i7 8700k 4,7Ghz MAX Turbo Limit

[nightmare392]

Hi guys,

I have an intel I7 8700K mounted on ASUS Z370-E Gaming motherboard with a Cooler Master MasterLiquid 240 Pro AIO cooler. I tried running 4.7Ghz on all cores. I set my VCORE to Adaptive mode and a value of 1.2 V, Load Line Calibration to 4 (with 3 it fails in Prime95 10 sec into the test). I maxed out the power and current limits. Now when i run Prime95 Small FFT test, i get 4,7Ghz at first, but it drops to 4,6 and remains there, with the test running stable. Looking in HWiNFO i see "IA: Max Turbo Limit" set to YES. There are no other events (no throttling due to temperatures or power draw). The CPU temperature never goes over 79 degrees during the prime95 test. What could be causing the CPU to throttle down to 4,6Ghz? I attached a screenshot from HWiNFO.

 

[faalin]

Is prime running on all cores?

A 8700k with boost will run at these speeds per core used

1 core 4.7
2 core 4.6
3 core 4.5
4-5 core 4.4
6 core 4.3

So if prime for some reason is only using 2 cores then it cpu would default to 4.6 ghz on boost

 

[biglizard]

Max Turbo Limit - is telling you just that you've reached what you have set as turbo limit.

What version of prime?

Do you have an AVX offset set?

 

[nightmare392]

My prime95 version is 29.4. I have AVX offset set to 0. I set the limit on all cores to 47, and they all go to 47 for the first 20-30 seconds, dropping to 46 after that. prime95 is running on all cores

 

[biglizard]

My prime95 version is 29.4. I have AVX offset set to 0. I set the limit on all cores to 47, and they all go to 47 for the first 20-30 seconds, dropping to 46 after that. prime95 is running on all cores

You have said you set vcore to 1.2, your showing 1.152 in hwinfo.

I'm curious if setting LLC to 5 would make a difference. Too much droop?

How long did you run prime for?

Run Version 26.6, see if same behavior.

 

[nightmare392]

My vcore is set to 1.2 adaptive, not fixed. I will try to set LLC to 5, but it might bring the temperatures up. I ran prime for about an hour. In large fft (512-4096) it does the same.

 

[biglizard]

My vcore is set to 1.2 adaptive, not fixed. I will try to set LLC to 5, but it might bring the temperatures up. I ran prime for about an hour.

Any increase in core voltage will increase temps, as you know.

Seems strange that your losing a bin with no power or temp reasons.

 

[nightmare392]

That is what i can't understand. There is no reason to throttle. Temps are good, power limit is not reached, vrm is not too hot to cause throttling. Maybe it's not getting enough vcore. Should i try to increase to 1.25V and see if it can hold 4,7? It seems a bit too much 1.25V for 4.7Ghz.

 

[biglizard]

That is what i can't understand. There is no reason to throttle. Temps are good, power limit is not reached, vrm is not too hot to cause throttling. Maybe it's not getting enough vcore. Should i try to increase to 1.25V and see if it can hold 4,7? It seems a bit too much 1.25V for 4.7Ghz.

I'd try upping LLC first.

Low voltage is my thinking also. Since your using Prime ver 29 which use AVX instructions, it may be dropping a bin. Normally this would just be unstable, though.

Have you set AC/DC load lines to 0.01?

How do you know that VRM is not hot? I wouldn't think that they are considering vcore, Just curious.

 

[nightmare392]

The AC/DC load lines are set to 0.01. Leaving them on Auto was causing VR Thermal Alert and throttling (the motherboard was probably setting a very high value). I will try upping LLC tomorrow, when i get back from work. If it doesn't work, i will try to increase Vcore. In the mean time, i had to increase VCCSA to 1.3 and VCCIO to 1.25 to get my G.Skill 3200Mhz memory to work with XMP enabled (i gradually raised the values from 1.1 and 1.15, until i got no fail in Prime95 512k-4096K). The values seem high, but i could not get it stable with anything under these. The DRAM voltage is set to 1.35 by the XMP. I tried 1.36 with VCCIO and VCCSA to 1.15 and 1.2 and it would fail in Prime95. It has been running for 2h now without errors, but i am worried that the 1.3V VCCSA might be a little high.

 

[biglizard]

The AC/DC load lines are set to 0.01. Leaving them on Auto was causing VR Thermal Alert and throttling (the motherboard was probably setting a very high value). I will try upping LLC tomorrow, when i get back from work. If it doesn't work, i will try to increase Vcore. In the mean time, i had to increase VCCSA to 1.3 and VCCIO to 1.25 to get my G.Skill 3200Mhz memory to work with XMP enabled (i gradually raised the values from 1.1 and 1.15, until i got no fail in Prime95 512k-4096K). The values seem high, but i could not get it stable with anything under these. The DRAM voltage is set to 1.35 by the XMP. I tried 1.36 with VCCIO and VCCSA to 1.15 and 1.2 and it would fail in Prime95. It has been running for 2h now without errors, but i am worried that the 1.3V VCCSA might be a little high.

