Dodgy H100i, dodgy OC (TL;DR warnings)

Verponian

Verponian

Verponian

Associate
Joined
3 Mar 2012
Posts
246
Location
South Yorkshire
Over the past couple of days I've seen huge jumps in my CPU temps and more than I'd expect from just hotter weather. I'm wondering if the pump on my H100i is on the fritz somehow (3 months old) or if my OC has just gone unstable (also 3 months old).

Background:

I picked up a new GTX 670 for a SLI setup, first ever SLI and was running a benchmark on the new card on its own to see if it ran faster than the one I had (same model, no OC on the GPU). I realised the H100i was running like a hairdrier at the end. I have a 3770k CPU OCed to 4.5 Ghz on a 1.2 vcore using MSI Z77 Mpower. It's not a fixed OC but a dynamic or boost or whatever you want to call it.

Symptoms:

I used to get great idle temps of between 26 to 28 degrees, maybe hitting low 30s with a light load. Gaming for a few hours would maybe see temps rise to Mid forties, 50 at the very highest. Prime95 test over 10 hours when I first OCed saw a peak temp of 63 on the highest core, average of around 55-57.

Now, idle temp is high 30s to mid 40s. A task like installing new nvidia drivers saw temps hitting 60+ degrees, temps I used to get on prime95 burn in. Getting WHEA errors at these temps.

I've tried reinstalling the corsair link software, deleting all profiles, uninstalling, reinstalling etc. Tried all connections, new USB header for the cooler, new fan headerson the board and on the pump etc. Uninstalled the radiator and checked for dust etc. I've reseated the cooler with no improvement, maybe marginally worse, checked the backplate is tight. Also tried swapping in the old GPU just in case that was somehow the cause.

A few weeks back, I got a similar problem during gaming when I was installing mods. What seemed to fix it then was either reinstalling/repairing the link software and/or removing the radiator and blowing dust out even though there wasn't really much. Not sure which, it just improved after I did both but neither have worked this time.

Reason I'm wondering if the pump is on the fritz is because it almost seems as if coolant isn't reaching the radiator properly, even though one hose appears warm, the other cool and corsair software saying the fan is working. I can't feel any vibration from the pump. I've found that fan speed basically doesn't seem to affect CPU temps at all. At 400rpm or 2200rpm, CPU temp will stay roughly the same. H100i temp readings always used to be pretty static even when the CPU was under load. They'd stay around 25-28 degrees, maybe nudging over 30 after a period of time at load. Now the H100i temp shoots up straight away into the 30s, even 40s straight out of boot, which is what causes the fans to go crazy. Despite these high temps, the actual radiator seems cool to the touch.

I was hoping to avoid it but my next step was to possibly disable my OC and see if things normalise then. Another option was to salvage my Hyper 212 from my other PC and see if that can cope with the temps any better, logic being if a Hyper 212 can cope better than a h100i, the h100i is screwy. If not, maybe it is the OC.

I welcome any suggestions.
 
Agree with switching to air. If your pump is working correctly, both hoses should be roughly the same temperature. The problem with these corsair AIO loops seem to be building up on the forum, thank god they have reliable RMA.
 
The performance of the H100i was amazing...when it worked properly but it is starting to be a huge headache. A nice straight forward air cooler is looking more and more attractive, less stuff to go wrong.

By the way, I thought one warm hose, one cool was correct. Doesn't one carry hot coolant to the radiator and the other take cool stuff back to the CPU?
 
comparing like for like what are your prime temps now? remember compared to 3 months ago weve had the best part of 15c increase in temps, ive got a full custom loop and ive seen an increase in around 5c in the last week alone
 
Last edited:
Verponian said:
By the way, I thought one warm hose, one cool was correct. Doesn't one carry hot coolant to the radiator and the other take cool stuff back to the CPU?
Click to expand...

If you were to sketch up a diagram on paint, thats roughly how you would think it works. Rather, the pump should be quick enough to make temperatures uniform. Any working pump is quick enough in a single block/radiator loop. For the coolant to be noticeably cooler exiting the radiator means that the liquid has been in the radiator long enough to cool down, this in turn means that you have very little flow (it is much more efficient if the temperature is roughly uniform).

Think of it like this, if the liquid in the loop is slow enough to make a temperature difference between the radiator inlet and outlet, than there is also going to be a large temperature difference between the inlet and outlet of the CPU block. If your liquid is staying in the block long enough to take up that heat difference, it will have a huge negative effect on temperatures. The reason why watercooling is advantageous is that a good loop dissipates heat quicker to prevent it building up around the CPU and reaching a higher equilibrium temperature. The way a loop does this most efficiently is to make it so the liquid travels fast enough round the loop so that it is roughly the same temperature, it would mean the liquid inside the block at any one point would be cooler compare to if it was running with much lower flow.
 
Last edited:
stuie said:
comparing like for like what are your prime temps now? remember compared to 3 months ago weve had the best part of 15c increase in temps, ive got a full custom loop and ive seen an increase in around 5c in the last week alone
Click to expand...

