I won the the 8600K silicon lottery

[BigJohnson]

Hello from Australia,

 

I am brand spanking new to the LTT community. I have been watching every video I can on YouTube and have been having a great time doing so. After seeing how much voltage and heat people need just to get to 5Ghz on their 8600K's I feel I have truly won the silicon lottery. I have been creeping up to 5Ghz the last couple of days expecting my temps to skyrocket at some point. To my surprise this has not happened yet. I have just done a quick 1 hour stress test in RealBench to demonstrate the clock and temps. I know its only 1 hour, but with these temps and voltage who really cares, there is so much room to add more voltage if need be. Ambient temp is 20 degrees. At 1 hour my temps did not pass 62 degrees. I was browsing and typing this while the stress test was running so to me that is pretty stable, but only time will tell. I can't take any credit, this one was all Intel. I feel very lucky.

 

Settings:

Vcore: 1.21v adaptive

Clock: 5.0Ghz

Intel Enhanced Turbo: Off

Ratio mode: Variable

EIST: On

C-State: Auto

Cache voltage: Auto

Cache: 4.1Ghz

Idle temp: 29 Degrees

Max temp: 62 Degrees

 

 

PC Specs:

Motherboard: MSI Z-370 Sli Plus

Ram: Corsair 16gb @ 2666mhz XMP

Video Card: Galax GTX 1080 

CPU: 8600K

Cooler: H100i 4 years old

Case: Cougar Panzer-G

 

Thanks for reading,

 

Johnson.

 

 

 

[WereCat]

That's very impressive if it's truly stable.

You can most likely do 5.2GHz... heck maybe even 5.3GHz.

Have you delided the CPU?

[Technomancer__]

wow that's good

[Jurrunio]

Try Prime95, though even if you have to up the voltage to 1.25v for stability it's plenty impressive. 5.2GHz should be doable.

[BigJohnson]

10 minutes ago, WereCat said:

That's very impressive if it's truly stable.

You can most likely do 5.2GHz... heck maybe even 5.3GHz.

Have you delided the CPU?

I haven't delided yet, but I have my conductonaut ready to go!

[Jurrunio]

1 minute ago, BigJohnson said:

I haven't delided yet, but I have my conductonaut ready to go!

at 62C you dont even need to delid.

[BigJohnson]

6 minutes ago, Jurrunio said:

Try Prime95, though even if you have to up the voltage to 1.25v for stability it's plenty impressive. 5.2GHz should be doable.

I will give it a run now and see what I get. What version do you recommend?

[Jurrunio]

3 minutes ago, BigJohnson said:

I will give it a run now and see what I get. What version do you recommend?

the newest one. Versions before 27 dont use AVX, avoid them.

 

To be more extreme, set the AVX offset to 0.

[BigJohnson]

5 minutes ago, Jurrunio said:

the newest one. Versions before 27 dont use AVX, avoid them.

 

To be more extreme, set the AVX offset to 0.

Ooo sounds fun! Thanks mate.

[Froody129]

29 minutes ago, Jurrunio said:

at 62C you dont even need to delid.

This is not about need. Hehehe

 

Awesome OC. Keep us up to date. With a delid and like 1.4V I'd be very interested in seeing where you end up

[WereCat]

57 minutes ago, Jurrunio said:

at 62C you dont even need to delid.

At 62°C because of 1.21V.

Once he goes to 1.4V the temp will skyrocket.

[O9B0666]

I'm scratching my head here, cuz with my 8700k i can get 5.2 stable, and run 5ghz 24/7....so the 8086 is the same?

[MageTank]

8 hours ago, WereCat said:

At 62°C because of 1.21V.

Once he goes to 1.4V the temp will skyrocket.

It's not even the 1.21v. His stress test is only pulling 90w on the package. If he were to throw small FFT Prime95 or Linpack at it, the temps would surely rise.

 

What I find most puzzling, is that this is supposedly without a delid, on a 240mm AIO, in a case that is completely surrounded by tempered glass. For this to be the result, he would need to live in an igloo. Either that, or he is suffering the same issue I had with my Coffeelake build, and his vcore reporting is wrong. My board required a beta BIOS update to get proper vcore readings. This isn't a specific vendor issue, as it happened for basically every board that I've tried, and others had the same issue on different brands as well: 

 

@BigJohnson Would you be able to run some performance benchmarks with that CPU? I'd like to compare some results with that chip, and see if it's able to hold it's clock without throttling in the background. If it's numbers add up, you just might have the worlds greatest 8600k.

