• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Delidded my 8700K, are the temps OK, or should I expect better?

Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
Heyya,

So I took the dive and delidded my 8700K (mats: Dr. Delid tool, Thermal Grizzly Conductonaut and Kryonaut, some clear nail "coating"/polish for the 4 small contact points near the die, Victor Reinz's Reinzosil high temp sealant).

I am curious if I did this correctly though, as this is the first time I try it. I mean, the temps look fine, I think, but I'm unsure if more could have been done. Could use somebody with a delidded 8700k willing to either run it at the same voltage, or maybe tell me their voltage and a modest OC (my VRMs are shit, I don't wanna drive more than 150W through them if possible).

My test(s): 5GHz without AVX offset, +0.040V adaptive offset (it puts the CPU at 1.344V in AIDA64 so for static voltage that should do, if you wanna replicate this), CPU/FPU/Cache/memory stresstest for 10mins). Temps peaked at max 73C/core and around 130W were drawn. I'd say that 66-70C temps were the "average".

Another test: prime95, AVX turned off, inplace FFTs 1344K, power draw was higher, around 140W, temps were maybe 5C lower though. I assume AVX from AIDA64 was responsible for the increased temps.

Kinda afraid to use AVX2 or AVX inside p95, would likely require more voltage and also maybe even fry my shitty 4phase VRM. So please not that test, thanks!

Edit: oh yeah, the CPU cooler (D15S from Noctua, 120mm fan in front) didn't really break a sweat, I think the 140m fan topped at around 666RPM and the 120mm fan at around 1K rpm. The max is 1500rpm for the 140mm and 2000rpm for the 120mm. The case fans also ran at kinda 50%. Just putting this here as additional info.
 

Attachments

Joined
Nov 1, 2008
Messages
4,093 (1.00/day)
Location
Vietnam
System Name Gaming System / HTPC-Server
Processor i5 8700K (@4.8 Ghz All-Core) / R5 3600
Motherboard Z370 Aorus Ultra Gaming / MSI B450 Mortar Max
Cooling CM ML240L / CM Seidon 120XL
Memory 16Gb Hynix @3200 MHz / 16Gb Ballistix @3600Mhz
Video Card(s) Colorful 1080Ti / Gigabyte RX 5700 (XT Bios)
Storage 750G MX300 / 8+4 TB WD Reds + 250 Sabrent NVMe
Display(s) 2x Dell U2515H/18D / 50" L50U50 4K TV
Case Xigmatek Aquarius Plus / Zalman HTPC case
Audio Device(s) On Board Realtek
Power Supply FSP 850 W Gold / Andyson TX-700 Platinum
Mouse Logitech G502 Hero / K400+
Keyboard Wooting Two / K400+
Software Windows 10 x64
Benchmark Scores Cinebench R15 = 1542 3D Mark Timespy = 9758
Heyya,

So I took the dive and delidded my 8700K (mats: Dr. Delid tool, Thermal Grizzly Conductonaut and Kryonaut, some clear nail "coating"/polish for the 4 small contact points near the die, Victor Reinz's Reinzosil high temp sealant).

I am curious if I did this correctly though, as this is the first time I try it. I mean, the temps look fine, I think, but I'm unsure if more could have been done. Could use somebody with a delidded 8700k willing to either run it at the same voltage, or maybe tell me their voltage and a modest OC (my VRMs are shit, I don't wanna drive more than 150W through them if possible).

My test(s): 5GHz without AVX offset, +0.040V adaptive offset (it puts the CPU at 1.344V in AIDA64 so for static voltage that should do, if you wanna replicate this), CPU/FPU/Cache/memory stresstest for 10mins). Temps peaked at max 73C/core and around 130W were drawn. I'd say that 66-70C temps were the "average".

Another test: prime95, AVX turned off, inplace FFTs 1344K, power draw was higher, around 140W, temps were maybe 5C lower though. I assume AVX from AIDA64 was responsible for the increased temps.

