Are 5.0GHz+ overclocks on Coffee Lake really typical?

[IllogicalGlory]

You can see from my signature that I have an i7-8700K, you can also see that I'm claiming to have hit 4.9GHz, but this isn't true. Yes, 4.9GHz at around 1.35V is stable for 10-20 minutes of OCCT or prime95 (actually, I've never seen it crash in a stress test), however it's not even close to stable once I start actually using the computer. In fact, I can't even get 4.7GHz with 1.35V stable.

4.7GHz is relatively stable for a few hours, but playing games like Battlefield 1, I experience crashes and other errors. Last night I saw the screen go white for a few seconds and then the textures stopped loading properly, which was strange. Chrome tabs crash, my FreeBSD VM notifies me of a watchdog timer error, the Final Fantasy XV demo crashes after about 30 minutes, Discord crashes in the background, etc. I don't get these on lower CPU clocks, so I'm almost certain the CPU is to blame.

HWMonitor shows voltage holding pretty steady at about 1.33-1.35V. My temperatures are also very reasonable, holding at around the high 60s to low 70s. I'm using an Asrock K6 Fatality motherboard with LLC at level 2.

All said, I'm stuck at 4.6GHz and really hoping that that's fully stable.

Have I got the worst 8700K ever? Are those of you out there with 5.1GHz overclocks actually able to sustain these clocks 24/7 while stressing the CPU with realistic workloads?

 

[moonbogg]

We have similar GPU's. Mine was stable for almost a full year at 2114mhz core. I started getting crashes and it was driving me insane trying to trouble shoot it. The GPU was the culprit because, over time, it had settled to a lower stable clock. Now its at 2088mhz and games no longer crash. I had no reason to consider it, but that's what it was.
What's your SA voltage at to stabilize that RAM speed of yours?

 

[frozentundra123456]

You can see from my signature that I have an i7-8700K, you can also see that I'm claiming to have hit 4.9GHz, but this isn't true. Yes, 4.9GHz at around 1.35V is stable for 10-20 minutes of OCCT or prime95 (actually, I've never seen it crash in a stress test), however it's not even close to stable once I start actually using the computer. In fact, I can't even get 4.7GHz with 1.35V stable.

4.7GHz is relatively stable for a few hours, but playing games like Battlefield 1, I experience crashes and other errors. Last night I saw the screen go white for a few seconds and then the textures stopped loading properly, which was strange. Chrome tabs crash, my FreeBSD VM notifies me of a watchdog timer error, the Final Fantasy XV demo crashes after about 30 minutes, Discord crashes in the background, etc. I don't get these on lower CPU clocks, so I'm almost certain the CPU is to blame.

HWMonitor shows voltage holding pretty steady at about 1.33-1.35V. My temperatures are also very reasonable, holding at around the high 60s to low 70s. I'm using an Asrock K6 Fatality motherboard with LLC at level 2.

All said, I'm stuck at 4.6GHz and really hoping that that's fully stable.

Have I got the worst 8700K ever? Are those of you out there with 5.1GHz overclocks actually able to sustain these clocks 24/7 while stressing the CPU with realistic workloads?

I have no personal experience, but silicon lottery says 99% of 8700k were able to hit 4.9 ghz. They claim to do a rigorous stress test. But I think their results are for delidded processers. So your value seems low. Have you tried more voltage, and are you using an AVX offset?

 

[Kenmitch]

Disclaimer: I pretty much lost interest in overclocking around the 4790k era and currently just use a laptop.

Maybe you just need to play around with other voltages beyond the vcore. I'll assume you've already looked at some guides already. Maybe take another look and see if you missed something. Have to tried lowering your memory clocks and timings to see if it effects your stability?

 

[IllogicalGlory]

I have lowered the memory clocks a touch. When I was running it at 4.7GHz, the memory was at 3466, obviously that's not a huge amount. Maybe I'll take it down to 3200. The timings aren't particularly tight. I don't experience any problems at 4.6GHz with 3600.

I haven't adjusted any other voltages so far, other than Vcore, as I kind of assumed that it would only be necessary to do so if I were targeting a high overclock (5GHz or so). I'd be happy with 4.8GHz at this point. Regardless, I'll do some investigation on that.

@moonbogg I'm just running the XMP profile (right now I have it 3466) on this set, so 1.35V.

 

[moonbogg]

I have lowered the memory clocks a touch. When I was running it at 4.7GHz, the memory was at 3466, obviously that's not a huge amount. Maybe I'll take it down to 3200. The timings aren't particularly tight. I don't experience any problems at 4.6GHz with 3600.

I haven't adjusted any other voltages so far, other than Vcore, as I kind of assumed that it would only be necessary to do so if I were targeting a high overclock (5GHz or so). I'd be happy with 4.8GHz at this point. Regardless, I'll do some investigation on that.

