Did I just get a bad i5-8600k?

[Runscream]

First of all, first time poster and overclocker here. Here are my specs:

MB: Gigabyte Z370P D3
CPU: i5-8600k
CPU Cooler: Cyrorig H7
GPU: MSI GTX 1080 ARMOR 8G OC
RAM: HyperX Fury 2x8GB 2400mhz
PSU: Cooler Master G750M 80 PLUS Bronze

Now, there are two issues worrying me about my new OCed CPU:

1. Temperatures: at 4.5ghz, 1.35v, my i5-8600k peaks to 100c during Prime95 stress testing, then going down to around 80c (77c-83c). Then it peaks again for a while, like a cycle. This is the main reason I haven't run the test for more than 40 minutes, out of fear that the high temps will damage the CPU. I have also tested rendering an After Effects composition, since I know as a fact that this will be the most heavy use I'll realistically give to my CPU. The render lasted one hour and a half and temps never went beyond 85c, averaging at 75 the rest of the time. Then I checked the temps while running Heaven Benchmark and 3D Mark's Time Spy, assuming this would be representative of heavy gaming, which will be the most common use for this PC. Temps never went beyond 80c, averaging at 70 the rest of the time.

So, should I be worried about these temps? Specially the 100c I get during Prime95 stress testing. This is one of the main reasons why I don't up the voltage, which would allow me to get stable 4.7ghz.

2. Frequency: I had read that it was very easy to get 5ghz on these CPUs, so I was expecting to get at least stable 4.8ghz, but even at 1.45v, one core always fails (the same one every time), other than that, everything works well at this frequency. What am I risking if I decide to run at 4.8ghz in this scenario? Also, did I just get a really bad chip or should I assume these limitations are because of my MB and/or CPU cooler?

Btw, when I use EasyTune, I can get stable 4.7ghz with the voltage set on auto, but the stress test temps worried me, so I just set it to 1.35v in the BIOS and sticked to 4.5ghz, which is stable at this voltage.

I look forward to learning a lot here. Thanks for your time

EDIT: forgot to add that this is the first time I have installed an aftermarket CPU cooler and I'm worried the bad temps are because I did it wrong (didn't do the x pattern thing).

 

[EarthDog]

Its a $35 air cooler, I wouldn't expect that much out of them at 1.35V...

I wouldn't worry too much, no. Consider using an AVX offset, -3 or -4, since your workloads don't appear to use those instruction sets. That will keep temps down in P95.

 

[diaz]

You could try remounting it and see if your results improve.. tim application plays a part, but also important is the mechanical mounting stability of the cooler. Sometimes the mounting brackets are slightly loose or something similar happens. The X-pattern tightening of the brackets will also help spread the tim more evenly.. I wouldn't expect a very substantial difference in temperatures if you get an optimal mount, but it may yield 2-8C if there was an issue.

 

[trents]

Something seems wrong also with your settings. Do you really need 1.35 vcore to only get to 4.5 ghz ? Tell us what you changed in bios.

How did you go about overclocking? Did you start by setting the vcore to 1.35? You should do overclocking in increments. Start with stock voltage and start increasing the core multiplier by 1x increments. Stress test after each increment of increase by running Prime95 blend for 20 minutes. Monitor for temps whenever you stress test. When you can no longer pass the 20 minute stress test give the vcore a .01 increase and restest. Etc. When you think you have topped out run a 2 hr. Prime95 stress test. If you pass it, you will likely be stable in real life computing. If you fail it, either increase the vcore or decrease the multiplier.

The Coffee Lake CPUs run hot when overclocked. Many people are "delidding" them. Are you familiar with that? Putting liquid metal TIM under the lid will typically lower temps 10-20c. This would likely render your present cooler adequate.

Your motherboard may be part of the problem as it has only 4+2 power phase. Kind of an entry level Z chipset motherboard. And having the vcore at 1.35 is not helping that. Your VRM may be overheating.

