9700K Temps

[-Synchronicity-]

The 95w is a joke. How can a 8 core cpu take less power then a 6 core cpu(8700k) rated at 150w with higher clockspeeds? Logic much? On the same 14nm no less.

 

[InvincibleTofu]

iirc, Intel will refer you back to retailer cos it will talk to you only if it is past 30 days post purchase.

Hmm.. I'll give my retailer a call. See if I can swap the chip.

 

[Kairos]

Best order a binned 9700K from Silicon Lottery?

 

[InvincibleTofu]

Best order a binned 9700K from Silicon Lottery?

Then what do I do with this one?

 

[Kairos]

Then what do I do with this one?

Scenario 1: RMA successful; sell BNIB at a good price.

Scenario 2: RMA unsuccessful; sell USED at not so good price.

 

[InvincibleTofu]

Damn sian to get a new rig only to encounter such issues.

 

[LiLAsN]

Have you tried saving your current BIOS profile settings and then load the default Bios settings and turn off MCE instead?


This is mine when in a game in Overwatch. Running at 240 FPS (my custom fps limit) but at 1080p from my second monitor.
Room set to game mode so aircon automatically set to 23°C for room ambience.

Please ignore fan speeds. NZXT cam is a bit screwy when there are 2 NZXT Kraken X62s in the system.
They displayed my GPU fan speed in the CPU fan speed instead.
Similarly, I disabled the FPS counter. So the 217 FPS for OW is based on a very old gameplay. Not the current one. So ignore that as well.

But both CPU and GPU are being cooled by NZXT X62.
Radiator placed at the worst spot for CPU cooling in the Evolv X case which is the top.


So stats were 20% load +/- across the cores with average frequency being 4.7 and temps were averaging at 36°C with the 9900K.


GPU might show 99% load but graphics were set to low. When gaming in AC Odyssey, temps go higher than 36°C.

 

[InvincibleTofu]

Have you tried saving your current BIOS profile settings and then load the default Bios settings and turn off MCE instead?


This is mine when in a game in Overwatch. Running at 240 FPS (my custom fps limit) but at 1080p from my second monitor.
Room set to game mode so aircon automatically set to 23°C for room ambience.

Please ignore fan speeds. NZXT cam is a bit screwy when there are 2 NZXT Kraken X62s in the system.
They displayed my GPU fan speed in the CPU fan speed instead.
Similarly, I disabled the FPS counter. So the 217 FPS for OW is based on a very old gameplay. Not the current one. So ignore that as well.

But both CPU and GPU are being cooled by NZXT X62.
Radiator placed at the worst spot for CPU cooling in the Evolv X case which is the top.


So stats were 20% load +/- across the cores with average frequency being 4.7 and temps were averaging at 36°C with the 9900K.


GPU might show 99% load but graphics were set to low. When gaming in AC Odyssey, temps go higher than 36°C.

Woah those are really good temps.

My rig is stable during gaming.

It is AVX loads that worry me. I trend towards 100 Deg C at stock speeds/auto vCore during AVX loads because Core 7 requires a ton of juice to prevent faults.

I feel like it is not unreasonable to be able to expect the chip to be stable at 100% AVX loads at stock given that I am liquid cooling.

But I will make some tweaks at home later and post screenies.

 

[Encrypted11]

These things when used together produce a noticeable temp improvements if you're on full auto from any particular board vendor. You can skip some of them if necessary, but the more you skip the lesser the temperature improvement.
Also if you don't set the appropriate power limit manually, fact is you are running auto power limit overclocks set by mb vendor. It's just like precision boost overdrive.

Actual defaults
-Short Duration Power Limit: 119W
-Long Duration: 95W
-Current Limit Max: 193A.

System Agent / PCIe / IMC
-Check for excessive VccSA VccIO (Spec'd at 1.05/0.95 bef xmp)
-Some vendors default with 1000MHz system agent while intel defaults at 800MHz per intel spec (so your SA/IO needs more volts)

Cores - Power Management & Core Parking
-EIST + SpeedShift enabled, at least up to C6 (C1E optional)
-Use default per core frequency, than synced all cores. Even if you place them all on 50 for example. So the core clocks move asynchronously depending on usage.
-Adaptive/DVID/Offset vcore mode
-PLL Bandwidth/CPU PLL/VccPLL (Check for the equivalents of ur mb vendor because of different terminology. Should be 1-1.2v. Has an impact on core oc, to lock high multipliers.

At full defaults per Intel spec, input manually..
the 1-8 core active ratios

The CPU is not faulty like what elmariachi conveying. If you'd try RMA with Intel the reality is the agent will want to look at your IPDT/XTU details which will reveal motherboard vendor set default OCs if you do not sanitise them you'd hit more troubles with the support agent. If you're running an all-core MCE the agent will spot it and refuse your warranty claims fyi, word of advice given..