Kinda afraid to use AVX2 or AVX inside p95, would likely require more voltage and also maybe even fry my shitty 4phase VRM. So please not that test, thanks!

Edit: oh yeah, the CPU cooler (D15S from Noctua, 120mm fan in front) didn't really break a sweat, I think the 140m fan topped at around 666RPM and the 120mm fan at around 1K rpm. The max is 1500rpm for the 140mm and 2000rpm for the 120mm. The case fans also ran at kinda 50%. Just putting this here as additional info.
Temps seem inline for 1.34 V on Kryonaut. If you want better, you need liquid metal. It's a nice chip that you can do 5Ghz on that voltage. Mine only goes to 5 Ghz on ~ 1.39V and temps jump to the lower 80's

For reference, I'm currently running delidded @ 4.8 Ghz 1.3 V and temps get to the mid 70's in AIDA, but I live in a pretty warm climate so room temps are higher.
 
Joined
Feb 3, 2019
Messages
1,005 (2.81/day)
Location
Chicago Land
Processor Lidless PGA 2700X
Motherboard Asus B450-I Gaming
Cooling Stock Air
Memory Corsair 3000nhz 13-15-15-15-32-53 2T
Video Card(s) Asus strix GTX 980 OC
Storage SSD
Display(s) 21" - 55"
Case None
Power Supply Antec CP series 850w
Mouse Razar Mamba Tournament Edition
Keyboard Logitech G910
Software W7
Benchmark Scores Max Cpu clock 7685Mhz FX-8300 WPrime 32m 2.886 seconds AMD 2700x
2nd that liquid metal. Current temps are good. Nice job on the de-lid.
 
Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
Temps seem inline for 1.34 V on Kryonaut. If you want better, you need liquid metal. It's a nice chip that you can do 5Ghz on that voltage. Mine only goes to 5 Ghz on ~ 1.39V and temps jump to the lower 80's

For reference, I'm currently running delidded @ 4.8 Ghz 1.3 V and temps get to the mid 70's in AIDA, but I live in a pretty warm climate so room temps are higher.
It has Conductonaut on the die, Kryonaut between the heatsink and Noctua D15S.

It's not stable at that voltage, at least not in prime95/inlineFFTs/1344K/no AVX (which seems to be the test many overclockers use, including der8auer). It did fine for maybe 1/2hr in AIDA64 though. The voltages also differ depending on workload, I kind of want to keep the adaptive voltage and power saving settings enabled, lot of my time is browsing and working in Photoshop/Indesign, and the CPU idles quite a bit in it. I'll use fixed voltage to match your settings or others' if that's what you use. Ambient temps are around 24C now, although it's kind of cold.

Interestingly enough, it's not workers stopping that causes instability, it's (self-correcting) WHEA cache errors that I can only see in HWinfo64 and the Event Viewer.

At +0.050V offset it seems I got rid of the cache errors, the CPU draws about 140-145W, vcore in p95 is 1.320-1.328-1.344, mostly stuck at 1.320. This might be the upper limit of what I can do on my motherboard, I mean, it's stable and nothing throttles, but the VRM are around maxing at 90C. In the summer this will just go 10-15C above that and will probably require a fan on top, at least while/if I stresstest.

Image for the current run of p95 below (with the +0.050V offset/around 1.33V average vcore).

Edit: it's still not stable dammit, the L0 cache doesn't like it, workers still don't fail, and you wouldn't know it fails without HWinfo/Event Viewer. Time to raise it more.

