TR Forums

Lore wrote:

The bad news is even with a normal core voltage(1.19Vmax) and all 3 case fans running at max it still reaches 100C in the end, it just takes much longer to reach that temp than it did with the higher voltage. I'll report back on the temps when the new cooler is here.

That makes me wonder if the heat spreader was not properly mated to the CPU at the factory. If this is the case, it does not matter how good your cooler is, as the heat has to make to to the cooler to be removed. You may want to consider an RMA of the CPU.

Lore wrote:

I'm kinda curius about total system power usage myself, but that doesn't really tell me exactly how much of that is used by the cpu. Money is also low after buying these parts, buying presents and having to order a better cooler for this hot cpu and a new receiver for my speakers since the old one died after only 5 years of use on thursday. The bad news is even with a normal core voltage(1.19Vmax) and all 3 case fans running at max it still reaches 100C in the end, it just takes much longer to reach that temp than it did with the higher voltage. I'll report back on the temps when the new cooler is here.

What is the load? Prime95 small fft's? What are the clock speeds on those cores before it hits 100c?

HERETIC wrote:

GUYS-Why is everyone still miss-using TDP as power usage?
TDP (Thermal design power) is there for heatsink manufacturers to know how much waste heat they have to design for.

He's not understanding the difference between total system draw and CPU draw so I wasn't even going there.

The Thermal Solution Specification for the 8700K is actually 130W (vs. 95W TDP), where the Thermal Solution Specification for the 8700 matches it's 65W TDP.

Lore wrote:

The bad news is even with a normal core voltage(1.19Vmax) and all 3 case fans running at max it still reaches 100C in the end

I was finally able to look through the manual for your MSI motherboard. Check in you OC settings for a setting called CPU Ratio Apply Mode. According to the manual, this defaults to All Cores. Try changing it to Per Core.

Also, I noticed that they default Windows 10 WHQL Support to Disabled, and Boot Mode Select to LEGACY+UEFI. You don't necessarily need to change the Windows 10 WHQL Support setting, but I would recommend settings Boot Mode Select to UEFI if it's not already there (Enabling Windows 10 WHQL Support forces the Boot Mode Select option to UEFI). Note that this may require a reinstall of Windows if it's currently set to LEGACY+UEFI and Windows is running in Legacy mode. It depends on how they implemented it. You can see which mode Windows is currently running in by running msinfo32.exe and checking the BIOS Mode (you're good if it says UEFI). If it doesn't say UEFI then I would definitely go through the process of changing the setting and reinstalling Windows. It's possible that if you're in legacy mode Windows might not be able to do all of the power management for the CPU/motherboard.

If you do end up reinstalling Windows, then make sure you have a good backup and let it recreate the partitions on your boot drive. UEFI uses a GPT partition for the boot drive where legacy mode would use an MBR partition type.

curtisb wrote:

Also, I noticed that they default Windows 10 WHQL Support to Disabled, and Boot Mode Select to LEGACY+UEFI. You don't necessarily need to change the Windows 10 WHQL Support setting, but I would recommend settings Boot Mode Select to UEFI if it's not already there (Enabling Windows 10 WHQL Support forces the Boot Mode Select option to UEFI). Note that this may require a reinstall of Windows if it's currently set to LEGACY+UEFI and Windows is running in Legacy mode. It depends on how they implemented it. You can see which mode Windows is currently running in by running msinfo32.exe and checking the BIOS Mode (you're good if it says UEFI). If it doesn't say UEFI then I would definitely go through the process of changing the setting and reinstalling Windows. It's possible that if you're in legacy mode Windows might not be able to do all of the power management for the CPU/motherboard.

If you do end up reinstalling Windows, then make sure you have a good backup and let it recreate the partitions on your boot drive. UEFI uses a GPT partition for the boot drive where legacy mode would use an MBR partition type.

What does changing to UEFI do to affect thermals?

