Jump to content
Corsair Community

i7 7700k @5.00Ghz temp blance between 70 and 98C


Alexis39
 Share

Recommended Posts

Hi !

 

I overclocked my i7 7700k at 5.00Ghz with static Vcore at 1.300 (but Intel Tuning Extreme show 1.36V sometime ... but I think it's VID, not VCore) and I use a h100i V2 to cool it.

 

I get peak at 98C according to CPUID Package temp.

 

Here are 2 screen to show the stress test from Intel Extreme Tuning, you can see the peak and low temp of the CPU + VCore. Cooler Temp is about 35~37C

 

https://drive.google.com/open?id=1bhi6WY6pT5hW6VuEoapSnP6TV7qL-JiF

https://drive.google.com/open?id=1RRo8UE0SZzrSAgP67G-u3D0MF_zQPeEE

 

Also, the 2 fan of my radiator, connected with the Y cable doesn't match the Liquid cooler temp but the CPU temp even I choose "H100i V2 cooler temp" group on the fan setting in the Corsair Link app. anything I missed ?

 

Thank you for your attention !

Edited by Alexis39
Link to comment
Share on other sites

That's far too warm.

Can you post some images of your Link configuration? What are your pump and fan speeds? That's not on those screen shots and that'll be far more useful in diagnosing what's going on.

 

Hi ! Thank you for your quick answer.

 

I did 3 test: (I tested both with the h100i V2 profil on Quiet and Performance but nothing really changed)

 

Intel Extreme Utility at base boost clock speed (4.6Ghz)

https://drive.google.com/open?id=12715uDKaUGnk85ZLfW1xMobvJZjJ-txd

 

Intel Extreme Utility OC at 5.00Ghz

https://drive.google.com/open?id=1E2ywnzJW8g5Kk1hos2dSQNiYxiNRB9fR

 

You can find the max reached temp on the CPUID windows and a graph on the Corsair Link window.

 

and IntelBurn at 5.00Ghz

https://drive.google.com/open?id=1A-cpQNkym8_Kvz0iI6EkHTihS_K6ER_W

 

The IntelBurn test really push the CPU to the limit and reached 100C so the h100i V2 turned to max throttle but temps didn't change so I just stopped it.

Link to comment
Share on other sites

What is your pump set to? You shouldn't be seeing that kind of variation in pump speeds. That kind of variation in pump speeds is usually an indication that the pump is not being powered properly. You need to set the header to 100%/Full Speed.

 

Second ... your coolant temperatures won't rise anywhere near as quickly as your CPU temperature. That's because the coolant has a higher heat capacity ... it takes a while to get to max temperature, even with a hot CPU.

 

I would recommend disabling any and all overclock until we get the thermal issues resolved.

 

We aren't seeing your Max VCore on any of those shots either. But I did see a 1.30V vCore at 1% utilization. That's far too high. But then ... there was also a 1.25 vCore @ 99%. So that's a little odd.

Link to comment
Share on other sites

The H100i v2 needs 12v at all times. Don't use the BIOS to control pump or fans, use Link instead. That part is resolved, but with coolant temps only in the low 30's, there always was another issue.

 

Seems like a big swing between XTU and IBT temps. Certainly not unusual since they are on opposite ends of the spectrum, but it may suggest the origin of the problem. If you hashed the mount or TIM, both would equally terrible and immediately terminable. The disparity usually suggests a need to refine your BIOS settings. This is particularly true with Kaby and Coffee. There is no more pop it in a let it rip or you will have a truckload of voltage bearing down on you. I thought I saw 1.37v in one of the VID requests. What is your Vcore set to? Adaptive? Manual?

 

If adaptive or auto, you will need to address the IA/DC loadline settings. If you have not read it before, take a look at the Asus Kaby Guide. AI/DC on page 3.

 

Package temp line graph looked a little warbly, but I have not run XTU on 7700/8700k yet. That may not be out of character for those.

 

EDIT: I went back and ran XTU on my 8700K@5.0/1.30v. My core and package lines are parallel, smooth, and follow the traditional sine wave pattern. Yours looks a little jagged. That might be TIM, but it could also be the difference between a delidded and dialed in BIOS and the stock IHS plus stock voltage behavior.

