h100i water cooler-- installed wrong?

Joined
Mar 23, 2013
Messages
53
Hi--
I built my rig a few weeks ago, but haven't done anything special with it.
Not that experienced at this.

I have a 3770k cpu, asrock extreme4 motherboard, and h100i water cooler, all installed in an Antec900. The h100i is installed in the front; with some tinkering, you can get it mounted in a push-pull set up with the existing 120mm fans there.

I just installed core temp, and it idles at 25-35C (no load), but when I ran prime95, it shot to 80C, then kept on climbing after a few minutes to 95C!
Granted, it's loading all 4 cores to 100%. The 4th core remained cooler, maybe 80C; cores 1-3 got quite hot though.

Is coretemp inaccurate?
Or... is something installed wrong? Like too much thermal paste or poor contact? I feel a $100 water cooler shouldn't let things get that hot!

The case has 4 120mm fans in line (blowing the same direction, before you suggest that sort of mistake) in front, 1 240mm fan on top, a 120mm in back, and a 120mm on the side.
 
the h100i came with thermal paste, but, I had to reinstall it (DOA motherboard, and to remove cpu to RMA it), so I wiped it all off and reapplied some arctic silver.
I used a razor blade like a fine putty knife, to get what I thought was a thorough but thin layer.
 
alright-- pulled off the cooler block.
found 50% contact.

Wiped it all down,
Tried again: very thin and even layer on the cpu, with a credit card, tightened the nuts in an balanced criss cross pattern.

same results!

I actually counted the minutes more accurately; within 2 minutes of prime95 and 100% usage, core 2 got to 98C. I shut it off. Cores 1 and 3 were at 92, 94C. Again, core 4 was coolest at 85C.

Should I remove the cooler block again?
 
tried again for 5 minutes.
coretemp says it reached the same levels-- 98C on core 2, but steady.

this time also ran speedfan
speedfan says 59C :) HWMonitor is another program, and seems lock-step with speedfan-- must be getting the same number from the same source. Also says 59C
speedfan does not break it down by cores: it just says "cpu" is 59C

Err, which one's right? The one that says I'm doing great for 100%cpu, and time to overclock! Or the one that says my chip is about to liquefy?
 
Last edited:
I have the same MB, CPU and cooler.

Make sure that you are using the correct parts for the cooler hold down. The backplate is notched to miss 2 screws on the CPU backplate. Are you using all 4 studs that screw into backplate the same length? I don't use any of the supplied washers in the install.

I've never used coretemp or speedfan. I DO know that AIDA64 gives correct temps for all 4 cores on that MB. My temps are approx. 30C idle, and 65C under full stress.

Make sure that all your fans are pushing air in the same direction, not at each other.
 
There are only a few things that cause such massive temp spike in a water loop.

Either an awful waterblock mount with the sticker still on... Or a pump that is not working properly. OP, verify that your pump is plugged in a nd spinning, I suspect it is not if Speedfan is correct.
 
Make sure that you are using the correct parts for the cooler hold down. The backplate is notched to miss 2 screws on the CPU backplate. Are you using all 4 studs that screw into backplate the same length? I don't use any of the supplied washers in the install.

^ this by far its the most majority of people mistake... The backplate have a shape/marks that must be the same as the 2 CPU screws on the back... Check this: http://s1.postimg.org/5f30yulcf/img_7385.jpg

Also be sure that you are using the correct standoff for 1155s and not the 2011s.. If you make any mistake with the backplate or standoff the h100i will perform badly..
 
I was reluctant, but I tore out the motherboard, and sure enough, as you guys suggested, I did not have it oriented correctly! Now that I reinstalled it, I can tell that it is much firmer. In theory the flexibility of the thin steel tabs might have accommodated this little issue, but it really did not.

Now, I was double checking the manual to be sure i had the right offsets (I had), and I noticed something... a usb cable.
I did NOT have that installed! Well, I did, but not plugged in. Must have forgotten it in my rush to install the new mobo after the annoying RMA. wow!
I had the power plugged in, and the fans, all three of them... it never is explained what the usb cable does. But it was unplugged. Does that mean the pump wasn't running at all?!?

If so, I'm lucky I didn't hurt anything! But, the computer never shut off, and it's supposed to if it gets much too hot... so hopefully I'm ok...?

So I reinstalled everything, being more careful with the thermal paste as well, and It's holding steady at 80-85C on all cores according to coretemp (drifts up and down between the two numbers), and 27C according to HWMonitor and speedfan (why the continued appalling discrepancy?)

Whichever is right, no doubt it's cooler! I feel like such an idiot. A cable unplugged! And orienting the back brace incorrectly!
It's also running at 3.7Ghz and 1.19V... I understand that I might be able to turn up the clock speed and turn down the voltage a little bit?