Lore wrote:

I'm kinda curius about total system power usage myself, but that doesn't really tell me exactly how much of that is used by the cpu. Money is also low after buying these parts, buying presents and having to order a better cooler for this hot cpu and a new receiver for my speakers since the old one died after only 5 years of use on thursday. The bad news is even with a normal core voltage(1.19Vmax) and all 3 case fans running at max it still reaches 100C in the end, it just takes much longer to reach that temp than it did with the higher voltage. I'll report back on the temps when the new cooler is here.

Install and run "CORE TEMP"
It'll give you-core temperature-cpu power usage-ram usage,and cpu usage.
also cpu speed,multi and bus.

thecoldanddarkone: The load is occt or p95 in place large ffts. The clock speed before it hits 100C and starts thottling just enough to not go beyond 100C is 4.3Ghz on all cores - which is correct for this cpu loaded on all cores. After that it goes down to 4.0-4.1 and a 1.11-1.14V core voltage.

HERETIC: I'm using occt, it has all the information you are speaking about and I posted imgur links to pictures of it.

curtisb: The power management appears to be correct from what I'm seeing and I'm not keen on reintalling windows. Sadly windows can't regulate the impedance of the cpu other than reducing the multiplier. The problem is the heat produced by the cpu even at what should be reasonable speeds and voltages, and the ability to remove this heat. About the "CPU Ratio Apply Mode", if it somehow was able to circumvent this not being a k cpu, and run 4.6Ghz loaded on all cores, then maybe this setting would help. In reality in never runs the 4.6 turbo, because there is always another thread there so 4.5 is the max speed observed on light loads, 4.3 on heavy ones for a short while.

At this point I we can say that it is not an automatic overclock issue. It is possible that there is a heat transfer issue. I looked at the base of the cooler used (Akasa nero 3) and noticed that the outer 2 heatpipes are unlikely to overlap the cpu underneath by much. Then there is the possibility of a poorly attached heatspreader further hurting thermal performance.

I would give it one more try with the new cooler but I would also seriously consider returning the cpu. These days, in addition to the silicon lottery, we have to play the heat spreader lottery as well.

There is always double-D: delid and D15. (/shamlesplug for noctua - or similar, but don't fall into the CLC trap) Practically required to get your money's worth out of the K version.

I would look into your heatsink mounting some more, it is an older model and might not be making the best contact with the chip. The 115x socket itself has not changed at all but the cpus have gotten thinner starting with skylake.

The main problem with unsoldered intel cpus is not the actual paste (its fine) but the sloppy assembly and excessive sealant that typically leave a poor or uneven seal between the heatspreader and die. After you've popped a dozen or so it becomes pretty obvious that this is the new normal and they all suck. You can go nuts with liquid metal or whatever if you really want but the main improvement is cleaning off the crap and putting it back together with some actual precision, any decent paste is fine. The aftermarket bang/$ is probably a giant tube of gelid extreme (noctua's included stuff is almost as good too) decently close to liquid metal performance, same viscosity as typical oem paste and no conductive crap to worry about.

Lore wrote:

HERETIC: I'm using occt, it has all the information you are speaking about and I posted imgur links to pictures of it.

The reason I suggested you run Core Temp was your interest in how much power your CPU was using-
"I'm kinda curius about total system power usage myself, but that doesn't really tell me exactly how much of that is used by the cpu."

Core Temp should tell you how much power your CPU is drawing...........

If by impedance you mean voltage, that should be dropping in Windows too, when downclocking and idling. You and Mentawl seem to be the right track - I’m betting on MSI UEFI applying crazy voltages. Can you make sure that both SpeedStep and SpeedShift are enabled? They do different things slightly, with the latter being newer and CPU-based. Did you manually set a CPU voltage yet? Try 1.28 to 1.32 with both Auto LLC and then low-level LLC. Use offsets if you can’t type one in (i’ve kept up with the thread but lost track.).

I like HWMonitor, but to each his or her own.

HERETIC you are absolutely right about TDP. So sick of it. Actual Processor Power is part of the Silicon Lottery function. IF this was a K proc, I don’t doubt that it could approach 150W with heavy heavy over clocking. Heavy.

Heretic: sorry I missed that, actually a nice little program. 85-87W at load and 9-12W at idle. I do find myself doubting the load number a little bit in my particular case thinking it may be even higher than this, though it does sound reasonable for a cpu that didn't lose the silicon lottery. Do we know for sure if the program really gets direct information about the amperage flowing through the mosfets to the cpu, or if this is some kind of estimation ?