 

Attachments

Joined
Nov 1, 2008
Messages
4,093 (1.00/day)
Location
Vietnam
System Name Gaming System / HTPC-Server
Processor i5 8700K (@4.8 Ghz All-Core) / R5 3600
Motherboard Z370 Aorus Ultra Gaming / MSI B450 Mortar Max
Cooling CM ML240L / CM Seidon 120XL
Memory 16Gb Hynix @3200 MHz / 16Gb Ballistix @3600Mhz
Video Card(s) Colorful 1080Ti / Gigabyte RX 5700 (XT Bios)
Storage 750G MX300 / 8+4 TB WD Reds + 250 Sabrent NVMe
Display(s) 2x Dell U2515H/18D / 50" L50U50 4K TV
Case Xigmatek Aquarius Plus / Zalman HTPC case
Audio Device(s) On Board Realtek
Power Supply FSP 850 W Gold / Andyson TX-700 Platinum
Mouse Logitech G502 Hero / K400+
Keyboard Wooting Two / K400+
Software Windows 10 x64
Benchmark Scores Cinebench R15 = 1542 3D Mark Timespy = 9758
It has Conductonaut on the die, Kryonaut between the heatsink and Noctua D15S.

It's not stable at that voltage, at least not in prime95/inlineFFTs/1344K/no AVX (which seems to be the test many overclockers use, including der8auer). It did fine for maybe 1/2hr in AIDA64 though. The voltages also differ depending on workload, I kind of want to keep the adaptive voltage and power saving settings enabled, lot of my time is browsing and working in Photoshop/Indesign, and the CPU idles quite a bit in it. I'll use fixed voltage to match your settings or others' if that's what you use. Ambient temps are around 24C now, although it's kind of cold.

Interestingly enough, it's not workers stopping that causes instability, it's (self-correcting) WHEA cache errors that I can only see in HWinfo64 and the Event Viewer.

At +0.050V offset it seems I got rid of the cache errors, the CPU draws about 140-145W, vcore in p95 is 1.320-1.328-1.344, mostly stuck at 1.320. This might be the upper limit of what I can do on my motherboard, I mean, it's stable and nothing throttles, but the VRM are around maxing at 90C. In the summer this will just go 10-15C above that and will probably require a fan on top, at least while/if I stresstest.

Image for the current run of p95 below (with the +0.050V offset/around 1.33V average vcore).

Edit: it's still not stable dammit, the L0 cache doesn't like it, workers still don't fail, and you wouldn't know it fails without HWinfo/Event Viewer. Time to raise it more.

I found it easier to get it 5Ghz stable, at a lower voltage, by using fixed voltage and Turbo LLC. You might want to try reaching stability that way to determine the voltage targets, then try switching it to offset after.
 
Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
I found it easier to get it 5Ghz stable, at a lower voltage, by using fixed voltage and Turbo LLC. You might want to try reaching stability that way to determine the voltage targets, then try switching it to offset after.
I'll try that as well. The motherboard has plenty of LLC settings, like 8 or 9 steps. With 4 or Auto it doesn't seem to "overshoot" on the voltage, so I assume if I max it there will be a lot of overvolting when I start the stresstest. I bumped Vcore to +0.055V in BIOS after the previous L0 cache error, will leave it for maybe 30 mins, and then sleep, tomorrow, testing with fixed voltage and will try to see what temps I get at fixed 4.8 and 1.3V like you.

Temps now 66-67C peaking at 70, 133-144W package power, fans still around 50%.

I guess I should be somewhat satisfied with 70C at 5GHz and 140W.

BTW, I can absolutely recommend the Dr. Delid tool. Because it applies pressure through rotating the IHS, there's no "pop", it comes off super easy and without much force applied at all. I think it was also cheaper that both der8auer's and Rokit's kits. Chose it mostly because the shipping was 5EU only.

--------------

70 mins in this p95 run, stable so far, +0.055V Adaptive+Offset (vcore auto BIOS). It would usually give those L0 Cache errors within 10 mins, so I assume this is at least better, if not yet actually stable, hell, I've got errors as late as 11hrs into P95 stresstesting in the past, and until I fixed those, I had occasional app crashes and even BSODs, rare, maybe 1-2 times/month, but still annoying.