It is somewhat high for 3200mhz, but if I'm not mistaken higher SA can be needed to stabilize a weak IMC. Higher LLC may raise IO and SA it does on my board. You may want to compensate for that.

 

[nightmare392]

Is there any way that something else is affecting the stability of the RAM, and raising VCCSA is compensating? Nothing comes to my mind, but maybe i'm missing something

 

[biglizard]

Is there any way that something else is affecting the stability of the RAM, and raising VCCSA is compensating? Nothing comes to my mind, but maybe i'm missing something

With SA too high can cause instability, its the touchiest of all the voltages in my experience.

According to this write up, your on the high side but ok. http://edgeup.asus.com/2017/kaby-lake-overclocking-guide/3/

 

[nightmare392]

I found the problem with the multiplier dropping to 46. I had Intel Extreme Tuning Utility installed and it had 46 set (i must have set it at some time). Now it goes to 4.7, but i have other problems. I set 1.2V in the BIOS, and i get 1,152 during full load in Prime95. If i increase LLC to 6 i get VR Thermal Alert (the VRM is probably too hot). Should it be getting so hot at only 1.2V Vcore, and why the huge voltage drop? Am i doing something wrong? I also tried setting 1,225V, and i was getting 1,120V during load, with LLC set to 5.

 

[biglizard]

Hi guys,

I have an intel I7 8700K mounted on ASUS Z370-E Gaming motherboard with a Cooler Master MasterLiquid 240 Pro AIO cooler. I tried running 4.7Ghz on all cores. I set my VCORE to Adaptive mode and a value of 1.2 V, Load Line Calibration to 4 (with 3 it fails in Prime95 10 sec into the test). I maxed out the power and current limits. Now when i run Prime95 Small FFT test, i get 4,7Ghz at first, but it drops to 4,6 and remains there, with the test running stable. Looking in HWiNFO i see "IA: Max Turbo Limit" set to YES. There are no other events (no throttling due to temperatures or power draw). The CPU temperature never goes over 79 degrees during the prime95 test. What could be causing the CPU to throttle down to 4,6Ghz? I attached a screenshot from HWiNFO.

I found the problem with the multiplier dropping to 46. I had Intel Extreme Tuning Utility installed and it had 46 set (i must have set it at some time). Now it goes to 4.7, but i have other problems. I set 1.2V in the BIOS, and i get 1,152 during full load in Prime95. If i increase LLC to 6 i get VR Thermal Alert (the VRM is probably too hot). Should it be getting so hot at only 1.2V Vcore, and why the huge voltage drop? Am i doing something wrong? I also tried setting 1,225V, and i was getting 1,120V during load, with LLC set to 5.

Try more air flow over VRM.

Are VRM heat sinks fastened tightly?

If those don't help, I'd RMA or return board.

 

[nightmare392]

I tightned the vrm heatsink screws a bit and i got the termal alert later. It seems like setting ac/dc load lines both to 0.01 causes the huge difference between the set vcore and the actual one. I reverted back to stock, but with the load lines on auto i get 83-84 degrees at 4.3Ghz. Something is not right. I should't be getting these temps with an 240mm AIO cooler.

 

[4Ryan6]

I tightned the vrm heatsink screws a bit and i got the termal alert later. It seems like setting ac/dc load lines both to 0.01 causes the huge difference between the set vcore and the actual one. I reverted back to stock, but with the load lines on auto i get 83-84 degrees at 4.3Ghz. Something is not right. I should't be getting these temps with an 240mm AIO cooler.

Temperature has everything to do with any CPUs performance and just how fast the CPUs temperature is approaching higher numbers depends on how much the CPU throttles itself to protect itself and those protective features are built into the CPU. The first protection features are sent by the CPU to the motherboard to tell the motherboard to increase the CPU fan speed but if the response is not fast enough the CPU activates internal protective features. These protective features may not even show up as drastic as the multiplier dropping, but are measurable in a benchmark when you suddenly see your previous scores dropping.

That said I would like to help you reach a higher overclock, with the cooling you are running you should at least be able to run 4.7ghz on all cores. I am running an 8700K with all cores overclocked to 5ghz at 1.325v, however to reach that I am running chilled water cooling and the CPU has been delided, but I was running 5ghz before I delided the CPU and the delid has allowed running a lower Vcore and lower temperatures.