Mosfet: Impedance is the resistance the cpu represents in the circuit under different speeds. So like you are speaking about, individual cpus of the same type at the same voltage and speed will spend slightly different amounts of power because of the silicon lottery and now even the heatspreader lottery. Power use goes up with temps. The first transistor radios would burn up at the hot beach because of this phenomenon. A more important part of this lottery is needing more or less voltage to be stable. In my case one of 6 cores is unstable at what would be a nice low voltage, damn you core number four ! The voltages are ok now, was manually set thanks to Mentawl.

Flying Fox wrote:

curtisb wrote:

Also, I noticed that they default Windows 10 WHQL Support to Disabled, and Boot Mode Select to LEGACY+UEFI. You don't necessarily need to change the Windows 10 WHQL Support setting, but I would recommend settings Boot Mode Select to UEFI if it's not already there (Enabling Windows 10 WHQL Support forces the Boot Mode Select option to UEFI). Note that this may require a reinstall of Windows if it's currently set to LEGACY+UEFI and Windows is running in Legacy mode. It depends on how they implemented it. You can see which mode Windows is currently running in by running msinfo32.exe and checking the BIOS Mode (you're good if it says UEFI). If it doesn't say UEFI then I would definitely go through the process of changing the setting and reinstalling Windows. It's possible that if you're in legacy mode Windows might not be able to do all of the power management for the CPU/motherboard.

If you do end up reinstalling Windows, then make sure you have a good backup and let it recreate the partitions on your boot drive. UEFI uses a GPT partition for the boot drive where legacy mode would use an MBR partition type.

What does changing to UEFI do to affect thermals?

Probably nothing, but there are other advantages in making sure you're on UEFI. In all likelihood, he's running on UEFI already...I was just making a suggestion to double check, which takes all of two seconds in msinfo32.


I still think it would be worth him setting the CPU Apply Ratio Mode to Per Core to see if it makes a difference. That's how the CPUs were designed and intended to run. It may not since the apps he's using are designed to stress the CPU by running them at full load, but it could help with normal workloads.

Checked now, the system info page that came up from msinfo32 says bios uefi. I think even my ancient sandy bridge system was running uefi.

Yeah, UEFI has been around for a while. Initial UEFI-based systems were forced into Legacy mode, then it went to selectable with Legacy being the default, and eventually defaulted to UEFI. The manual for your motherboard just caught me off guard because it defaulted to LEGACY+UEFI, which is why I suggested you check. What that setting means is that you can boot your system with either type of boot media...for example, we have a DOS-based flash drive at work for BIOS updates and it requires that the system be able to boot into legacy mode, but we install Windows in UEFI mode. The way Windows installs depends on how you boot the install media, although it'll try it's best to use UEFI.

Lore wrote:

Heretic: sorry I missed that, actually a nice little program. 85-87W at load and 9-12W at idle. I do find myself doubting the load number a little bit in my particular case thinking it may be even higher than this, though it does sound reasonable for a cpu that didn't lose the silicon lottery. Do we know for sure if the program really gets direct information about the amperage flowing through the mosfets to the cpu, or if this is some kind of estimation ?

It's even nicer using Core Temp Gadget-I have it sitting bottom right corner of screen-all information available,and one temp and cpu usage on toolbar for when running something full screen.
That's on W7 thro-MS might have killed off the use of gadgets on later OS's.
As to your last question-I think there's a certain amount of guestimation to power and temps,and possibly voltage as well(on this MB bios tells me my 12 volt is 7 volts)
If you remember the old speedfan-you used to be able to add offsets to compensate dodgy temp info........................

A $20 plug/wall meter is a useful tool,that would give you system power,a bit of calculation could then confirm if that's correct......
And to make it pay for itself-Use it to check all the "ghost power" all your appliances are using............................

