stability with higher fixed voltage and less aggressive LLC, you will have less variance in your system voltage and generally a more stable system. That said, from my personal experience, I notice voltage to be highest during idle states when there isn't much actual current going through it so that's not really meaningful. 

 

What I can tell you is that tests like OCCT and P95 are so ridiculously more demanding than your stated uses that you won't even come close to those numbers in real life.

 

I wouldn't be worried about going to 1.3v on your system for gaming and general use.

 

My 8700k will hit the high 80s in P95 AFTER a delid, but in real life it never goes above 65 even in demanding games.

[ThreePeenSoup]

1 hour ago, Plutosaurus said:

I presume LLC 6 is the maximum LLC setting?

 

LLC "guesses" how much vdroop you will have under load and compensate; this sometimes results in boosting above your specified target and generally speaking if you can maintain stability with higher fixed voltage and less aggressive LLC, you will have less variance in your system voltage and generally a more stable system. That said, from my personal experience, I notice voltage to be highest during idle states when there isn't much actual current going through it so that's not really meaningful. 

 

What I can tell you is that tests like OCCT and P95 are so ridiculously more demanding than your stated uses that you won't even come close to those numbers in real life.

 

I wouldn't be worried about going to 1.3v on your system for gaming and general use.

 

My 8700k will hit the high 80s in P95 AFTER a delid, but in real life it never goes above 65 even in demanding games.

Yea I'm not seeing much max temp difference between 4.5ghz 1.26v and 4.6ghz 1.28v 

So far i've run an hour of occt big data set and 5 realbench benchmark runs + cinebench one right after the other and the max temp was 86c which is an outlier like the min temp.  I see average temps of 68 - 76c 

 

Gonna throw a few games at it see what hitches.  Thank you for your help.

 

Hit 88c with cinebenchr20

[Mister Woof]

2 minutes ago, TheNoid said:

Yea I'm not seeing much max temp difference between 4.5ghz 1.26v and 4.6ghz 1.28v 

So far i've run an hour of occt big data set and 5 realbench benchmark runs + cinebench one right after the other and the max temp was 86c which is an outlier like the min temp.  I see average temps of 68 - 76c 

 

Gonna throw a few games at it see what hitches.  Thank you for your help.

Delid helped me, a lot.

[ThreePeenSoup]

12 minutes ago, Plutosaurus said:

Delid helped me, a lot.

Honestly $40 + shipping i'm assuming for a delidded and binned 8700k sounds like a good deal.  

[DarkSmith2]

2 hours ago, Plutosaurus said:

What I can tell you is that tests like OCCT and P95 are so ridiculously more demanding than your stated uses that you won't even come close to those numbers in real life.

Overclocking without reasonable stability testing is like cooking without tasting the food. 

If you think prime95 is stressing your CPU with unnecessary high and unrealistic loads, you just dont know how to use it properly.

59 minutes ago, TheNoid said:

Yea I'm not seeing much max temp difference between 4.5ghz 1.26v and 4.6ghz 1.28v 

So far i've run an hour of occt big data set and 5 realbench benchmark runs + cinebench one right after the other and the max temp was 86c which is an outlier like the min temp.  I see average temps of 68 - 76c 

 

Gonna throw a few games at it see what hitches.  Thank you for your help.

 

Hit 88c with cinebenchr20

You sir have a problem, and its not your CPU, its your cooler. No matter what you are currently using to cool your CPU its either undersized, not seated properly or defective. Hitting close to 90°C with a core voltage of 1.28v or below is ridiculous unless you have a 45°C+ room temperature, no matter what, even undelided.

[Mister Woof]

13 minutes ago, DarkSmith2 said:

Overclocking without reasonable stability testing is like cooking without tasting the food. 

If you think prime95 is stressing your CPU with unnecessary high and unrealistic loads, you just dont know how to use it properly.

You sir have a problem, and its not your CPU, its your cooler. No matter what you are currently using to cool your CPU its either undersized, not seated properly or defective. Hitting close to 90°C with a core voltage of 1.28v or below is ridiculous unless you have a 45°C+ room temperature, no matter what, even undelided.

"Reasonable stability testing" is totally arbitrary, IMO.