I tried installing the trial version of aida64extreme, and forgive the dumb question, but where in the menu was CPU temperature? I could not after 10 minutes of browsing everything (including first of all, cpu, cpuid, and motherboard), find that number.
 
glad you got it sorted, the discrepancy between core and case looks to me like a awful job on the TIM on the die.
 
is that something that 'unliding' would help, and if so --since I don't have $300 to waste on a mistake-- is there a forum member who offers unliding services?
 
There's almost no difference at idle. But after several minutes of prime95, there's a huge difference.
40-55C difference (before, and after fixing my h100 install gaff), with coretemp reading fully 40-55C higher on the highest core than HWMonitor or speedfan, which both give just a single temp and are reading the same thing..
Is it cooling well enough to safely try OCing a little bit? Like to 4.2ghz maybe? ...not reaching for a record or anything.
 
now, the difference between the cores-- that has to do with manufacturing variance?
So if core 4 is consistently at least 15C cooler than core2, which is consistently the hottest, is there a way to instruct the computer to somehow favor core 4 in every task? Assign a priority list if you will. If a given program can only use 2 cores, or, if it uses 1 core for a main task and the other three to a much lesser extent-- to tell it to use the coolest core for the primary task always?

I don't know what I'm talking about.
But it seems plausible that such a setting *might* exist.
 
I was reluctant, but I tore out the motherboard, and sure enough, as you guys suggested, I did not have it oriented correctly! Now that I reinstalled it, I can tell that it is much firmer. In theory the flexibility of the thin steel tabs might have accommodated this little issue, but it really did not.

Now, I was double checking the manual to be sure i had the right offsets (I had), and I noticed something... a usb cable.
I did NOT have that installed! Well, I did, but not plugged in. Must have forgotten it in my rush to install the new mobo after the annoying RMA. wow!
I had the power plugged in, and the fans, all three of them... it never is explained what the usb cable does. But it was unplugged. Does that mean the pump wasn't running at all?!?


If so, I'm lucky I didn't hurt anything! But, the computer never shut off, and it's supposed to if it gets much too hot... so hopefully I'm ok...?

So I reinstalled everything, being more careful with the thermal paste as well, and It's holding steady at 80-85C on all cores according to coretemp (drifts up and down between the two numbers), and 27C according to HWMonitor and speedfan (why the continued appalling discrepancy?)

Whichever is right, no doubt it's cooler! I feel like such an idiot. A cable unplugged! And orienting the back brace incorrectly!
It's also running at 3.7Ghz and 1.19V... I understand that I might be able to turn up the clock speed and turn down the voltage a little bit?

I tried installing the trial version of aida64extreme, and forgive the dumb question, but where in the menu was CPU temperature? I could not after 10 minutes of browsing everything (including first of all, cpu, cpuid, and motherboard), find that number.

According to what I've read on Guru3D about that cooler, it is possible you had no pump spinning. The most I've seen said is that is a controller cable for Corsair's "Link" software, which I guess controls the LED as well as the pump. This is q question that can only be answered by a Corsair rep for sure.

Presuming that that cable being unplugged made your pump not spin....which I'm not certain of...that is some poor decision making by Corsair. Even a vanilla 35X waterpump is smart enough that if it doesn't have a controller signal to run although on highest RPM.

In AIDA64 look under Computer->Sensor for thermal readouts.
 
I was reluctant, but I tore out the motherboard, and sure enough, as you guys suggested, I did not have it oriented correctly! Now that I reinstalled it, I can tell that it is much firmer. In theory the flexibility of the thin steel tabs might have accommodated this little issue, but it really did not.

Now, I was double checking the manual to be sure i had the right offsets (I had), and I noticed something... a usb cable.
I did NOT have that installed! Well, I did, but not plugged in. Must have forgotten it in my rush to install the new mobo after the annoying RMA. wow!
I had the power plugged in, and the fans, all three of them... it never is explained what the usb cable does. But it was unplugged. Does that mean the pump wasn't running at all?!?

If so, I'm lucky I didn't hurt anything! But, the computer never shut off, and it's supposed to if it gets much too hot... so hopefully I'm ok...?

So I reinstalled everything, being more careful with the thermal paste as well, and It's holding steady at 80-85C on all cores according to coretemp (drifts up and down between the two numbers), and 27C according to HWMonitor and speedfan (why the continued appalling discrepancy?)

Whichever is right, no doubt it's cooler! I feel like such an idiot. A cable unplugged! And orienting the back brace incorrectly!
It's also running at 3.7Ghz and 1.19V... I understand that I might be able to turn up the clock speed and turn down the voltage a little bit?