 

[InvincibleTofu]

These things when used together produce a noticeable temp improvements if you're on full auto from any particular board vendor. You can skip some of them if necessary, but the more you skip the lesser the temperature improvement.
Also if you don't set the appropriate power limit manually, fact is you are running auto power limit overclocks set by mb vendor. It's just like precision boost overdrive.

Actual defaults
-Short Duration Power Limit: 119W
-Long Duration: 95W
-Current Limit Max: 193A.

System Agent / PCIe / IMC
-Check for excessive VccSA VccIO (Spec'd at 1.05/0.95 bef xmp)
-Some vendors default with 1000MHz system agent while intel defaults at 800MHz per intel spec (so your SA/IO needs more volts)

Cores - Power Management & Core Parking
-EIST + SpeedShift enabled, at least up to C6 (C1E optional)
-Use default per core frequency, than synced all cores. Even if you place them all on 50 for example. So the core clocks move asynchronously depending on usage.
-Adaptive/DVID/Offset vcore mode
-PLL Bandwidth/CPU PLL/VccPLL (Check for the equivalents of ur mb vendor because of different terminology. Should be 1-1.2v. Has an impact on core oc, to lock high multipliers.

At full defaults per Intel spec, input manually..
the 1-8 core active ratios

The CPU is not faulty like what elmariachi conveying. If you'd try RMA with Intel the reality is the agent will want to look at your IPDT/XTU details which will reveal motherboard vendor set default OCs if you do not sanitise them you'd hit more troubles with the support agent. If you're running an all-core MCE the agent will spot it and refuse your warranty claims fyi, word of advice given..

Thanks. Will try those tweaks when I'm home later.

It seems disingenuous of Intel to work with mobo manufacturers who almost all turn on MCE by default, then state that running the system in that state technically counts as overclocking and voids chip warranties.

 

[Encrypted11]

System agent is usually FCLK on early power on or FCLK frequency on most vendors.

At XMP-3200 type memory overclocks assuming 800MHz FCLK, I believe under 1.15v sa/io would already be more than enough for most if not all cpu samples. Some vendors like to set autos to 1.2-1.3v+ on higher than 3200 kits.

 

[InvincibleTofu]

System agent is usually FCLK on early power on or FCLK frequency on most vendors.

At XMP-3200 type memory overclocks assuming 800MHz FCLK, I believe under 1.15v sa/io would already be more than enough for most if not all cpu samples. Some vendors like to set autos to 1.2-1.3v+ on higher than 3200 kits.

I agree that XMP auto profiles have overly aggressive voltages as well.

But work on processor first =s

Once that's sorted I'll tighten the RAM voltages.

Honestly, I never thought it could be this much "fun" to try to run an efficient system. My goal isn't to overclock, stock with cool temps and lower power draw works for me.

 

[kimsix]

MCE should not be issue.

it is the XMP and auto-vcore settings of some boards, very aggressive. dont use XMP and use adaptive vcore set the voltage around native values +/- to get stable OC is enough.

asus bios was best for haswell/skylake and before, not sure is same for new generation.

 

[lifeishard]

MCE should not be issue.

it is the XMP and auto-vcore settings of some boards, very aggressive. dont use XMP and use adaptive vcore set the voltage around native values +/- to get stable OC is enough.

asus bios was best for haswell/skylake and before, not sure is same for new generation.

Huh,don't use xmp? So ram to run at default base of 2666mhz?

 

[kimsix]

Huh,don't use xmp? So ram to run at default base of 2666mhz?

just manually switch the ram speeds and change ram timings if needed. xmp is auto-oc, some bios also auto up the voltage with it.

 

[Outramkia]

Huh,don't use xmp? So ram to run at default base of 2666mhz?

Manual adjust clock and voltage lor

 

[InvincibleTofu]

This is getting frustrating. I had a WHEA error once after I reset BIOS then disabled MCE.

XMP was off as well.

But BSOD means no screenshot

 

[elmariachi]

Looks either to me like board problem or memory or BOTH. I'm having almost the same setup as you and have no issues so far. I seriously doubt it is the CPU man.

 

[LiLAsN]

Looks either to me like board problem or memory or BOTH. I'm having almost the same setup as you and have no issues so far. I seriously doubt it is the CPU man.

Since BIOS saved profiles are transferable, why not you grab the saved profile from the one you upload to your USB and pass it to him to see if it helps?


For me, the last time my 6700K got super hot, it turns out, the pump in the cheap china brand AIO died.


Try googling the exact problem like WHEA CPU error fix.
Some say you need to increase voltage. Who knows.

For me, I am currently running the Asus AI overclock. So I just let the AI decide what's the best Overclock for my CPU. The AI will update the clocks/frequency if need be upon every boot up of your PC. Quite convenient for me.

The latest 0602 update for the Mobo lets me idle at around 50+W when checking the wattage being pulled from the wall. Not bad considering I have a few software and chrome open.

 

[InvincibleTofu]

Mmm.. Yea this is really weird.

I'm pretty sure the pump is working on the X62.

At this stage I still think the most likely culprit is the CPU since that is the component that fails the stress test most often.