Edited by c-attack
Link to comment
Share on other sites

I've never seen a 7700K not jump around temp-wise like an insane bean. The heat bursts it has, especially with the poor TIM below the lid, are infamous. But it really shouldn't be hitting 98 regardless. Though it "can" be okay at 98 according to Intel, it just feels wrong.
Link to comment
Share on other sites

LLC was on Auto and I changed it to 5. @4.8Ghz 1.3v on IBT, max temp was 92C. Idle temp are between 32 and 42C.

 

I did my TIM 3 times because of those strange temp but nothing really changed. I think the only thing I didn't follow was to remove the previous thermal paste with special tim removal liquid (I used colon stick and toilet paper) ... maybe I should do it again, I actually have the Arctic liquid (bought just "in case I need to do it again"). about the cooler mount, I turned the screw with my hand and gave one more turn with the screw driver.

Edited by Alexis39
Link to comment
Share on other sites

LLC was on Auto and I changed it to 5. @4.8Ghz 1.3v on IBT, max temp was 92C. Idle temp are between 32 and 42C.

 

I did my TIM 3 times because of those strange temp but nothing really changed. I think the only thing I didn't follow was to remove the previous thermal paste with special tim removal liquid (I used colon stick and toilet paper) ... maybe I should do it again, I actually have the Arctic liquid (bought just "in case I need to do it again"). about the cooler mount, I turned the screw with my hand and gave one more turn with the screw driver.

 

That matches what I and other folks have seen on a custom loop without delidding.

 

TIM Removal Liquid: 90+% Rubbing Alcohol. :P XD I wouldn't advise using toilet paper though. Too much chance for leaving fibers/lint behind.

 

You could possibly squeeze more heat reduction out of it, but at this point it would involve experimenting with voltage (Keep it stable!) or involve delidding (In which case go with CooLaboratory's or ThermalGrizzly's liquid metal thermal solution under the lid for decent results.)

 

If you do delid, get a reputable delidding tool for safety unless you're comfortable doing it manually with the risk of destroying the CPU. The tool makes it "Still scary, but if you follow the directions, no longer (as) risky"

Link to comment
Share on other sites

What happened to peak Vcore when you adjusted LLC to 5? It would be good to differentiate between "can't run IBT" and an actual physical problem. Leave IBT alone, particularly since it is unlikely have any real application for you. XTU is a better tool when you might have an issue and you won't fry your CPU trying to figure it out. Try running XTU again on LLC 5. Before you start, go into the setting wrench on the graph and enable core 1-4 to show along with package temp. It would like to see if there is any difference under load.

 

Aside from IBT, what are your normal use temps like?

Link to comment
Share on other sites

When I play game, max CPU temp was 85C (Playing Ark Survival). Seems like XTU get about the same temps.

 

here is the test with 4Core showing, LLC at 5, VCore on Adaptive: vary between 1.18v and 1.2v. Screenshot took at the end of 5min stress test.

 

https://drive.google.com/open?id=19g_DKaoiBbSl4-sNS6jsqCeHOKT5MHUv

 

About deliding, I don't want to take the risk/ void the waranty as it won't make a huge difference in game (with frequency @5.0 ~ 5.2). I'll redo my TIM this evening in ~5h (Japan time) and tell you how it goes :)

 

Thank you all for your help ! I appreciate :)

Edited by Alexis39
Link to comment
Share on other sites

Your max vCore was 1.296. You seem to be confusing VID with VCore.

 

With that vCore, your temps are about right for a stock Kaby Lake i7. Your particular draw in the Silicon Lottery may not be the best sample but it's far from the worst either. My guess is that you have a little extra or not quite enough of the craptastic Intel TIM. You won't get much better temps without delidding.

 

Completely understand that you don't want to take the risk. That said, delidding dropped my stock temps by 20C. You may be able to dial down the vCore a bit, which will help. Even .005V could drop you a degree or two. And your CPU seems to think that it doesn't need quite so much voltage (that's what VID is ... the requested voltage) ... assuming HWMonitor is reading accurately.