What's clear is that I will better get the OC done in the winter because there's no way I can stresstest in the summer, not on this motherboard with the flimsy VRM. Peak 93C on the VRM, mostly stabilized on 90C.

Core temps looked OK I think, from what I remember the CPU would thermally throttle if I tried to run 5GHz stresstesting on it with the original TIM. Tomorrow I'll see about the 4.8GHz@1.3V fixed, see what temps I get.

Attached the 1 and 2 screenshots for the curious.
 

Attachments

Last edited:
Joined
Feb 3, 2019
Messages
1,005 (2.81/day)
Location
Chicago Land
Processor Lidless PGA 2700X
Motherboard Asus B450-I Gaming
Cooling Stock Air
Memory Corsair 3000nhz 13-15-15-15-32-53 2T
Video Card(s) Asus strix GTX 980 OC
Storage SSD
Display(s) 21" - 55"
Case None
Power Supply Antec CP series 850w
Mouse Razar Mamba Tournament Edition
Keyboard Logitech G910
Software W7
Benchmark Scores Max Cpu clock 7685Mhz FX-8300 WPrime 32m 2.886 seconds AMD 2700x
70c at 5ghz is very good, spot on TIM application, everything went just fine for you.
Tested 8700K 5000Mhz 12 threads OCCT AVX2 topped at 73c 1.3600v. /Using 1.410v hits 79c.
Before delid ! 90c @ 1.360v ~ 5000mhz very poor conductivity of thermals with that stock paste.
4.8ghz under 70c should be no problem. Everything in your screen shots look nearly identical to mine.

additional comment:
That is with paste application. I have in the cart LM for purchase, should be here by xmas.
 
Last edited:
Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
Alright, some fixed voltage numbers, 4.8GHz@1.3V, LLC is around the middle (level 4), Aida64 stress test CPU/FPU/Cache/RAM. It's not that demanding, only 120W power draw or so, which you can easily do even without delid from experience. Had 2 coolers on this PC, first one was a cheap Zalman CNPS10X Optima 120mm, one fan, and it could hold the 8700k without delidding up to maybe 130W, 85C or so. Afterwards I got this Noctua D15S. The interesting part is that the LM/delid seems to make a huge difference on AVX2 loads and higher OCs, but not that much at moderate OC like the 4.8GHz or even stock. So weirdly enough, at 5GHz/1.36V and AVX2 in LinX, the CPU would peak around 72C and draw 165W (not stable sadly, getting things stable under AVX2 stresstesting is completely another thing for these Intel CPUs). In AIDA64 4.8/1.3, with 118W-ish power draw, it peaked at 71C, which is both a meh result for the lower OC and a wow result for the higher OC. SO LM doesn't really seem to do as much for stock/low OCs?

Running LinX at 4.8GHz/1.3V is not stable either. Power draw 140W, temps still around 70. I guess there's no way to get my CPU fully stable without going nuts on the voltage. After all, Silicon Lottery uses 1.4V for 5GHz on the 8700K I believe.
Prime95 SmallFFTs with AVX is also not stable. 180W draw, worker fails immediately. The power draw for 5GHz/1.4V must be insane for SmallFFTs, 220-240W I guess.
 