I used idle at 26c, now between 38c and 44c. Under mild load it used to be about 30c-33c, now around 50c-55c. Anything heavier and now I'm pushing into 60s when before it would have been 40s. I daren't try an actual heavy load because I'm fairly sure it'd hit 70-80.

The thing is, had the rise been gradual over the week, I'd be more inclined to put it down to higher ambient temps but it pretty much went from being 26c one evening to 38c the next. The fans no longer seem to effect temps regardless of speed and the cooler itself has jumped from a dependable constant 27c temp to a fluctuating 35-40c one.

Also thanks Avenged7Fold, that makes more sense. So effectively only small amounts of heat are conducted to the coolant but that's offset by the fact that the coolant is being pumped so quickly that it is able to get more heat away from the CPU overall and the coolant coming into contact with the CPU will be relatively consistent temps instead of extremes (which can't be ideal).

I'm going to uninstall the h100i tonight and try out the hyper 212. It's incredibly frustrating not being able to play with my new SLI setup. :(
 
go for a noctua d14 or a k2 doom if they can fit

best thing with aircoolers is they are reliable,use some top quality paste aswell for a few more degrees c drop
 
wazza300 said:
go for a noctua d14 or a k2 doom if they can fit

best thing with aircoolers is they are reliable,use some top quality paste aswell for a few more degrees c drop
Click to expand...

Still have half a syringe of MX-4 which has always been pretty good for me though I welcome any other suggestions. I don't really want to use AC5. I think if I go for either of the air coolers, I'll go for the Noctua. I have a NZXT Phantom which is pretty awkward with regards to cooler clearance, especially if you want to have a side intake. I remember seeing a mod someone managed with a 200mm Betfenix fan and a Noctua D14.


Also, installed the Hyper 212 and I think it's safe to say it's definitely a screwy H100i. Though it's pretty loud with one of the corsair fans attached, it's giving me idles of 29c (8 lower than the corsair), load temps of 40c (10 lower) and a peak of 62c during the combined test in 3dmark11 (H100i hit 62c installing a driver!) and that's all with just one fan.

Also P15915 in 3dmark11! :D
 
Can recommend the silent akasa Apache fans for your cooler

Coollaboratory liquid metal will give you a further 4-6c drop over any paste on the market,not to be used on aluminium tho
 
Corsair link was reporting the pump running at around 2200rpm. Can't remember the exact number but I don't think I've ever seen it fluctuate so don't know how accurate it was. Seen some threads on the corsair forums where users said Link reported a perfectly functional pump even when it was demonstrably not pumping coolant at all. Apparently some piece can come loose inside so it no longer shifts coolant but still registers revolutions.
 
Verponian said:
Corsair link was reporting the pump running at around 2200rpm. Can't remember the exact number but I don't think I've ever seen it fluctuate so don't know how accurate it was. Seen some threads on the corsair forums where users said Link reported a perfectly functional pump even when it was demonstrably not pumping coolant at all. Apparently some piece can come loose inside so it no longer shifts coolant but still registers revolutions.
Click to expand...

It should be running at 2350 rpm, however when i swapped my p8p67 for a p8p67 deluxe i noticed my pump was running at 2450 rpm.
 
Don't want to bump am old thread but thought I'd update just in case anyone with similar symptoms with a H100i is wondering what the problem is.

RMA'd the H100i and just got feedback saying the problem is a faulty pump. Apparently the RPM of the pump was too low but more importantly something detached meaning it wasn't even pumping coolant much at all.

So above problem = faulty pump.
 
The pump seems to be running at expected range, which is between 2200 to 2400rpm. A gradual rise in CPU temp would normally suggest a faulty pump or a poor contact between the coldplate and CPU. I might have missed it, but did you try re seating the pump on top of the CPU and apply a new thermal paste? I'm kinda suspecting there was a poor contact between the two causing your CPU temp to gradually rise overtime. Unless of course Corsair Link pump rpm value was inaccurate where it says 2200rpm but in reality it was really running lower than spec.
 
The ambient temperature has gone up ! Mine was like your around 26c & with the warmer weather it gone up to 35c !
 
Greybeard said:
The pump seems to be running at expected range, which is between 2200 to 2400rpm. A gradual rise in CPU temp would normally suggest a faulty pump or a poor contact between the coldplate and CPU. I might have missed it, but did you try re seating the pump on top of the CPU and apply a new thermal paste? I'm kinda suspecting there was a poor contact between the two causing your CPU temp to gradually rise overtime. Unless of course Corsair Link pump rpm value was inaccurate where it says 2200rpm but in reality it was really running lower than spec.
Click to expand...

Yeah, I tried re-seating it a couple of times and saw no improvement. I know some mobos have had an issue with the H100i and getting good contact between the pump and the CPU but mine seemed pretty good. The thing is the temperature rise wasn't gradual. It was okay one evening, the next evening it had jumped 15 degrees. It looks like the corsair link software was indeed giving inaccurate readings of the pump, engineers have confirmed the pump unit was broken.
 
You must log in or register to reply here.
Back
Top Bottom