High load temps on new W/C Loop

[Kangaxx]

I recently partook in a new watercooling loop after many years away from watercooling. I picked up a v8 bundle from OcUK here, with the HL black ice rad, and some Mayhems pastel fluid.

Now, its been many years since I did my last loop (I was cooling a socket A AMD chip, so yes, its been a while!) so I am not totally foreign to the principles and how-to on settings up a loop, but this problem is really bizarre.

The loop is installed, with flow going...

Pump/Res -> CPU -> Rad -> pump/res

nothing too exotic there. I did the usual leak testing, but this is where things have changed a bit. In the dim and distant past, I assembled my loop outside of the chassis, and ran it off the PSU for x hours for leak testing. The advantage of this being that you could really shift the Rad and blocks about to get all the air out. Problem is, with this new kit, that ain't possible, so I followed the instructions and installed, primed and bled it in the chassis. Did I get all the air out? I think I did, but its hard to say.

Anyway, onto the problem. It's all running now, great. Idle temps are slightly lower than they were with my K2 Mount DOOOOM (about 30 degrees now), but the load temperatures are off the scale. I have seen 3 of my cores hit 105 degrees, which I believe is the max temp for this CPU (which incidentally is an Intel i5-3570k running at 4.4 ghz on 1.32v). I should add that for reasons I have never understood, core 1 always sits about 20 degrees cooler than the others. Did on the air cooler and does on the water block. I will also add that on air, the temp would only get to 80-90 degrees on heavy load (prime).

So, what have I done to try and sole this puzzle? Well, here you go...

1). Re-seated the CPU block.
2). Checked (best I can) for air in the Rad. Don't think there is any in there.
3). Lowered the OC on the CPU to 4 ghz, but the temps are still shooting up to the max as soon as the CPU is stressed.
4). Checked the loop for any kinks etc. and there are none.

Now, to my knowledge, it would seem that the pressure/flow through the block isn't high enough, as the coolant isn't shifting the temperature away from the block fast enough. However, this is a simple loop with a reasonable pump/res combo (XSPC 750), so I can't see how it would be that. I have got the flow going the correct way through the block before you ask.

Too much thermal paste maybe? I did clean off the old stuff (AS5 i believe it was) and used the XSPC paste instead. Could too much of that cause this issue?

So, I put it to you, can anyone think of any reason why the loop would be doing this?

Thanks in advance!

 

[Kangaxx]

Did you peel the clear film off the bottom of the CPU block? I had your symptoms once and had forgotten to do that.

Checked, i definitely removed it.

Failing that the only thing I can think of is terrible contact between the CPU cooler and block. Assuming you haven't absolutely caked it in thermal paste, is it definitely screwed down tight? What did the paste look like when you removed the block - were there clear signs of contact?

Yes, all 4 top screws are tightened up as far as they will go, and the lower nuts are pretty much as tight as they will go as well. I must admit, I am a bit paranoid to tighten it much more. Don't really wanna crush the CPU. The thermal paste had spread out, and was covering the entire heat spreader. I wouldn't say it was caked, i did put more on the normal (more accident than anything), but we are talking maybe 3 grains of rice worth?

Lastly it would be useful to know what CPU you are running and how many volts you are putting through it... just to have a more accurate idea of the temps we should be expecting.

I did put it in my post, guess you missed it, so here is it again..

Intel i5-3570k running at 4.4 ghz on 1.32v

if your cores are hitting tjmax, then there's definitely something major wrong. my 1st guess would be pump (assuming you fitted the block correctly)

The pump is definitely on. Faulty? How can I say or tell? The damn thing is near silent, and went hell for leather when i was priming the system, so AFAIK, its working fine.

usually this is air in the rad. have you really tilted you case over 70 degrees or more? my front rad needs a big tilt to get all the air out.

Yep, tilted it a good 60 degrees in all directions (apart from turning the whole chassis upside down, but as its a corsair 750D, it is rather heavy )

I think I will re-seat the block again, check the rear mounting plate and re-apply the thermal paste. And maybe try inverting the case completely, just to ensure the air is all out

Thanks for your help and suggestions so far guys. if anyone has any other bright ideas, please share

 

[Kangaxx]

Once again, thank you for all your replies.

I took the PC apart again, checked the mounting of the block. It was pretty tight, but the bottom screws could have gone tighter. In any case, I removed the block, removed the paste from the block and gave the CPU a good clean. I remounted it using Arctic Cooling MX-2, this time with only 1 grain of rice worth, checked the backing plate (which was fine) and remounted the block. I tightened the top bolts evenly and as tight as they will go, and then did the same for the lower knurls. They are now as tight as I can make them, without using a wrench (shredded my fingers in the process!)