I tried installing the trial version of aida64extreme, and forgive the dumb question, but where in the menu was CPU temperature? I could not after 10 minutes of browsing everything (including first of all, cpu, cpuid, and motherboard), find that number.

Glad to see you've found your mistake installing. Really important to pay attention.

The USB cable IS for the Corsair link software monitoring. It is not mandatory to use the USB cable and software for the cooler to work, although I DO recommend using it to monitor your temps and such.

Your pump power is connected through the SATA power cable coming off the pump directly into the PSU, so as long as the computer is running, it's running. The pump running has nothing to do with the MB in relation to power.

To start AIDA64 Extreme monitoring, start AIDA64/Computer/Sensor. That'll get you to all the temp/voltage/speed readings.
 
is that something that 'unliding' would help, and if so --since I don't have $300 to waste on a mistake-- is there a forum member who offers unliding services?


Delidding is a pretty amazing thing... I successfully delidded my 3570k about 2 weeks ago. I of course used coollabs ULTRA for the replacement TIM. I also re-mounted the IHS(cpu lid) as i am not going to be pushing for max possible OC, just high and stable. My over all temp drop was ~20 under max load... my idle temps are in the 20's. REMEMBER THOUGH i do not have hyper threading on my cpu, which iv'e read generates a bit more heat. My tips if you decide to delid...

Find a guide. read it 5 times. watch the video 10 times.
TAKE YOUR TIME, GROUND YOUR SELF!
Use the thinnest strongest sharpest razer blade you can find
put a piece of tape on lid to mark dye.
start with the corners, find an easy one to get in to.
USE COOLLABS LIQUID ULTRA. not pro, ULTRA. and use an extremely thin layer.
 
the USB cable its totally necesary to run the h100i specially if the pwm fans are connected directly to the pump.. why.? cuz corsair link is wich control the speed of the Fans, and the default fan profile for the corsair h100i without plug the cables USB its "QUIET" so the fans are running on VERY low speed like 600-800RPM wich can be the cause of the extremely low performace of the cooler, so i recoomend to plug the USB cable on the mobo header, download the Corsair Link Software and monitor the software.. there you can have control over the temps on the chip, on the h100i fans and other things like LED colours, fan profile etc... but i repeat if you have fans connected to the pump plug the USB to the mobo header that its necesary...
 
There's almost no difference at idle. But after several minutes of prime95, there's a huge difference.
40-55C difference (before, and after fixing my h100 install gaff), with coretemp reading fully 40-55C higher on the highest core than HWMonitor or speedfan, which both give just a single temp and are reading the same thing..
Is it cooling well enough to safely try OCing a little bit? Like to 4.2ghz maybe? ...not reaching for a record or anything.

can you post screen shots of hwmonitor, coretemp and other programs? this its just to check the readings on your system i think the problem on the hwmonitor or speedfans, its you are reading the Global Chip temp from the mobo sensor... core temps do not read from that sensor on the mobo, it read in real time the temp from the individual Cores from the Digital Thermal Sensor on the chip.. i'll post some pictures that im making right now to show you from stock settings on my 2600 stock clocked on a 32C Room Temp thats I a huge room temp that can show the worst case scenario.. and you can compare it with your temps and different programs.. i'll post images later..
 
"In AIDA64 look under Computer->Sensor for thermal readouts."
Still couldn't get aida64 to help me: looked there, and all it wants to tell me is the hard drive temperature, and for that, it says-- 'trial version' [that is to say, it refuses to tell me]

May be that the trial version is severely neutered now; and may be it was better before.


hmm... so pump may have been running (if it weren't, I'd think it would be overheating much worse at all loads), but the fan speeds were not ramping up to match when I loaded it with prime95? Makes sense.
The fans had always been running of course. I did observe that.

Forgot about the corsair software. Opened it up, and it agrees with coretemp, and thinks things are running at around 80C under full load after a few minutes.

Others have reported 60C under load with the h100i. Is there more I should be doing? Do I still have a poor thermal paste job?
I even have the extra 2 fans in a push pull set up past the radiator (blowing the same direction of course, but I did just double check that last night also, to be absolutely sure).
 
err, also, the previously observed 95C temps were scary.
And the computer still had not shut itself down.

Should I tell it to shut down at a certain temperature, if so, what temperature, and if so, where's the best place to set that instruction? Would that be in the bios, or a windows 7 setting?

"TjMax" is 105C-- so it might be that the computer only shuts down to save itself, at 105?
 
Last edited:
found one forum post where a fellow bought nylon washers to lower the backing plate.
(the washers that came with the h100i were not meant for that).