@moonbogg I'm just running the XMP profile (right now I have it 3466) on this set, so 1.35V.

System agent voltage needs to be looked at. If using fast memory and XMP, it will auto to a very high voltage usually. This can be high enough to actually wreck your CPU. If you aren't already familiar with what I'm talking about, then don't ignore or discount this warning. Research it as it might save your CPU. This is different from vcore. Its the system agent voltage, not the RAM voltage. RAM voltage being sent to the actual RAM sticks is 1.35 and that's fine. System agent voltage is a CPU voltage that lets the CPU handle the RAM speed. High RAM speeds are advertised by the motherboard makers to compete with eachother. They will slam your CPU with enough SA voltage to kill the damn thing just so they can claim that their board supports really fast RAM. Watch it.

 

[StinkyPinky]

Would be very surprised if you can't get 4.7 stable on a 8700k. Something is not right.

 

[Excessi0n]

Are you setting the cache to the same frequency as the cores? The cache on my 6700K caps out a couple hundred megahertz below the cores.

The stock turbo on the 8700K goes up to 4.7, so all of the cores should be able to manage at least that...

 

[IllogicalGlory]

Are you setting the cache to the same frequency as the cores? The cache on my 6700K caps out a couple hundred megahertz below the cores.

The stock turbo on the 8700K goes up to 4.7, so all of the cores should be able to manage at least that...

You would think so; I even got my 4790K to run at 4.7GHz 24/7. Of course, that was only considered average as well.

I have the cache frequency at 3.7 right now to isolate the CPU overclock. However, I have noticed that increasing it causes some interesting effects. When I put it at 4.2GHz, I scrolling on web pages stops working correctly, and some browser applications don't work correctly either. It's fine at anything below that. I know that sounds like a non-sequitur, but I'm pretty confident that it's true.

Like I said, it seems stable in stress tests at higher frequencies, but fails in actual usage, and the CPU isn't necessarily even being hammered in these cases either (50-80% CPU usage).

Also, it fails even with a -2 AVX offset.

@moonbogg The BIOS tells me it's 1.25V; I have it on the auto setting. HWmonitor shows the offset as being 0V. This looks like it's pretty high based on the research I've done. Should I try reducing it?

 

[moonbogg]

You would think so; I even got my 4790K to run at 4.7GHz 24/7. Of course, that was only considered average as well.

I have the cache frequency at 3.7 right now to isolate the CPU overclock. However, I have noticed that increasing it causes some interesting effects. When I put it at 4.2GHz, I scrolling on web pages stops working correctly, and some browser applications don't work correctly either. It's fine at anything below that. I know that sounds like a non-sequitur, but I'm pretty confident that it's true.

Like I said, it seems stable in stress tests at higher frequencies, but fails in actual usage, and the CPU isn't necessarily even being hammered in these cases either (50-80% CPU usage).

Also, it fails even with a -2 AVX offset.

@moonbogg The BIOS tells me it's 1.25V; I have it on the auto setting. HWmonitor shows the offset as being 0V. This looks like it's pretty high based on the research I've done. Should I try reducing it?

I'm no expert, but from what I've been advised and from what I've gathered, 1.25v is probably on the outside of what should be considered safe. If it were my chip, I wouldn't be happy unless I was comfortably under 1.2v. I have my own chip at 1.13. I know they are different chips, but I think the system agent is not as robust as the core itself on either of these CPU's. I would go under 1.2v or use 1.2v as a max, and whatever RAM speed you can get, well, it is what it is. Blame the motherboard companies for advertising RAM speeds that they knew would require some really high SA voltage to stabilize. It might work fine for 9 months or a year or something, then you get crashes. Then you have to add more and more voltage until the thing becomes unstable junk. It might be fine at 1.25v, but again if it were mine, I would lower it. Maybe you'll be fine though...maybe.

 

[Insomniator]

So why are you lying in your sig?

 

[justoh]

Running @ 5 without avx offset. Set the voltage to 1.26, but it goes up to 1.3, according to various monitors. I don't care enough to mess with llc or whatever I'd have to do to stop that. Seems stable after a few hours of non avx prime with a max temp of ~80. Games see about 55-60 max. X62

 

[IllogicalGlory]

So why are you lying in your sig?

Because I thought it was stable for a while and I haven't bothered to change it.

Speaking of which, I can't entirely remember if I did anything all that stressful with it at 4.9GHz, but I remember that Chrome would crash a lot during browsing. It didn't seem all that unstable, but it won't become entirely stable without adjusting it to 4.6.

It reminds me a bit of my i5-3570K; it did 4.2GHz fine the stock cooler and I couldn't push it any further even with water cooling. It's like I have a hard wall at 4.6.

Also I'm slightly optimistic that it might become true in the future.