As others have said, you can set the AVX offset to -2 and that should help. Recent versions of Prime95 use AVX 2 instructions set which produce a terrific amount of heat and stress on a CPU. By using the AVX offset you can allow the CPU to run at 200 mhz lower clock speed when it encounters AVX instructions in applications. And since AVX is not used extensively or at all in most common applications it will allow you to run most of the time at the overclock setting you have settled on. So when AVX instructions were encountered the CPU would run at 4.3 ghz in stead of 4.5 ghz.

Also, you neglected to tell us anything about your case and case ventilation. Poor case ventilation can contribute to high temps.

 

[Runscream]

Its a $35 air cooler, I wouldn't expect that much out of them at 1.35V...

I wouldn't worry too much, no. Consider using an AVX offset, -3 or -4, since your workloads don't appear to use those instruction sets. That will keep temps down in P95.

You're right, I guess I should consider investing on a better cooler later on. I guess it's actually doing good for what it costed. AVX offset helped, now temps peak at 92c and then go lower.

Something seems wrong also with your settings. Do you really need 1.35 vcore to only get to 4.5 ghz ? Tell us what you changed in bios.

How did you go about overclocking? Did you start by setting the vcore to 1.35? You should do overclocking in increments. Start with stock voltage and start increasing the core multiplier by 1x increments. Stress test after each increment of increase by running Prime95 blend for 20 minutes. Monitor for temps whenever you stress test. When you can no longer pass the 20 minute stress test give the vcore a .01 increase and restest. Etc. When you think you have topped out run a 2 hr. Prime95 stress test. If you pass it, you will likely be stable in real life computing. If you fail it, either increase the vcore or decrease the multiplier.

The Coffee Lake CPUs run hot when overclocked. Many people are "delidding" them. Are you familiar with that? Putting liquid metal TIM under the lid will typically lower temps 10-20c. This would likely render your present cooler adequate.

Your motherboard may be part of the problem as it has only 4+2 power phase. Kind of an entry level Z chipset motherboard. And having the vcore at 1.35 is not helping that. Your VRM may be overheating.

As others have said, you can set the AVX offset to -2 and that should help. Recent versions of Prime95 use AVX 2 instructions set which produce a terrific amount of heat and stress on a CPU. By using the AVX offset you can allow the CPU to run at 200 mhz lower clock speed when it encounters AVX instructions in applications. And since AVX is not used extensively or at all in most common applications it will allow you to run most of the time at the overclock setting you have settled on. So when AVX instructions were encountered the CPU would run at 4.3 ghz in stead of 4.5 ghz.

Also, you neglected to tell us anything about your case and case ventilation. Poor case ventilation can contribute to high temps.

Honestly, I followed Gigabyte's guide (https://overclocking.guide/gigabyte-z370-overclocking-coffee-lake/), I changed everything it said I should change. It also made me overconfident and I did it the wrong way... But now, with patience and a friend's help, I've been doing it in increments like you said. In my latest testing I'm running 4.7ghz at 1.35 with temps peaking at 92c. However, one core failed at 50 minutes. Since I'll mostly be gaming with this PC, I'm not sure if I should just ignore that, lower the frequency to 4.6ghz and be happy with that or up the voltage to 1.36 and test again. I guess the safest bet would be going for 4.6ghz and test for 2 hours.

Yeah, I'm familiar with delidding. However, I don't live in the US, so I'm gonna have to wait some months before I can get a delidding kit and some liquid metal. I definitely plan on doing it though, as well as upgrading my case and cooler.

I'm not sure how to check VRM temps, but on HWMonitor, of all of the temps listed under the motherboard, 82c is the highest of them all. But yeah, I could tell my motherboard was not the best for overclocking based on the price. It's what I could get at the time, though.

About my case, I'm using a Cooler Master K380 with two intake fans (front+bottom) and two exhaust fans (rear+top). I've been wanting to upgrade, but I can't right now due to space limitations in my desk.

 

[Woomack]

I had no chance to play with 8600K ( have one new that will test this week ) but voltage supposed to be much lower. I'm not sure how this generation is scalling but 1.35V is voltage which in last series is required at 5.0-5.2GHz, not 4.5GHz.