Joined
Oct 21, 2005
Messages
5,718 (1.10/day)
Location
USA
System Name Small ATX Desktop
Processor Intel i5 8600K @ 4.6 GHz Core & 4.3 GHz Uncore @ 1.168V
Motherboard Asrock Z370 Taichi
Cooling Phanteks PH-TC14PE, MonoPlus, Fans: 2xThermalright TY143 / 2xCorsair SP-120L / 2xYate Loon D14BH-12
Memory G-Skill TridentZ 2X8GB DDR4 3200 CL14 F4-3200C14D-16GTZ @ 3200 14-14-14-34 2T @ 1.35V
Video Card(s) Zotac 1060 6GB Mini ZT-P10600A-10L with Arctic MonoPlus and Yate Loon D14BH-12 Fan
Storage OS Samsung 970 Pro 512GB NVMe, Games Phison E12 NVMe 1TB, HDD WD Black 4TB WD4005FZBX, 2xWD10EZEX
Display(s) 2x Asus PB258Q 2560x1440 25" IPS
Case Lian Li PC A05NB (Inverted Mobo)
Audio Device(s) Audiotechnica ATH M50X, Antlion Mod Mic 4, SYBA SD-CM-UAUD, Acoustic Research 2Ch Speakers
Power Supply Seasonic SS-660XP2 660 Watt Platinum
Mouse Zowie EC2A Mouse, Razer Naga Chroma '14, Corsair MM600, Inateck 900x300 XL pad, Tiger Gaming Skates.
Keyboard Filco Majestouch II Ninja TKL, Goldtouch GTC 0033 Ten Key, PS3 Controller
Software Win 10 LTSC 1809
Your temperature and voltage for a delidded 8700K are pretty average
 
Joined
Feb 3, 2019
Messages
1,005 (2.81/day)
Location
Chicago Land
Processor Lidless PGA 2700X
Motherboard Asus B450-I Gaming
Cooling Stock Air
Memory Corsair 3000nhz 13-15-15-15-32-53 2T
Video Card(s) Asus strix GTX 980 OC
Storage SSD
Display(s) 21" - 55"
Case None
Power Supply Antec CP series 850w
Mouse Razar Mamba Tournament Edition
Keyboard Logitech G910
Software W7
Benchmark Scores Max Cpu clock 7685Mhz FX-8300 WPrime 32m 2.886 seconds AMD 2700x
Alright, some fixed voltage numbers, 4.8GHz@1.3V, LLC is around the middle (level 4), Aida64 stress test CPU/FPU/Cache/RAM. It's not that demanding, only 120W power draw or so, which you can easily do even without delid from experience. Had 2 coolers on this PC, first one was a cheap Zalman CNPS10X Optima 120mm, one fan, and it could hold the 8700k without delidding up to maybe 130W, 85C or so. Afterwards I got this Noctua D15S. The interesting part is that the LM/delid seems to make a huge difference on AVX2 loads and higher OCs, but not that much at moderate OC like the 4.8GHz or even stock. So weirdly enough, at 5GHz/1.36V and AVX2 in LinX, the CPU would peak around 72C and draw 165W (not stable sadly, getting things stable under AVX2 stresstesting is completely another thing for these Intel CPUs). In AIDA64 4.8/1.3, with 118W-ish power draw, it peaked at 71C, which is both a meh result for the lower OC and a wow result for the higher OC. SO LM doesn't really seem to do as much for stock/low OCs?

Running LinX at 4.8GHz/1.3V is not stable either. Power draw 140W, temps still around 70. I guess there's no way to get my CPU fully stable without going nuts on the voltage. After all, Silicon Lottery uses 1.4V for 5GHz on the 8700K I believe.
Prime95 SmallFFTs with AVX is also not stable. 180W draw, worker fails immediately. The power draw for 5GHz/1.4V must be insane for SmallFFTs, 220-240W I guess.
Run system stock and test the memory then apply cpu oc.
Most errors I get are fro. Memory tweaking more so than cpu overclock, however this is my experience while I do Not use the pc everyday but for benchmarking only. So errors for me are much less concern.

Example.
4000mhz cas 14 bench stable but immediately throws errors.
Cas 15 can get 15-30 minutes before errors.
Cas 16 4000mhz much more stable longer duration without errors.
All this at stock cpu speeds during my testing.
I use Aida for latency test.
Pimod for effeciency. (32m)
This generally how I test.
OCCT AVX and AVX2 and none. Test all 3.
Then Linpack for heavier memory loads and stability testing can choose amount of memory used.
 
Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
Your temperature and voltage for a delidded 8700K are pretty average
My main preoccupation is with the delid process itself and if I applied the LM correctly (since this is the very first time I do this), not with how far I can OC the chip. To be completely fair, I mostly bought the whole thing because I was bored and wanted something new to tinker with. The 5GHz literally makes no difference with the 1070ti in the few games I rarely play, and for my work which is mostly Indesign, Photoshop, Illustrator etc., there's no way I can notice +700MHz.
The chip is either a dud or half decent I guess, depending on who you are talking with. Almost nobody uses AVX2 for stresstesting OCs, although I've seen some with incredible chips that can do 5GHz at 1.25V even with AVX. This chip is nowhere near that. I'd say it's mostly on par with the 5GHz chips Sillicon Lottery sells, that run at 1.4V and -2 AVX offset, so 4.8AVX at 1.4V basically, maybe a bit better, as for me 1.33V at 4.8GHz seem to be stable with AVX2.

Run system stock and test the memory then apply cpu oc.
Most errors I get are fro. Memory tweaking more so than cpu overclock, however this is my experience while I do Not use the pc everyday but for benchmarking only. So errors for me are much less concern.

Example.
4000mhz cas 14 bench stable but immediately throws errors.
Cas 15 can get 15-30 minutes before errors.
Cas 16 4000mhz much more stable longer duration without errors.
All this at stock cpu speeds during my testing.
I use Aida for latency test.
Pimod for effeciency. (32m)
This generally how I test.
OCCT AVX and AVX2 and none. Test all 3.
Then Linpack for heavier memory loads and stability testing can choose amount of memory used.
The memory is a super weird ADATA wityh Spektek chips, it gave me 1 year of troubles, without exaggeration. I was unable to get things stable at its XMP of 3000MHz and CL16 at all for months, because I made the mistake to try and increase voltages. It was exactly the DRAM voltage that made things unstable. It's at 1.21V now and manually entered XMP timings, fully stable, no app crash in months, 0 BSODs, 48 hrs of stresstesting in both LinpackX and Blend with AVX2 and 10GB RAM used, also with the memory test suggested by ASUS inside Linux.
In short, the RAM is rock solid stable (now...). Had to compromise in a few places to get everything I wanted when I got this in 2017, and this RAM was cheapest, I guess for obvious reasons.

I think I will see if everything remains stable at +0.055V adaptive and no AVX offset, 5GHz, in day to day usage. If I see apps crashing or BSODs I'll probably use an AVX offset. The CPU seems to have dropped 25C or so at higher OC/power draws (150-160W+). I remember very well having it thermally throttle if I tried SmallFFTs no AVX at 5GHz/1.33V, reaching 100C. Now it tops at 74C max core temp and the core averages are 68-72. Package temp peaks at 75C like this, average 72C. The -25C seems in line with the expectations for LM I guess?

Below is SmallFFTs no AVX, 5GHz/+0.060V, increased it a bit to make the CPU hotter, see how the LM does.