Link to comment
Share on other sites

I think you can skip another TIM go if you want. When that is the issue, there is normally more random variability in the rolling waves. If there is a contact issue here, it is more likely to be centered around the backplate. The curve peaks seem just a little too sharp to me.
Link to comment
Share on other sites

So I redone my TIM and nothing changed.

 

Then in Intel XTU, I changed the VCore offset to -0.025 but nothing change ... -0.050 nothing again ... I check the VCore reading on HWMonitor and it doesn't change ! (I used to watch the VID but, as DevBiker said, I was confusing both: XTU says "Core Voltage" but that just the VID !)

 

So @48x I went to the BIOS and change Core Voltage to Manual and tested how low I could go: just done a stress test for 20min @4.8 1.225v and my temp don't get higher than 85C. Didn't really try @5.0 as it didn't boot at 1.225v (1.25v will probably bring me to 90+ temps and it doesn't make huge difference in game)

 

Thanks all for your help :)

Link to comment
Share on other sites

If you are on manual/fixed voltage, the LLC setting is really the only other thing to affect the actual Vcore value. On auto or adaptive, there are several more. I guess since I don't use HWMonitor, I am not as able to be sure on the results as something like HWiNFO or AIDA. HWINFO sensors have a LOT of data. A screen shot of that during or after the XTU test might put a nail in this subject.

 

Have you been using XTU to alter the BIOS settings? Or going directly into the BIOS? I am not sure the XTU settings will always take and software overclocking/tweaking is always a bit dicey.

Link to comment
Share on other sites

Excellent info and very informative.

Note the individual core temps ... there is a 10C delta between the coolest and the hottest. That's likely due to uneven TIM applied by Intel during the manufacturing process. Note that there is also some imprecision in the measurements, although it does get more accurate as it gets hotter.

 

This is pretty much inline with what I saw from both my 6700 and 7700 CPU before delidding.

Link to comment
Share on other sites

Yup, so that ends the voltage overshoot speculation. That really leaves two possibilities. As described above, this is the performance of your CPU when overclocked. It is pretty amazing to me there is a near 30C difference between your 7700K@4.8/1.20v and my 8700K@5.0/1.30v, but then I guess that is the value of delidding. Presumably, it is also Summertime where you are, so this is bad as it gets.

 

The last thing you might want to verify is there backplate is positioned properly and turned the right way. A very slight gap might cause a general increase in all CPU temps. You might be able to have previously seen this in the TIM spread. Uneven pressure or a misaligned backplate would probably leave sections of paste not spread evenly or left as chunks on the CPU lid. I would not fiddle with this too much, but a general check of the backplate position might be a good idea.

Link to comment
Share on other sites

Yup, so that ends the voltage overshoot speculation. That really leaves two possibilities. As described above, this is the performance of your CPU when overclocked. It is pretty amazing to me there is a near 30C difference between your 7700K@4.8/1.20v and my 8700K@5.0/1.30v, but then I guess that is the value of delidding. Presumably, it is also Summertime where you are, so this is bad as it gets.

 

The last thing you might want to verify is there backplate is positioned properly and turned the right way. A very slight gap might cause a general increase in all CPU temps. You might be able to have previously seen this in the TIM spread. Uneven pressure or a misaligned backplate would probably leave sections of paste not spread evenly or left as chunks on the CPU lid. I would not fiddle with this too much, but a general check of the backplate position might be a good idea.

 

 

Thanks for your answers guys.

 

I took pictures of the cooler before I clean it (but forgot the CPU)

 

You can see how it was before and I probably applied the same way before. (I think it's a bit spreading more on one side because the tube for the liquid coolant push the pump a bit)

 

https://drive.google.com/open?id=1GlNBXo0YI3Vcx9DIx8PDgVqkX6me_o30

https://drive.google.com/open?id=1R1ykeHnhgleT9EvLpkguzxQNMxk7HO-d

 

As you answered, I guess I don't a good pick at the silicon lottery but it should be fine, it's not getting that hot in game.

Link to comment
Share on other sites

 Share

×
×
  • Create New...