This would mean that you could tighten the screws further and exert more clamping force.
Worth trying? Would more force between the water block and the cpu ensure better contact?
 
hmm... now this is more interesting. Someone else with similar issues-- he started with too much paste, but continued to have high temps:

"He's not even checked his Vcore, there is no way that at full load he's running it on 0.9v, if he's got his voltage set to auto it'll pump through a stupid amount. You're jumping to the worst case scenario without even checking the basics."


According to the coretemp gadget in the corner, my voltage --which I have NOT tampered with-- is defaulting to around 1.235V !
And according to replies in that other thread, 1.235v is a lot of voltage. Might be where the temperature is coming from, now that my h100i install error is fixed?
Haven't changed the clock at all.

So in the bios, should I manually set that much lower? What's your advice to be stable?
If at 1.235V it's reaching 80C at max load, not overclocked; what's a good starting point to try to overclock? 1.1V and try for 4.2ghz? (just making up a pair of numbers). That is to say, if I do attempt an overclock, it will have to be stable at a lower voltage if it's not going to overheat, given that it's already hot at 1.235v and stock clock.
 
finally, question on the thermal paste.

I've been putting on a few tiny, tiny dabs around the cpu, then spreading with a credit card or razor blade.

Someone else said just to put a rice-sized piece in the middle--- that the clamping pressure will spread it all by itself? (or did I misunderstand).

Seems to me that the clamping pressure is really not that much, and the paste is really not that thin, and I'm concerned that if I just left a blob in the middle that it would not spread out.
 
err, also, the previously observed 95C temps were scary.
And the computer still had not shut itself down.

Should I tell it to shut down at a certain temperature, if so, what temperature, and if so, where's the best place to set that instruction? Would that be in the bios, or a windows 7 setting?

hmm... now this is more interesting. Someone else with similar issues-- he started with too much paste, but continued to have high temps:

"He's not even checked his Vcore, there is no way that at full load he's running it on 0.9v, if he's got his voltage set to auto it'll pump through a stupid amount. You're jumping to the worst case scenario without even checking the basics."


According to the coretemp gadget in the corner, my voltage --which I have NOT tampered with-- is defaulting to around 1.235V !
And according to replies in that other thread, 1.235v is a lot of voltage. Might be where the temperature is coming from, now that my h100i install error is fixed?
Haven't changed the clock at all.

So in the bios, should I manually set that much lower? What's your advice to be stable?
If at 1.235V it's reaching 80C at max load, not overclocked; what's a good starting point to try to overclock? 1.1V and try for 4.2ghz? (just making up a pair of numbers). That is to say, if I do attempt an overclock, it will have to be stable at a lower voltage if it's not going to overheat, given that it's already hot at 1.235v and stock clock.

Make sure that you are not reading the VID. To measure a correct voltage you can check using CPU-Z or Aida64.. about the issue with aida64 free version its a crap. all its resctricted...


err, also, the previously observed 95C temps were scary.
And the computer still had not shut itself down.

Should I tell it to shut down at a certain temperature, if so, what temperature, and if so, where's the best place to set that instruction? Would that be in the bios, or a windows 7 setting?

totally right... the best you can do its set a overheat protection maybe on CoreTemp. to shutdown/sleep when reach certain temp.. I do not like nothing more than 80C buts thats just me... some people put overheat protection on 90C.

finally, question on the thermal paste.

I've been putting on a few tiny, tiny dabs around the cpu, then spreading with a credit card or razor blade.

Someone else said just to put a rice-sized piece in the middle--- that the clamping pressure will spread it all by itself? (or did I misunderstand).

Seems to me that the clamping pressure is really not that much, and the paste is really not that thin, and I'm concerned that if I just left a blob in the middle that it would not spread out.

if the contact in the H100i its good then no matter how you spread it.. it will make good contact and spread nice.. just be sure its a small amount of TIM.

Also on the corsair Link2 Software be sure that you have the High Performance Profile at the moment of make any temp test... that will be shown on the next post on images...
 
here are the pictures, I have to remember again that this is a 32C Controlled Room Temp for the Time of the Test.. this present worst case scenario temp.. at least for the average Tested methodology, when you check to the most review, the ambiente are controlled to 25C for a average of comfortable room temp.. all depend too on your room temp and case temps.. my case have amazing Airflow.. also I use only Corsair h100i Stock Fans and only pushing air soutside of the case through the radiator..

according to Aida64:
aida64.png


according to Hwmonitor PRO:
hwmonitorpro.png


according to RealTemp And CoreTemp:
realtemp.png


According to CorsairLink2:
corsairlink2.png
 
Last edited:
any new info to update us about your temp.. or how are performing your h100i?
 
Back
Top