 
Last edited:
Joined
Oct 21, 2005
Messages
5,718 (1.10/day)
Location
USA
System Name Small ATX Desktop
Processor Intel i5 8600K @ 4.6 GHz Core & 4.3 GHz Uncore @ 1.168V
Motherboard Asrock Z370 Taichi
Cooling Phanteks PH-TC14PE, MonoPlus, Fans: 2xThermalright TY143 / 2xCorsair SP-120L / 2xYate Loon D14BH-12
Memory G-Skill TridentZ 2X8GB DDR4 3200 CL14 F4-3200C14D-16GTZ @ 3200 14-14-14-34 2T @ 1.35V
Video Card(s) Zotac 1060 6GB Mini ZT-P10600A-10L with Arctic MonoPlus and Yate Loon D14BH-12 Fan
Storage OS Samsung 970 Pro 512GB NVMe, Games Phison E12 NVMe 1TB, HDD WD Black 4TB WD4005FZBX, 2xWD10EZEX
Display(s) 2x Asus PB258Q 2560x1440 25" IPS
Case Lian Li PC A05NB (Inverted Mobo)
Audio Device(s) Audiotechnica ATH M50X, Antlion Mod Mic 4, SYBA SD-CM-UAUD, Acoustic Research 2Ch Speakers
Power Supply Seasonic SS-660XP2 660 Watt Platinum
Mouse Zowie EC2A Mouse, Razer Naga Chroma '14, Corsair MM600, Inateck 900x300 XL pad, Tiger Gaming Skates.
Keyboard Filco Majestouch II Ninja TKL, Goldtouch GTC 0033 Ten Key, PS3 Controller
Software Win 10 LTSC 1809
Yes, I mean to say for a delidded 8700K, it is performing average as in not poor and not spectacular. I state this to reassure you as your title asks "are the temps OK?"

Your temperatures are where one would expect them to be. I would go for the max stable AVX2 and not use the offset.

Also, I am skeptical about the Silicon Lottery stats and I also think they sell a lot of poor to average CPUs as 'high binned' under the guise of AVX offset.
 
Joined
Feb 3, 2019
Messages
1,005 (2.81/day)
Location
Chicago Land
Processor Lidless PGA 2700X
Motherboard Asus B450-I Gaming
Cooling Stock Air
Memory Corsair 3000nhz 13-15-15-15-32-53 2T
Video Card(s) Asus strix GTX 980 OC
Storage SSD
Display(s) 21" - 55"
Case None
Power Supply Antec CP series 850w
Mouse Razar Mamba Tournament Edition
Keyboard Logitech G910
Software W7
Benchmark Scores Max Cpu clock 7685Mhz FX-8300 WPrime 32m 2.886 seconds AMD 2700x
Where can we get Liquid Metal?
Im looking at newegg. But sure any water cooling site or PC hardware sites.
Its not cheap though.

Thermal Grizzly kit.
 
Joined
Feb 18, 2009
Messages
347 (0.09/day)
Processor i7 8700K
Motherboard MSI Z370 Gaming Plus
Cooling Noctua NH-D15S + NF-A12x25 PWM + 4xNF-A14 PWM
Memory 16 GB Adata XPG Dazzle DDR4 3000 MHz CL16
Video Card(s) Gigabyte GTX 1070 Ti Gaming 8G
Storage Samsung 970 EVO Plus, Samsung 850 Evo
Display(s) Samsung C24FG73 144Hz 1080p
Case Fractal Design Meshify C
Audio Device(s) Steelseries Arctis 3
Power Supply Superflower Leadex II Gold 650W
Mouse Steelseries Rival 600
Keyboard Steelseries Apex M750
Software Windows 10 x64 Pro
Realbench is quite hardcore, no AVX offset @5GHz. Looks like Adaptive recognizes AVX workloads and bumps the voltage, interesting. For example, stresstesting without AVX seems to get a Vcore of 1.328, and AVX workloads result in 1.368V. To make things even more weird, the Realbench workload seem to produce a very large amount of CPU heat, without a very large power draw, which is unexpected. The heat output always seem to match very well the "Package Power Draw" reading. Not for Realbench though, which "only" seems to run at 138-142W, yet it results in +8-10C temps, with peaks of 79C. SmallFFTs nonAVX runs at consistently higher power draw, 155W or so, yet the temp peaks were approx. 5C lower, while averages were less dramatic.

The power draw seems to be legit, as the VRM temps match it, they're 10C lower with Realbench @135-140W compared to SmallFFTs @155W.


EDIT:

Alternative for OC - running at stock speeds, undervolted, on an (almost) fully quiet PC. Fans at most 33%. Peak temps 63C, during SmallFFTs, with AVX2, which results in about 118W power draw. Fully stable.

 
Last edited:
Top