New cooler installed yesterday, and it is so much better it is a little unreal. Msi Frozr isn't the best cooler out there, but it does manage 63-68C(78C max for a little while) even after working p95 for many hours, instead of the 100C i was getting with the Akasa Nero and stock cooler. I don't know if it is the thicker heat pipes, the backplate and screws instead of plastic pushtrough pins, 900+grams instead of 600grams or a combination of all these, but it works a lot better.

Lore wrote:

New cooler installed yesterday, and it is so much better it is a little unreal. Msi Frozr isn't the best cooler out there, but it does manage 63-68C(78C max for a little while) even after working p95 for many hours, instead of the 100C i was getting with the Akasa Nero and stock cooler. I don't know if it is the thicker heat pipes, the backplate and screws instead of plastic pushtrough pins, 900+grams instead of 600grams or a combination of all these, but it works a lot better.

WONDERFULL
Bolted on heatsinks can achieve more pressure than pushpins,which is good. Might drop a few degrees more once your compound has cooked in..............................

Just read a post by Strangerguy here-
viewtopic.php?f=29&t=81745&p=1372323#p1372323
He's undervolted his 8700K to 1.05V-That would really drop temps........................................

For comparison, I've got a non-delidded i7-8700K mated to a Hyper 212 EVO cooler in a Fractal Design case. Just now ran Prime95 for about five minutes and the CPU temp stuck to between 60 and 70 Celsius. Cores are all at 3.8 GHz and I'm presently not overclocking.

Same system can reach just under 5 GHz with automatic overclocking and is pretty (but not perfectly) stable at that speed.

Lore wrote:

New cooler installed yesterday, and it is so much better it is a little unreal. Msi Frozr isn't the best cooler out there, but it does manage 63-68C(78C max for a little while) even after working p95 for many hours, instead of the 100C i was getting with the Akasa Nero and stock cooler. I don't know if it is the thicker heat pipes, the backplate and screws instead of plastic pushtrough pins, 900+grams instead of 600grams or a combination of all these, but it works a lot better.

I am having same problem with my 8700. I have contacted Intel and they are investigating it.
They asked me to run Intel® Extreme Tuning Utility (Intel® XTU) to stress CPU and checked the highest package TDP during that. Mine is 117W. My motherboard is ASRock Z370M.

HERETIC wrote:

GUYS-Why is everyone still miss-using TDP as power usage?
TDP (Thermal design power) is there for heatsink manufacturers to know how much waste heat they have to design for.

...because the power usage of a CPU almost directly translates into heat load. They're very good at converting power into heat - the work done is a side effect.

flyxman007 wrote:

I am having same problem with my 8700. I have contacted Intel and they are investigating it.
They asked me to run Intel® Extreme Tuning Utility (Intel® XTU) to stress CPU and checked the highest package TDP during that. Mine is 117W. My motherboard is ASRock Z370M.

Tried the Intel® Extreme Tuning Utility, max TDP was 87W. Most of the time it showed betweem 61-70W. What cooler are you using, and what core voltage do you get under load in programs like occt or core temp ?

flyxman007 wrote:

Lore wrote:

New cooler installed yesterday, and it is so much better it is a little unreal. Msi Frozr isn't the best cooler out there, but it does manage 63-68C(78C max for a little while) even after working p95 for many hours, instead of the 100C i was getting with the Akasa Nero and stock cooler. I don't know if it is the thicker heat pipes, the backplate and screws instead of plastic pushtrough pins, 900+grams instead of 600grams or a combination of all these, but it works a lot better.

I am having same problem with my 8700. I have contacted Intel and they are investigating it.
They asked me to run Intel® Extreme Tuning Utility (Intel® XTU) to stress CPU and checked the highest package TDP during that. Mine is 117W. My motherboard is ASRock Z370M.

I'd ask the same question of you flyxman007 - have you checked what voltages the motherboard is supply to the CPU?

Lore wrote:

flyxman007 wrote:

I am having same problem with my 8700. I have contacted Intel and they are investigating it.
They asked me to run Intel® Extreme Tuning Utility (Intel® XTU) to stress CPU and checked the highest package TDP during that. Mine is 117W. My motherboard is ASRock Z370M.