 

Either way, I have a spare non-delidded 8700k sitting on my desk, I'll turn hyperthreading off and see if I can obtain similar results with my 280mm AIO. Curious to see if Realbench really has that low of a power draw. 

[BigJohnson]

14 hours ago, MageTank said:

It's not even the 1.21v. His stress test is only pulling 90w on the package. If he were to throw small FFT Prime95 or Linpack at it, the temps would surely rise.

 

What I find most puzzling, is that this is supposedly without a delid, on a 240mm AIO, in a case that is completely surrounded by tempered glass. For this to be the result, he would need to live in an igloo. Either that, or he is suffering the same issue I had with my Coffeelake build, and his vcore reporting is wrong. My board required a beta BIOS update to get proper vcore readings. This isn't a specific vendor issue, as it happened for basically every board that I've tried, and others had the same issue on different brands as well: 

 

@BigJohnson Would you be able to run some performance benchmarks with that CPU? I'd like to compare some results with that chip, and see if it's able to hold it's clock without throttling in the background. If it's numbers add up, you just might have the worlds greatest 8600k.

 

Either way, I have a spare non-delidded 8700k sitting on my desk, I'll turn hyperthreading off and see if I can obtain similar results with my 280mm AIO. Curious to see if Realbench really has that low of a power draw. 

 

 

Hi MageTank,

 

Absolutely, it would be my pleasure! I just ran CineBench R15.0 and get variable score of around 1243. If you have any suggestions of other programs you would like to see me run and I will be happy to give them a try. I was watching my core clock the whole time and I could not see any dips. However with the CPU demand of Cinebench meant that HWmonitor was slow to update so I could have missed them. This result in CineBench is pretty common for a 8600K @ 5Ghz so I've heard.

 

I just ran a 1/2 hour stress test in AIDA64 to give you another sensor feed to read through if you wanted to reference this current overclock to use in CineBench.

 

 

I had to bump my voltage up to 1.215v last night after crashing in Prime95 at 3 hours. Prime stressed my CPU more then any other program resulting in a max temp of 71 degrees. I truly loath that program and try not to use it. No other application can get my temps as high as what Prime95 can. I don't think that's a testament to the capability of the program. The concern for me is what it has to do in order to generate such conditions.

 

 

Corsair Panzer-G has AMAZING airflow. I have replaced the glass panels for mesh panels on the front and top of my case to aid the cooling.The stock fans have great performance with little to no noise at all. Not only that it looks amazing. I have a high static pressure fan at the bottom along with 2 high static pressure and 2 high CFM fans on my radiator.

 

I have really enjoyed everyone's input and comments.

 

Cheers big ears, all the way from Australia.

 

Johnson.

[MageTank]

1 hour ago, BigJohnson said:

 

 

Hi MageTank,

 

Absolutely, it would be my pleasure! I just ran CineBench R15.0 and get variable score of around 1243. If you have any suggestions of other programs you would like to see me run and I will be happy to give them a try. I was watching my core clock the whole time and I could not see any dips. However with the CPU demand of Cinebench meant that HWmonitor was slow to update so I could have missed them. This result in CineBench is pretty common for a 8600K @ 5Ghz so I've heard.

 

I just ran a 1/2 hour stress test in AIDA64 to give you another sensor feed to read through if you wanted to reference this current overclock to use in CineBench.

 

 

 

I had to bump my voltage up to 1.215v last night after crashing in Prime95 at 3 hours. Prime stressed my CPU more then any other program resulting in a max temp of 71 degrees. I truly loath that program and try not to use it. No other application can get my temps as high as what Prime95 can. I don't think that's a testament to the capability of the program. The concern for me is what it has to do in order to generate such conditions.

 

 

 

Corsair Panzer-G has AMAZING airflow. I have replaced the glass panels for mesh panels on the front and top of my case to aid the cooling.The stock fans have great performance with little to no noise at all. Not only that it looks amazing. I have a high static pressure fan at the bottom along with 2 high static pressure and 2 high CFM fans on my radiator.

 

I have really enjoyed everyone's input and comments.

 

Cheers big ears, all the way from Australia.

 

Johnson.

