[SOLVED] i9-9900KS Lowering Temp & Undervolt settings check

Jun 2, 2020
1
0
10
0
So I'm a bit on the OCD type and easily worried, my gaming temp sometimes can hit as far as 94c on stock settings BIOS. Which is totally not cool at all, so I take an initiative to learn undervolt but yeah you can easily find OC guide result more often then undervolt and i9-9900k as dominant result compare to i9-9900KS. Therefore, I decided to use OC settings but lowering the Vcore w/o disturbing the clock ratio , so I remain it as stock 5ghz.

So I begin my test with the settings I used (can find the screenshots below):
*note that I live in hot climate , temp during day around 31-34c.
*checked on stock BIOS settings my stock Vcore is 1.404v which is I think is high

@1.3v = No BSOD on cinebench r20 and during gaming on heavy tittle like BFV and COD Warzone Didn't test on P95 because it is very overkill stress test. No aida test.

@1.28v = No BSOD during gaming and temp basically stays lower than 77c sometimes spike to reach 81-82c, Cinebench R20 temp is around 90c if I'm not mistaken. No aida test.

@1.25v = BSOD after gaming for few hours . Cinebench R20 passed . Aida64 stability test lasted for 1hours and 15mins then BSOD at max temp 95c no throttling.

@1.26v = No BSOD to date , feels like not doing aida64 test since gaming on 5-7hours nothing happened . Temp is also good at under 65c only sometimes spike 71-72c then maintain back under 65c . Idle temp is 5c more then my stock BIOS settings, which around 42-44c. If BIOS settings depends on liquid temp, if 35c on liquid temp usually CPU at 36c or 37c.

Screenshots(Album) current BIOS settings: View: https://imgur.com/a/aMs4Mg4


My main goal is to lower my CPU temp so my question is ,
  1. Is my BIOS settings okay?
  2. If I wanna pass aida64 stability test what changes is required?
  3. I read on some post that I need to change the value for VCCIO and VCCSA, but I don't have any experience on this or so. Please enlighten me on this.
  4. If there's any wrong settings/value please let me know I will change it.
Full Specs:
  • i9-9900KS
  • Z390 Aorus Master | BIOS version F11c (Latest)
  • G.Skill Trident Z 32GB(8GBx4) 3200Mhz CL14 (8GBx2 Trident Z RGB + 8GBx2 Trident Z NEO) B-Die
  • NZXT Kraken x72 mod with Noctua NF A12x25
  • RTX 2070 Super
  • EVGA G+ 1000w
  • Lian Li PC-O11 Dynamic top mount radiator as exhaust others(side and bottom) as intake.
 

CompuTronix

Intel Master
Moderator
HVNtime,

It looks like you've done a good job so far, however, Phaaze88 is correct; Prime95 is not overkill when used correctly. P95 v29.8 Small FFTs with all AVX test selections disabled is a steady-state 100% workload which conforms to Intel's Datasheets for valid thermal testing. As per the Datasheets, TDP and Thermal Specifications are validated “without AVX.

If you prefer, Prime95 v26.6 doesn't have AVX, but its Small FFTs is the same workload as v29.8 Small FFTs without AVX. If OCCT's first CPU test, called "OCCT", is configured for Small Data Set and No AVX, then it's a steady-state 97% workload that's nearly identical to Prime95's Small FFTs without AVX.

A steady-state workload is key for thermal testing so the PSU, VRMs, CPU, socket, motherboard and cooler can thermally stabilize for steady Core temperatures. This is how Intel eliminates software variables to develop thermal specifications.

AIDA64 has 4 CPU related stress test selections (CPU, FPU, Cache, Memory) which have 15 possible combinations that yield 15 different workloads and 15 different Core temperatures. That's a lot of variables.

The individual FPU test is about 115% TDP workload, the CPU/FPU combination is about 90%, all 4 tests combined is about 80% and the individual CPU test is only about 70%. All other AIDA64 test selections are fluctuating workloads which are suitable for stability testing, but not for thermal testing.

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Utilities that don't overload or underload your processor will give you a valid thermal baseline.

Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. As you can see at the top of the scale, if P95 is configured for AVX, then it's a brutal 130% workload. Without an AVX offset to reduce power consumption, Core temperatures skyrocket.

If you only game and don't anticipate ever using your rig for heavy CPU workloads such as rendering and transcoding which use AVX, then there's no need to concern yourself with configuring an AVX offset. Regardless, keep in mind that the AVX code that's used in recent games is much less demanding, and won't approach or exceed Prime95 Small FFTs without AVX.