Are you willing to run a manual Vcore, and how many levels of LLC does your motherboard BIOS have? I am running the ASUS ROG Maximus X Hero and we need to compare some notes BIOS wise between it and your ASUS Z370-E Gaming?

 

[4Ryan6]

One more thing, are you running the last available BIOS Flash for your motherboard, if not you need to flash it to the last released BIOS.

I have other things to get done today but I'll check back when I can. Ryan

 

[nightmare392]

Thank you for your answer. I know that temperature has everything to do with the performance of the CPU, but i am getting too high temperatures for an AIO cooler at stock frequencies (i have the CoolerMaster MasterLiquid 240 Pro). I have the latest BIOS (0606). My BIOS has 7 levels of LLC, but the max i could reach is 5 before my VRM begins to throttle. Something doesn't seem right. At stock (4.3 Ghz on all cores) i get a power draw of 141W (it seems a little high for stock). On Auto Vcore i get 1.232V with all cores at 4.3Ghz during Prime95 small fft test. I attached a screenshot from HWinfo, that shows temperatures when running prime95 small fft, and the power draw. I will try to remove the cooler and the CPU, look for any bent pins on the socket and install everything back applying thermal paste again (i am using Gelid GC-Extreme paste). I hope i don't have a deffective CPU that draws too much current. Since you have offered to help, any advice is welcome. Thank you!

 

[4Ryan6]

The red flag I see in your last HWiNFO is under your CPU Package Power 141.8w and for the clock on a 95w TDP CPU that's way too high, mine is no where near that high at a 5ghz OC, right off the bat I would say somethings wrong with your motherboard.

 

[nightmare392]

That power draw is under full load (prime95 small fft). How much power draw do you have while running prime95? If my motherboard is bad, i will send it back to get replaced. I am also worried about the fact that when my CPU is hot (80-90) degrees, the aio radiator is just a little warm, not hot. The pump is working at 100% rpm (the bios shows 9000 rpm, but it is probably 3000).

 

[4Ryan6]

That power draw is under full load (prime95 small fft). How much power draw do you have while running prime95? If my motherboard is bad, i will send it back to get replaced. I am also worried about the fact that when my CPU is hot (80-90) degrees, the aio radiator is just a little warm, not hot. The pump is working at 100% rpm (the bios shows 9000 rpm, but it is probably 3000).

My wattage readings running P95 26.6 Small FFTs will not be comparative to yours all my C-states are disabled and I am running coolant temperatures that are 10c below ambient, while your temperatures are constantly above ambient which means even at a 5ghz CPU overclock I am running less wattage than you are, a lot less wattage. But just looking at your wattage #s they are recording way too high 141w on a 95w TDP CPU? Somethings wrong there and it's probably the motherboard.

 

[nightmare392]

That power draw is under full load (prime95 small fft). How much power draw do you have while running prime95? If my motherboard is bad, i will send it back to get replaced. I am also worried about the fact that when my CPU is hot (80-90) degrees, the aio radiator is just a little warm, not hot. The pump is working at 100% rpm (the bios shows 9000 rpm, but it is probably 3000).

My wattage readings running P95 26.6 Small FFTs will not be comparative to yours all my C-states are disabled and I am running coolant temperatures that are 10c below ambient, while your temperatures are constantly above ambient which means even at a 5ghz CPU overclock I am running less wattage than you are, a lot less wattage. But just looking at your wattage #s they are recording way too high 141w on a 95w TDP CPU? Somethings wrong there and it's probably the motherboard.

I think there is something wrong with my AIO also, because i shouldn't be getting 85 C during prime (with AVX) with the stock turbo frequency of 4.3Ghz. I replaced the paste 3 times, so it's probably ok. I'm thinking that maybe the water block does not seat properly on the CPU (maybe the screws are too long on the intel mounting bracket). I'm thinking of going for a Corsair H115i instead of the CoolerMaster MasterLiquid 240 Pro i have at the moment. Wold it be a better choice? I also have the power limis to the max, so it never power throttles, that is probably why i get to 141W power draw, and because i am using prime95 with AVX, not 26.6. I tried 26.6 now and i get 107 W power draw and max 73 C.

 

[4Ryan6]

I tried 26.6 now and i get 107 W power draw and max 73 C.

That's looking more realistic!

That's closer to my #s at a 5ghz OC, but still too high for stock.

 

[nightmare392]

I tried 26.6 now and i get 107 W power draw and max 73 C.

That's looking more realistic!

That's closer to my #s at a 5ghz OC, but still too high for stock.

I ordered a Corsair H115i a few minutes ago. I will pick it up tomorrow and see if i get better temps. After i solve the cooling problem, i will test more OC settings to see if the MB is faulty.