I gotta say, that is one very impressive chip. I didn't have enough thermal paste on hand to swap out my chip to my non-delidded sample, so I had to use my delidded 8700k with hyperthreading disabled, clocked at 5ghz core, 4.1 cache, 2666 C16-18-18-35-2 and 1.2v on my EVGA CLC 280, but this was my result: 

 

I absolutely cannot wait to see what yours looks like after delidding.

[done12many2]

8 minutes ago, MageTank said:

I gotta say, that is one very impressive chip. I didn't have enough thermal paste on hand to swap out my chip to my non-delidded sample, so I had to use my delidded 8700k with hyperthreading disabled, clocked at 5ghz core, 4.1 cache, 2666 C16-18-18-35-2 and 1.2v on my EVGA CLC 280, but this was my result: 

 

I absolutely cannot wait to see what yours looks like after delidding.

 

I suspect that disabling hyperthreading really helped with the clocks / low voltages?  I have experienced that in the past with many i7 chips.   Running them in the same manner as their i5 counterparts often resulted in better clocks at lower voltages.  

 

I've also found that running stock cache and weak memory helps keep the voltage low with higher clocks.  Less data across the package = less work being done, hence less stress and more perceived stability on the overclock.  Keep those CPU cycles employed with faster cache/RAM and that shit falls apart fast.

 

I see why you asked for a benchmark of some kind now. 

[MageTank]

5 minutes ago, done12many2 said:

 

I suspect that disabling hyperthreading really helped with the clocks / low voltages?  I have experienced that in the past with many i7 chips.   Running them in the same manner as their i5 counterparts often resulted in better clocks at lower voltages.  

 

I've also found that running stock cache and weak memory helps keep the voltage low with higher clocks.  Less data across the package = less work being done, hence less stress and more perceived stability on the overclock.  Keep those CPU cycles employed with faster cache/RAM and that shit falls apart fast.

 

I see why you asked for a benchmark of some kind now. 

That is very true. From my experience, Hyperthreading adds about an extra 15w in power consumption alone, which is going to impact thermals to some degree. There is also the additional cache that can add more to the package power consumption depending on the workload. In regards to memory, specific tertiary timings will have a huge impact on thermals depending on the instruction set. Performance of AVX workloads are highly dependent on specific tertiary timings, such as tRDWR(DD/DR) and (while not really a timing) RTL/IO-L offset. Having "loose" memory results in lower AVX performance, and lower thermals as a result. The tighter your memory, the higher your AVX performance, the higher the thermals. This is a result of AVX being more efficient when it has enough bandwidth to complete it's workloads. As I've stated before, higher thermals are not always indicative of poor cooling, it can also be a result of higher efficiency in terms of resource usage. 

 

Regardless, I am really impressed by this 8600k's low thermals, especially knowing that he is located in Australia, and how hot their climate is. Hopefully @BigJohnson can give us an update on the thermals after the delid, as they will no doubt be even better than what they already are now. 

[O9B0666]

3 hours ago, MageTank said:

I gotta say, that is one very impressive chip. I didn't have enough thermal paste on hand to swap out my chip to my non-delidded sample, so I had to use my delidded 8700k with hyperthreading disabled, clocked at 5ghz core, 4.1 cache, 2666 C16-18-18-35-2 and 1.2v on my EVGA CLC 280, but this was my result: 

 

I absolutely cannot wait to see what yours looks like after delidding.

Would you mind posting your overclocking settings for your 8700K or pm me them? The 1800 score ones 

[MageTank]

1 hour ago, O9B0666 said:

Would you mind posting your overclocking settings for your 8700K or pm me them? The 1800 score ones 

Certainly.

 

This one was 5.4ghz core, 4.9ghz cache at 1.44v, with memory at 3600 C14-14-14-28-1, 1.39v vDIMM, tRFC 270, tREFI 65535, and tertiary timings as tight as they could go. 

 

Cooling was an EVGA CLC 280, and it was delidded with Conductonaut on the die, in a Cooler Master Mastercase Pro 5. Board was an ASRock Z370 Fatality K6, with BIOS version 1.61, I think it was one of the beta BIOS releases. This particular run was done on a brand new install of Windows 10, with only 18 background processes running, so do keep that in mind when benching.