My advice is to set 1.20V and check how high it will go. This CPU has 4.3GHz turbo so 4.5GHz should run at almost stock voltage.
Your cooler should handle 4.7GHz, just not at so high voltage. After digging in my tests, I found ~75°C max on [email protected] /1.20V and H7 QUAD LUMI cooler.
Problems will stability are probably caused because of overheating, not because of too low voltage ... and overheating is probably caused because of too high voltage.
If you need 1.35V to keep it stable at 4.3GHz then just RMA the CPU.

 

[EarthDog]

Exactly. My sample, a 8700k, hits 5ghz at 1.2v....yoyr may be different, but that goes to show how off 1.35v at a much lower clock is.

Consider setting 1.25v and see how far you can push..

 

[diaz]

Its a $35 air cooler, I wouldn't expect that much out of them at 1.35V...

I wouldn't worry too much, no. Consider using an AVX offset, -3 or -4, since your workloads don't appear to use those instruction sets. That will keep temps down in P95.

Quick question, the AVX offset of -3 means that the CPU will downclock by 300MHz when AVX instructions are present?

 

[wagex]

if it runs fine stock, its not a bad cpu. come on people.

 

[Johan45]

Yes that's what it means

 

[trents]

Quick question, the AVX offset of -3 means that the CPU will downclock by 300MHz when AVX instructions are present?

"Roger" as we used to say when CB radios were popular.

- - - Updated - - -

if it runs fine stock, its not a bad cpu. come on people.

This is true. No obligation on the part of Intel or the vendor to replace it. As I said earlier Runscream, it does not appear you won the silicon lottery with this chip. I would consider selling it on ebay and getting another. Just don't say anything about it's overclocking ability. However, before selling it you need to confirm it will pass a long stress test at stock speeds and voltages. Have you tried doing that? If it will not do that then I think you have reason to seek an RMA and a replacement.

 

[Runscream]

I had no chance to play with 8600K ( have one new that will test this week ) but voltage supposed to be much lower. I'm not sure how this generation is scalling but 1.35V is voltage which in last series is required at 5.0-5.2GHz, not 4.5GHz.

My advice is to set 1.20V and check how high it will go. This CPU has 4.3GHz turbo so 4.5GHz should run at almost stock voltage.
Your cooler should handle 4.7GHz, just not at so high voltage. After digging in my tests, I found ~75°C max on [email protected] /1.20V and H7 QUAD LUMI cooler.
Problems will stability are probably caused because of overheating, not because of too low voltage ... and overheating is probably caused because of too high voltage.
If you need 1.35V to keep it stable at 4.3GHz then just RMA the CPU.

One thing to note about voltage is that even when I Was testing between 1.30v and 1.35v, both HWMonitor and CPU-Z were reading 1.20v-1.24v most of the time, occasionally peaking to the set voltage for a little bit before going back to 1.20-1.24. Is this normal? I thought voltage was supposed to be constant.

 

[diaz]

Just looked at the VRMs on the Gigabyte Z370P D3 board

I would definitely stay away from OC'ing with this. I would lower the voltages as a minimum.

 

[EarthDog]

One thing to note about voltage is that even when I Was testing between 1.30v and 1.35v, both HWMonitor and CPU-Z were reading 1.20v-1.24v most of the time, occasionally peaking to the set voltage for a little bit before going back to 1.20-1.24. Is this normal? I thought voltage was supposed to be constant.

Its software which can be inaccurate....and its not constant anyway (just not normally that big of a difference).

Is there anything set to 1.2-1.24V in your BIOS, it may be reading that instead of Vcore. Try Coretemp and see what voltage it reads for the hell of it.

Also, since oyu lowered the AVX, you can likely lower vcore more as well since it isn't getting stressed out with those instruction sets.

Just looked at the VRMs on the Gigabyte Z370P D3 board

I would definitely stay away from OC'ing with this. I would lower the voltages as a minimum.

4 phase with a heatsink... not great, but, it will be fine for moderate overclocks surely.

 

[Runscream]

"Roger" as we used to say when CB radios were popular.