 

If you are doing scientific workloads or rendering 24/7, sure, something like P95 I would feel might be necessary.

 

P95 as a stress test for a gaming and YouTube rig though is totally unrealistic in my opinion.

 

But, feel free to get 100% 16 hour P95 stable for your Apex Legends / YouTube machine if you want to.

[DarkSmith2]

11 minutes ago, Plutosaurus said:

"Reasonable stability testing" is totally arbitrary, FYI.

 

If you are doing scientific workloads or rendering 24/7, sure, something like P95 I would feel might be necessary.

 

P95 as a stress test for a gaming and YouTube rig though is totally unrealistic in my opinion.

If you f.e. use the v26,6 of prime95 without avx2 and only check for core stability f.e. using a custom test with FFTs 1344 to 1344 ofc with small ffts in place checked, newer games wont even be stable enough and will crash. Its all about how you set prime95, you can adjust it for every usecase in existence. It doesnt get unreasonable only because you dont know how to use it properly specifically for those usecases. So your opinion doesnt matter because your lack of knowledge.

[Mister Woof]

4 minutes ago, DarkSmith2 said:

If you f.e. use the v26,6 of prime95 without avx2 and only check for core stability f.e. using a custom test with FTTs 1344 to 1344 ofc with small ffts in place checked, newer games wont even be stable enough and will crash. Its all about how you set prime95, you can adjust it for every usecase in existence. It doesnt get unreasonable only because you dont know how to use it properly specifically for those usecases. So your opinion doesnt matter because your lack of knowledge.

"Opinion"

 

Everyone's got one

 

My opinion is that everyone is obsessed with "stability" and spends more time obsessing about it than using their gear.

 

Like guys with PSA barrels worrying about how many rounds they can fire when they'll never even shoot that many and can't even afford the ammo anyway

 

For ME: any of the popular stress tests for a reasonable amount of time and your actual programs is enough.

 

Some people want specific criteria and others want completely ridiculous 24 hour tests..... That's up to them. This entire time I've given my opinion, and you're free to have your own.

[DarkSmith2]

2 minutes ago, Plutosaurus said:

"Opinion"

 

Everyone's got one

 

My opinion is that everyone is obsessed with "stability" and spends more time obsessing about it than using their gear.

 

Like guys with PSA barrels worrying about how many rounds they can fire when they'll never even shoot that many and can't even afford the ammo anyway

My opinion is that there are enough "ignorant people" weapon fetishist plattforms in the internet and that this has nothing to look for here, and that people are blaming games and software in general for "not working properly" when its their own fault because they want to overclock but dont want to spent the time to do it properly. 

[Mister Woof]

1 minute ago, DarkSmith2 said:

My opinion is that there are enough "ignorant people" weapon fetishist plattforms in the internet and that this has nothing to look for here, and that people are blaming games and software in general for "not working properly" when its their own fault because they want to overclock but dont want to spent the time to do it properly. 

It was an example.

 

You are very combative and I am unsure why you feel the need to belittle others for their opinions.

 

But, that's up to you. You seen unable to accept that other people have different opinions that might be valid.

[DarkSmith2]

50 minutes ago, Plutosaurus said:

It was an example.

 

You are very combative and I am unsure why you feel the need to belittle others for their opinions.

 

But, that's up to you. You seen unable to accept that other people have different opinions that might be valid.

Oh ill give you an example,

 

if we both sit infront of a banana, and your opinion is that this banana is "blue", which is obviously a false statement, your opinion disqualifies to be an opinion to me

because its so obviously wrong thats either a troll or that you need a doctor, but i couldnt care less, tbh.

[ThreePeenSoup]

 

11 minutes ago, DarkSmith2 said:

Oh ill give you an example,

 

if we both sit infront of a banana, and your opinion is that this banana is "blue", which is obviously a false statement, your opinion dequalifies to be an opinion to me

because its so obviously wrong thats either a troll or that you need a doctor, but i couldnt care less, tbh.

He was actually really helpful and I appreciate him taking the time to help me.  I don't know what your deal is though and "dequalifies" isn't a word.

But you obviously couldn't care less, tbh.