[BigJohnson]

3 hours ago, MageTank said:

That is very true. From my experience, Hyperthreading adds about an extra 15w in power consumption alone, which is going to impact thermals to some degree. There is also the additional cache that can add more to the package power consumption depending on the workload. In regards to memory, specific tertiary timings will have a huge impact on thermals depending on the instruction set. Performance of AVX workloads are highly dependent on specific tertiary timings, such as tRDWR(DD/DR) and (while not really a timing) RTL/IO-L offset. Having "loose" memory results in lower AVX performance, and lower thermals as a result. The tighter your memory, the higher your AVX performance, the higher the thermals. This is a result of AVX being more efficient when it has enough bandwidth to complete it's workloads. As I've stated before, higher thermals are not always indicative of poor cooling, it can also be a result of higher efficiency in terms of resource usage. 

 

Regardless, I am really impressed by this 8600k's low thermals, especially knowing that he is located in Australia, and how hot their climate is. Hopefully @BigJohnson can give us an update on the thermals after the delid, as they will no doubt be even better than what they already are now. 

Thanks MageTank,

 

It looks like you have won the lottery too! Your CPU has great temps with low voltage. Amazing temps in fact. That must be a very effective cooler. I would really like to know your O/C that produce those high CineBench results. Thank you heaps for doing that comparison. That is actually really interesting, and something I have always wanted to try my self.

 

That is very eye opening in regards to the memory having an effect on the heat and power consumption of the CPU. It makes perfect sense though. Higher bandwidth would mean a higher demand on the CPU to keep up, meaning more heat. I feel like I have hit the sweet spot for efficiency. I only play games on this computer, no real compiling or anything special so this should work perfect as memory speed has little to no affect on FPS

 

I forgot to mention that in the last picture I posted, I had the cache at 4.5Ghz, I'm slowly pushing that up but I suspect from other peoples overclocks that I wont get much more out of the cache.

 

Thanks to everyone who took the time to read or comment on this thread. I appreciate the feedback!

 

I will definitely keep posting with any progress I make after the delid.

 

Thanks again from down under!

 

1 hour ago, MageTank said:

Certainly.

 

This one was 5.4ghz core, 4.9ghz cache at 1.44v, with memory at 3600 C14-14-14-28-1, 1.39v vDIMM, tRFC 270, tREFI 65535, and tertiary timings as tight as they could go. 

 

Cooling was an EVGA CLC 280, and it was delidded with Conductonaut on the die, in a Cooler Master Mastercase Pro 5. Board was an ASRock Z370 Fatality K6, with BIOS version 1.61, I think it was one of the beta BIOS releases. This particular run was done on a brand new install of Windows 10, with only 18 background processes running, so do keep that in mind when benching.

Beat me to it.

Wowzer, 5.4Ghz! What temps did you hit with this O/C?

 

Thanks heaps for that MageTank. Very interesting!

[MageTank]

1 minute ago, BigJohnson said:

 

 

Beat me to it.

 

Wowzer, 5.4Ghz! What temps did you hit with this O/C?

 

Thanks heaps for that MageTank. Very interesting!

I didn't put that particular overclock under any real stress, as it was certainly not stable. It was enough to get a few runs in before I dialed it back down to my daily 5.2. Temps under Cinebench was near 60C I think, but I wasn't running any application to monitor the temps as I was worried they would impact the results. Going by what my CLC's LED showed me, it was high 50's, low 60's. 

 

6 minutes ago, BigJohnson said:

 

 

Thanks MageTank,

 

It looks like you have won the lottery too! Your CPU has great temps with low voltage. Amazing temps in fact. That must be a very effective cooler. I would really like to know your O/C that produce those CineBench results. Thank you heaps for doing that comparison. That is actually really interesting, and something I have always wanted to try my self.

 

That is very eye opening in regards to the memory having an effect on the heat and power consumption of the CPU. It makes perfect sense though. Higher bandwidth would mean a higher demand on the CPU to keep up, meaning more heat. I feel like I have hit the sweet spot for efficiency. I only play games on this computer, no real compiling or anything special so this should work perfect as memory speed has little to no affect on FPS

 

I forgot to mention that in the last picture I posted, I had the cache at 4.5Ghz, I'm slowly pushing that up but I suspect from other peoples overclocks that I wont get much more out of the cache.

 

Thanks to everyone who took the time to read or comment on this thread. I appreciate the feedback!

 

I will definitely keep posting with any progress I make after the delid.

 

Thanks again from down under!

 

Johnson.