I don't think I've done anything irresponsible with the chip; my case has decent airflow, and I've never set the voltage offset above +100mV. My temperatures have never exceeded the high 80s. I may have ran the computer with LLC at level 1 for a very short period of time (a few hours at most).

Is there a chance that the RAM is really having a major effect?

 

[Campy]

According to SiliconLottery "As of 12/01/17, the top 99% of tested 8700Ks were able to hit 4.9GHz or greater." However you might need 1.35-1.4 Vcore to achieve that, depending on your chip. Also using -2 AVX offset.

You should definitely be able to get over 4.6 if you want it.

 

[jpiniero]

Given that the single core turbo of the 8700K is 4.7, not being able to get full core turbo on 4.7 is rather unlucky. I'd guess that one of the cores has issues above 4.6.

 

[IEC]

I'm no expert, but from what I've been advised and from what I've gathered, 1.25v is probably on the outside of what should be considered safe. If it were my chip, I wouldn't be happy unless I was comfortably under 1.2v. I have my own chip at 1.13. I know they are different chips, but I think the system agent is not as robust as the core itself on either of these CPU's. I would go under 1.2v or use 1.2v as a max, and whatever RAM speed you can get, well, it is what it is. Blame the motherboard companies for advertising RAM speeds that they knew would require some really high SA voltage to stabilize. It might work fine for 9 months or a year or something, then you get crashes. Then you have to add more and more voltage until the thing becomes unstable junk. It might be fine at 1.25v, but again if it were mine, I would lower it. Maybe you'll be fine though...maybe.

That might be true on X99. On Z370 that doesn't seem to be the case. Using XMP for my G.Skill Samsung B-die @ 3600 CL15 gives a VCCIO of 1.2V and a VCCSA of 1.28V on a Z370 Taichi using the P1.30 UEFI. If my system spontaneously destabilizes in six months, I'll be sure to rant about it here.

I wouldn't go much beyond those voltages for 24/7 usage, but the sources I've seen seem to think < 1.3V is fine for air cooling.
https://overclocking.guide/gigabyte-z370-overclocking-coffee-lake/
https://overclocking.guide/rog-maximus-ix-x-apex-overclocking-guide/
https://www.tweaktown.com/guides/8481/coffee-lake-overclocking-guide/index5.html

 

[krumme]

Is it 100% stable at stock?

 

[IllogicalGlory]

Is it 100% stable at stock?

I don't know if it's really possible to be absolutely certain, but the main symptoms (Chrome tabs crashing, games crashing, watchdog errors, etc.) do go away if I decrease the clock to 4.6GHz. I have no reason to believe that it isn't.

 

[FIVR]

The obvious answer is that Silicon Lottery's "rigorous testing" probably involves getting the thing to post and that's about it. That would make sense, because it would take a lot of time and labor to actually test 1000s of CPUs with prime95, superPi, etc.


Most likely 99% of 8700ks will post at 4.9Ghz but many aren't stable, some (like OP's) aren't even stable at 4.7Ghz.

 

[Snarf Snarf]

Have you delidded your CPU? I just did an 8700k for a friend and he was having similar issues to you with shutdowns and crashes at around 4.8 GHz. After delidding, he's stable at 5.0 1.3v running around 70C at max load.

 

[LTC8K6]

SilconLottery guarantees stability during use. Posting at 5.0 is not stability.

This CPU is guaranteed stable when using the settings below and matching components from our QVL. We go through a rigorous stress test routine to ensure stability for the vast majority of use cases.

 

[IllogicalGlory]

Have you delidded your CPU? I just did an 8700k for a friend and he was having similar issues to you with shutdowns and crashes at around 4.8 GHz. After delidding, he's stable at 5.0 1.3v running around 70C at max load.

I haven't delidded it. I see that Silicon Lottery's binning is done on delidded CPUs, so that may explain why I'm having so much trouble getting similar results. I guess that's my only real option if I want higher clocks.

 

[Campy]

The obvious answer is that Silicon Lottery's "rigorous testing" probably involves getting the thing to post and that's about it.

That would be pants-on-head retarded, since they actually sell them with a guarantee to work at the stated clock and settings.

 

[Snarf Snarf]

I haven't delidded it. I see that Silicon Lottery's binning is done on delidded CPUs, so that may explain why I'm having so much trouble getting similar results. I guess that's my only real option if I want higher clocks.

Probably true, OCUK does the same thing. They delid in house and then do the binning process in order guarantee certain clocks.

I highly recommend it, the temperature gains when done properly are huge. Depending on where you live I would check and see if there' s a local hardware shop that will do it for you, here at my work we charge $49 for delid service, still cheaper than buying a Rockit tool and liquid metal TIM.

 

[moonbogg]

If 4.6ghz gets 80fps then 5.0 might get 86fps assuming perfect scaling with clock speed. Not really a big dead IMO but still. People want those chips for the YUGE CLOCKS.