Ran the pump on its own (yes, it does still turn on) I can hear it humming, and if that isn't enough, I totally inverted the chassis, to ensure there was no air still in the rad, and I could hear the pump working. Left it for a lil' while to ensure of no new leaks, and have now got it all hooked back up.

Has it made any difference? Not really. I still have the strange situation where core 0 is 20 degrees cooler than the other 3, but under intense load (using Intel Burn test), the temperatures are still skyrocketing to 100+ degrees.

Now Tj max for this CPU is 105 degrees, but when it hits that, the PC doesn't shut down or panic, which is what I thought would happen. Also, and this may just be me not knowing any better, but the temps vary VERY quickly. In the past, CPU core temps would increase over a period of time, if the CPU was under load. As soon as I start the stress test, and the load hits 100% at full tilt (4.4ghz) the temp immediately goes up. Is that correct? Is it possible there is a sensor awry somewhere?

In answer to the other questions, the radiator doesn't appear to be "hot", its warm, certainly, which is what I would expect it to be. The Fans are set to auto in the BIOS, and have been volt'd down with the resistors that came with the Corsair fans. They are definitely both working, and pushing a fair bit of air through the Rad. I used nearly the entire litre of Mayhems fluid. I think there is a tiny bit left in the bottle (which I could probably squeeze in if I wanted).

I will take some pictures if you really think it will help, and post the outputs I get from CPU temp as well in a bit.

Thats all I can think of at the moment. I appreciate all the suggestions so far

 

[Kangaxx]

Here is the picture of the temperatures from the last test I performed. You can see the jump from low temps to high temps as well with that picture.

 

[Kangaxx]

could you post some pics of your loop? something not rigt there

I will post some tonight, although I have more information which may render this unnecessary.

just a quick thought. on the cpu block, did you connect the inlet and outlet properly? if water is going the wrong way, that could explain the temps i guess

I have checked this, the water is def. flowing the right way

I know this may be a pain, but have you tried this CPU on the stock cooler, or another fan/heatsink cooler to see if you have some weird temp readings or a faulty CPU?

Yes, this would be a pain, but I know where you are coming from. I did have this on air before moving to the loop, and the core temp strangeness occurred with that too. All I can assume is that core 0 has a good contact with the heat spreader and the other 3 don't. However, I don't want to risk de-lidding it (not with my ham fisted approach to things anyway).

Anyway, I have been fiddling, and have done the following..

1). Flashed the BIOS to the latest, Beta version from gigabyte (the motherboard is a G1.Sniper M3, in case you are interested)
2). Altered the overclock slightly. Previously, the core was set to 1.3v. I have now lowered this to 1.25v

Since I have done the above, the peak temperatures have dropped. When running the stress test this morning, the hottest core (core 2) peaked at 96 degrees. Hot, yes, but not too hot. I have also noticed that under a "normal" 100% load (I use BOINC, so this is what I have been using to apply a "normal" 100% load), the temperatures now don't go above 80 degrees.

The idle temps do seem lower, even though nothing physical was changed after I did the above, so can only assume the new BIOS has improved the temperature sensing on the mobo, and the reduction in voltage will obviously help with this as well (unsure why it was set to 1.3v. System seems happy and stable at 1.25v with the multi at 44x)

Here is a pic of a test just performed with the stress test running..

I think I will monitor it for now. I rarely put the CPU under massive stress, so I am happy that the temps are not going to be a problem now.

Many many thanks for all the suggestions etc.

 

[Kangaxx]

Something still doesnt seem right over 90c under water at just 1.25v seems wrong. My 3770k also under water also at 1.25v 4.5ghz never goes above 80 normally in high 70s with IBT and I've got 2 670s in the loop too

Fair enough, but I think your loop has considerably more cooling power, if your sig is up to date, even with the additional graphics cards

I will post some pictures of the rig later on when I have a chance.

 

[Kangaxx]

Right, so following on from various advise, I have totally removed the O/C, so the CPU is now back at 3.4 ghz, stock volts (1.15v), and the max temp I have seen under stress is 73 degrees on one core. I left the stress test running for 8 loops, and that's the max temp seen.

I tried to feel the temperature of the Rad and the Piping, and neither seems particularly warm or hot to touch. Then, again, I am roasting, so may not be able to feel the temperature very well.

As promised, here are some hasty pictures of the loop...

Thanks again.

 

[Kangaxx]

Okay if that setup is hitting 73c at stock then something is very wrong so that rules out any issue with the OC or any settings. From what you describe about the rad/piping temps I would say this looks to be some kind of flow rate issue, it sounds to me like the coolant isn't moving fast enough to get the heat off the CPU as fast as it needs to and so the CPU is reaching higher temps than it should.