I don't think my cooler has much to do with my thermals, as it's mostly my ambient temperature carrying these results. I keep my house pretty cold at 65F (18.3C) so it's pretty easy to achieve much lower temperatures, even with these AIO coolers. As for my settings, I was dialed in at 5ghz core, 4.1ghz cache, 1.2v fixed (static), and ram was 2666 C16-18-18-35-2. I do not know what you have set for your secondary or tertiary timings, so I left those on auto and let the motherboard train them.

 

I wouldn't give too much thought into my Cinebench score being slightly higher than yours, as the i7's have extra cache over the i5's. It's impossible to have a real apples to apples comparison. My current case is the Thermaltake Core p3, which is an open air test bench, with a fan mounted directly over top of my VRM. I no longer have my Mastercase Pro 5, though I still consider it my favorite case.

 

If you are looking for any information on memory overclocking, I have a guide in my signature that should assist with that. It's a free way to obtain additional performance out of your setup, and can offer a pretty significant boost in minimum framerates while gaming. I think after delidding, you might even be able to push that chip even higher, especially with that much thermal/voltage headroom. I really can't wait to see what that chip can do. 

[O9B0666]

1 hour ago, MageTank said:

Certainly.

This one was 5.4ghz core, 4.9ghz cache at 1.44v, with memory at 3600 C14-14-14-28-1, 1.39v vDIMM, tRFC 270, tREFI 65535, and tertiary timings as tight as they could go. 

 

Cooling was an EVGA CLC 280, and it was delidded with Conductonaut on the die, in a Cooler Master Mastercase Pro 5. Board was an ASRock Z370 Fatality K6, with BIOS version 1.61, I think it was one of the beta BIOS releases. This particular run was done on a brand new install of Windows 10, with only 18 background processes running, so do keep that in mind when benching.

What settings did you disable in bios as well? I'm thinking I'm clicking some off I don't have to or some not off I should or my hoards a pile.... But that's a great oc 

[MageTank]

13 minutes ago, O9B0666 said:

What settings did you disable in bios as well? I'm thinking I'm clicking some off I don't have to or some not off I should or my hoards a pile.... But that's a great oc 

The only thing I actively disabled was Spread Spectrum. I left the current and power limits in auto, as I didn't observe any throttling of any kind in my tests, but I suppose depending on the board, you might need to alter those options if you find yourself throttling despite having thermals under control.

[O9B0666]

Throttling isnt it... I was wondering about the c states... I get 5ghz with 1.287 stable but trying to make sure I'm ticking and Un ticking the correct things in bios for a optimized 0c. I never oc a cpu before and couldn't find much on the settings to mess with in bios for my board. Wondering if I should get a different board but not sure what's a good over clocker board. 

[BigJohnson]

4 hours ago, MageTank said:

I didn't put that particular overclock under any real stress, as it was certainly not stable. It was enough to get a few runs in before I dialed it back down to my daily 5.2. Temps under Cinebench was near 60C I think, but I wasn't running any application to monitor the temps as I was worried they would impact the results. Going by what my CLC's LED showed me, it was high 50's, low 60's. 

 

I don't think my cooler has much to do with my thermals, as it's mostly my ambient temperature carrying these results. I keep my house pretty cold at 65F (18.3C) so it's pretty easy to achieve much lower temperatures, even with these AIO coolers. As for my settings, I was dialed in at 5ghz core, 4.1ghz cache, 1.2v fixed (static), and ram was 2666 C16-18-18-35-2. I do not know what you have set for your secondary or tertiary timings, so I left those on auto and let the motherboard train them.

 

I wouldn't give too much thought into my Cinebench score being slightly higher than yours, as the i7's have extra cache over the i5's. It's impossible to have a real apples to apples comparison. My current case is the Thermaltake Core p3, which is an open air test bench, with a fan mounted directly over top of my VRM. I no longer have my Mastercase Pro 5, though I still consider it my favorite case.

 

If you are looking for any information on memory overclocking, I have a guide in my signature that should assist with that. It's a free way to obtain additional performance out of your setup, and can offer a pretty significant boost in minimum framerates while gaming. I think after delidding, you might even be able to push that chip even higher, especially with that much thermal/voltage headroom. I really can't wait to see what that chip can do. 

Cheers for the guide there MageTank, I will definitely give it a look at. I can't wait either! Now that I am so happy with my chip I am starting to get scared about the delid and conductonaut.

 

But we have an old saying here in Australia, you gotta risk it for the biscuit!