For a better perspective, use HWiNFO "Sensors Only" to observe Vcore, Package Power (Watts) and Package temperature (hottest Core) during testing. Numbers, even when logged, don't show the big picture, so right click on those parameters to see the graphs. Remember that frequency, Vcore and workload drive power consumption, which in turn drives Core temperatures.

CT :sol:
 

Phaaze88

Titan
Ambassador
P95 isn't overkill. It is used to test cooler thermal stability - small FFT, AVX off. Make sure error reporting is checked in the options tab.
Cinebench R20 'infinite loop' covers voltage stability.

1)You should use an AVX offset of 2 or 3. Some applications and stress tests do use AVX instructions, and that running at the same frequency as the normal SSE instructions = :hot:
They also draw more power and require more voltage for stability VS SSE.

2)I'm not that familiar with Aida64, so someone else would have to answer this one.

3)They can help with memory overclocking, like 4000mhz or so. Not necessary for 3200mhz.

4)Just the AVX offset from what I can tell.
 

Danra

Distinguished
May 25, 2005
526
20
19,365
115
Temperatures on my i9 9900KS are 29 degrees coolest core to 33 degrees C hottest core while I type this, room temp is 24 degrees C. Also while typing this I did 11 runs of IntelBurnTest v2.54, maximum spiked temperatures: coolest core 74 degrees, hottest core 84 degrees C. Cooler is Noctua NH-U14S air cooler, case is Cooler Master HAF 922 with three 200mm fans, one 140mm and one 120mm. I use HWiNFO64 to monitor temps, voltages, etc. Volts from IR35201 VR OUT was 1.342 at idle and from 1.287 to 1.318 while IntelBurnTest was running.

My Vcore is much higher than yours. I use an AVX offset of 0, however I set my CPU at 4.8GHz all core because I cannot tell the difference between 5GHz and 4.8GHz in games and I want my CPU to last, it is not the best silicon, relatively poor in fact for an i9 9900KS and I had a devil of a time to get it to run stable, not something you need to worry about..

Motherboard is the same as yours F11c BIOS, 32GB G.Skill 3600 RAM

I used a variation of BIOS approach by another person, can't remember whom or a link for a how-to. I will get a USB drive and try to take some screen shots of my BIOS settings if you like.
 

Danra

Distinguished
May 25, 2005
526
20
19,365
115

CompuTronix

Intel Master
Moderator
HVNtime,

It looks like you've done a good job so far, however, Phaaze88 is correct; Prime95 is not overkill when used correctly. P95 v29.8 Small FFTs with all AVX test selections disabled is a steady-state 100% workload which conforms to Intel's Datasheets for valid thermal testing. As per the Datasheets, TDP and Thermal Specifications are validated “without AVX.

If you prefer, Prime95 v26.6 doesn't have AVX, but its Small FFTs is the same workload as v29.8 Small FFTs without AVX. If OCCT's first CPU test, called "OCCT", is configured for Small Data Set and No AVX, then it's a steady-state 97% workload that's nearly identical to Prime95's Small FFTs without AVX.

A steady-state workload is key for thermal testing so the PSU, VRMs, CPU, socket, motherboard and cooler can thermally stabilize for steady Core temperatures. This is how Intel eliminates software variables to develop thermal specifications.

AIDA64 has 4 CPU related stress test selections (CPU, FPU, Cache, Memory) which have 15 possible combinations that yield 15 different workloads and 15 different Core temperatures. That's a lot of variables.

The individual FPU test is about 115% TDP workload, the CPU/FPU combination is about 90%, all 4 tests combined is about 80% and the individual CPU test is only about 70%. All other AIDA64 test selections are fluctuating workloads which are suitable for stability testing, but not for thermal testing.

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Utilities that don't overload or underload your processor will give you a valid thermal baseline.

Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. As you can see at the top of the scale, if P95 is configured for AVX, then it's a brutal 130% workload. Without an AVX offset to reduce power consumption, Core temperatures skyrocket.

If you only game and don't anticipate ever using your rig for heavy CPU workloads such as rendering and transcoding which use AVX, then there's no need to concern yourself with configuring an AVX offset. Regardless, keep in mind that the AVX code that's used in recent games is much less demanding, and won't approach or exceed Prime95 Small FFTs without AVX.

For a better perspective, use HWiNFO "Sensors Only" to observe Vcore, Package Power (Watts) and Package temperature (hottest Core) during testing. Numbers, even when logged, don't show the big picture, so right click on those parameters to see the graphs. Remember that frequency, Vcore and workload drive power consumption, which in turn drives Core temperatures.

CT :sol:
 

ASK THE COMMUNITY

TRENDING THREADS