Tried the Intel® Extreme Tuning Utility, max TDP was 87W. Most of the time it showed betweem 61-70W. What cooler are you using, and what core voltage do you get under load in programs like occt or core temp ?

Thanks! That sounds a proper i7-8700! My reach to 117W with auto settings with BIOS. It shows max turbo power limit as 120W. It is more like my 8700 is faulty.

Mentawl wrote:

flyxman007 wrote:

Lore wrote:

New cooler installed yesterday, and it is so much better it is a little unreal. Msi Frozr isn't the best cooler out there, but it does manage 63-68C(78C max for a little while) even after working p95 for many hours, instead of the 100C i was getting with the Akasa Nero and stock cooler. I don't know if it is the thicker heat pipes, the backplate and screws instead of plastic pushtrough pins, 900+grams instead of 600grams or a combination of all these, but it works a lot better.

I am having same problem with my 8700. I have contacted Intel and they are investigating it.
They asked me to run Intel® Extreme Tuning Utility (Intel® XTU) to stress CPU and checked the highest package TDP during that. Mine is 117W. My motherboard is ASRock Z370M.

I'd ask the same question of you flyxman007 - have you checked what voltages the motherboard is supply to the CPU?

The voltages are high. In HWmonitor, it shows max 1.5v. I tried to set vcore voltage to fixed 1.25 - 1.3v, Windows crashed immediately after I start aida64 test. I also tried offset mode it doesn't work either.

I have hoped I can find some settings to get what a 8700 should perform. However, nothing seems work at the moment.

I ran octa and core temp. CPU power shows 95w (as I set PL1 limit to 95w to reduce the heat), vid voltage shows 1.24 - 1.33, frequency around 3.8GHz. Max temp 79.

Most appreciated if anyone can tell me what is wrong with my settings. I'd rather hope it is a settings issue as I really don't want to return the CPU and motherboard, redo everything. It is a time consuming job.

Tried to play with vcore voltage again. I change LLC from AUTO to 4, the lowest I can set in the BIOS, change vcore voltage from AUTO to 1.300 and change RAM from 1.350 to 1.20, leave Power limit to default AUTO value. The intel utility test now show max 105W TDP rather than 117 I saw before. I tried to change vcore voltage to 1.28, 1.275, 1.265, they didn't make different in terms of the max TDP reported. 1.265 or 1.255 windows frozen. I keep it to 1.300 now.

HW monitor still reports max package power 123W with those changes even if Intel tool reports max 105W.

flyxman007 wrote:

Tried to play with vcore voltage again. I change LLC from AUTO to 4, the lowest I can set in the BIOS, change vcore voltage from AUTO to 1.300 and change RAM from 1.350 to 1.20, leave Power limit to default AUTO value. The intel utility test now show max 105W TDP rather than 117 I saw before. I tried to change vcore voltage to 1.28, 1.275, 1.265, they didn't make different in terms of the max TDP reported. 1.265 or 1.255 windows frozen. I keep it to 1.300 now.

HW monitor still reports max package power 123W with those changes even if Intel tool reports max 105W.

Leave your RAM voltage on AUTO for now.
Experiment with lowering your POWER LIMIT.

HERETIC wrote:

Leave your RAM voltage on AUTO for now.
Experiment with lowering your POWER LIMIT.

Thanks. i tried that with no luck. limit the power limit 1 to 65 W, leave ram AUTO, the max TDP is capped to 65w and power liimit hit is reported. whatever power limit I set, it will hit that when running intel extreme utility to stress the CPU. The core frequency dropped by a lot.

another people's 8700 is just using 77 TDP at max.

i tried voltage 1.285, ran aida 64, system froze.

There isnot many things i can chaning in bios. i think i have tried all the poosible choice i can think of now.

i am seriously suspecting my i7-8700 is defective. I will see what Intel will say. I am going to contact ASRock raegarding this too. Howevrer, it looks like either get replacement or switch to AMD platform such as 1700.

i am using low profile Noctua L9i CPU Cooler with a case fan to cool thr CPU. it is capble to keep it under 80 in Aida64 test if i set the power limit 1 to 95w, and keep the pc quiet. this 8700 is more like a 8700k in terms of power consumption to me.