- - - Updated - - -



This is true. No obligation on the part of Intel or the vendor to replace it. As I said earlier Runscream, it does not appear you won the silicon lottery with this chip. I would consider selling it on ebay and getting another. Just don't say anything about it's overclocking ability. However, before selling it you need to confirm it will pass a long stress test at stock speeds and voltages. Have you tried doing that? If it will not do that then I think you have reason to seek an RMA and a replacement.

I left it stress testing with everything set at stock before coming to work to see if everything goes ok. It seemed weird to me though that all of the cores were running at 4.1ghz. Temps were good (under 70c) as far as I could see, and voltage stayed at around 1.12-1.15v. Shouldn't at least one or two cores be running at 4.3ghz?

Just looked at the VRMs on the Gigabyte Z370P D3 board

I would definitely stay away from OC'ing with this. I would lower the voltages as a minimum.

Would you recommend any of these boards?
MSI Z370 GAMING PLUS
ASUS TUF Z370 PLUS GAMING

I might consider selling the one I have and get one of those.

 

[trents]

I suggest using HWMonitor to monitor core speeds. It will display a column for the frequency of each core and record max, min and current frequencies. So if you run a stress test, over time you should see in the max column the stock turbo frequency on all cores. They will trade off but over time they should all show that stock turbo frequency was achieved at times. Now, this is with the assumption that no thermal throttling is occurring and that there is 0 AVX offset in bios. It is possible that on Auto the bios will apply an AVX offset so manually set that to 0 for the purpose of exploring the turbo question.

 

[Runscream]

I suggest using HWMonitor to monitor core speeds. It will display a column for the frequency of each core and record max, min and current frequencies. So if you run a stress test, over time you should see in the max column the stock turbo frequency on all cores. They will trade off but over time they should all show that stock turbo frequency was achieved at times. Now, this is with the assumption that no thermal throttling is occurring and that there is 0 AVX offset in bios. It is possible that on Auto the bios will apply an AVX offset so manually set that to 0 for the purpose of exploring the turbo question.

Yeah, I'm using HWMonitor. Ok then, I'll see when I get back home if any core reached turbo speed and then test with 0 avx offset. I'll report when I get home. Thanks

 

[diaz]

Would you recommend any of these boards?
MSI Z370 GAMING PLUS
ASUS TUF Z370 PLUS GAMING

I might consider selling the one I have and get one of those.

They are around the same for VRM. If you want better voltage stability without spending too much i would begin by researching these boards;

https://www.newegg.ca/Product/Product.aspx?Item=N82E16813144111
https://www.newegg.ca/Product/Product.aspx?Item=N82E16813157790
https://www.newegg.ca/Product/Product.aspx?Item=N82E16813157789
https://www.newegg.ca/Product/Product.aspx?Item=N82E16813144108 <- I got this one, but im not OCing
https://www.newegg.ca/Product/Product.aspx?Item=N82E16813119038

But again, if Earthdog isn't worried, I wouldn't go through the trouble. Tinker with your setup and find a sweet spot..

 

[TonyMc]

Something seems wrong also with your settings. Do you really need 1.35 vcore to only get to 4.5 ghz ? Tell us what you changed in bios.

I was about to ask the same question.
Why do you need 1.35v to hit a 4.5 OC? That seems a little excessive...
Drop your volts and get a different cooler.

 

[Der Kommandant]

While I was looking for definite information about OC'ing my 8700 found that people insist on going straight to 1.35V and sticking there. My personal experience is that excess voltage is bad too for overclocking.

Loafing thru many different threads noticed this, all go blindly and stick to 1.35V. Question is... why? My 8700 is happy at x50 multi with AVX -2 cache goes up at x46 multi too at 1.275V. Found other 8600K that roar at similar voltages too. VCSSA/VCCIO too can help to tame temps (everything helps a bit).

Have you tried to go lower (i.e: 1.3V), test... (1 Hour of Prime95) if passes down a notch and still? (I know that's a long and tedious process but wondering if I found an unicorn with my 8700K).