This was my train of thought as well. Unfortunately, I am not 100% sure of the best way to proceed. As I have alluded to before, the kit was bought from OcUK, but I cannot see how I can prove the Pump is under performing without some kind of flow monitor.

I haven't had as much chance to fiddle this week, as I have been working in the evenings. I may check the mounting of the block, again, to see if it can go tighter tomorrow, if I have time.

 

[Kangaxx]

Is the chip OEM or retail? Assuming it is a retail chip, then it must still be under warranty - given the temp differential - which you insist was also there with your air cooler, then there is a 99% chance (IMO) of there being a fault internally on the CPU. The temp of core #0 is what I would expect all cores to be at +/- 5 or 6 degrees at most.

It was part of another bundle from OcUK. It was pre-clocked by them, so you would have hoped, if the CPU is indeed faulty, that they would have noticed this. As such, I am not sure if its retail or OEM. I will dig up the invoice and see if it alludes to it on there.

Is the pump a d5? Try adjusting the speed on the back, put it to no 5 and re test.

Not a d5, no. XSPC 750.

 

[Kangaxx]

I dont believe its the cpu or the board as this setup was running fine on air

True, but I think the temperature readings are a bit strange in that 1 core is always about 20 degrees cooler than the others, and that did occur with the old cooler. It doesn't explain why things are warmer on the water than on air though.

Can you see any flow in your res?

Yes, I can see the liquid swirling in the Res window. Like I said, I am positive the pump is working and circulating fluid. Surely the idle temps would be sky high if the pump was not moving any liquid round the loop?

 

[Kangaxx]

The Raystorm has a terrible installation method from my recollection

I totally agree with you. The back plate mounting is pretty poor, as the adhesive area is minuscule on the 1155 backplate, so the backplate always falls off when attempting to screw in the mounting screws. In the end, i had to place a small box underneath the backplate, to stop it falling off while i was attempting to mount the block. Doesn't leave you feeling like you have a good mount in any case!

anytime I mount a cooler I try tighten on the diaganol and tighen top left bottom right at same time as each other and then top righ bottom left. Or vice versa, bit I always tighten 2 at same time on the diaganol.

I will look at re-seating the block again tonight (3rd times a charm) and follow the above. I was tempted to totally remove the mobo from the chassis, and attempt to mount the block externally, and then plumb it in once the mobo is re-seated in the chassis.

If things get desperate, I will look at trying this. Mainly, just to ensure I am getting a good mount on the CPU!

Take the pump and dip the inlet in a bucket of water, say 10 litres, and time how long it takes to empty the bucket. You can then work out your flow rate.

Good idea. After thinking about it, that should be possible. It will mean taking apart the loop, but at this stage, I think its going to be a necessary step to rule things out.

I do have an oldish pump which i got from Innovatek (W/C oldies may recognize that name), but the barbs on it are the wrong size. I will see if I can potentially get this hooked up on the loop, to see if it makes any difference as well. All depends if the threads on it are 1/4".

I will try and get at least some, if not all of the above done tonight, but work again will likely intrude, so some things may have to wait until I have more time.

 

[Kangaxx]

Ok, so I found time this evening to strip the PC down, drain the system, test the pumps, remove the motherboard, remount the block, re-install everything, prime the watercooling, and get it all back up and running.

Pump

This turns out to be operating fine. I performed the test, as advised by pastymuncher. Calculated the flow rate to what OcUK say it should be and it is bang on. As such, we must assume that the pump is working as it should. I did, for a laugh, manage to hook up my old eheim 1046, which actually has a better flow rate, but is more cumbersome, noisy and intrusive that the XSPC 750, so it's not going to be used.

CPU/Block

I bled the system, and removed the CPU block. The thermal paste had spread evenly over the block and the IHS. Removed the motherboard from the chassis, cleaned everything, and remounted the block. Was so much easier with the motherboard being flat on my desk. See attached pictures, showing the mountings once done...

and the other side...

Then tightened the lower bolts, til they would not turn any more...

Spoiler

Big picture, hence the spoiler window...

I followed the previous instructions, so tightened 2 corners at a time, until they were all pretty tight, and then finished them off individually.

Re-installed the motherboard, etc. etc. Re-plumbed the loop, re-filled it, bled it once again (tilted the chassis 90 degrees each way, to ensure all the air was out of the system). Leak tested and then re-booted PC.

So, have the temperatures improved?

No, not really, they are still where they were before. I suppose they are marginally better, but nothing to get excited about. Screen of latest burn test below..

Yes, we still have the weird Core 0 reading. That has always been there, so I am guessing I have a screwy CPU.

I am thinking I may stop there. The temps never get this high under "normal" use, so I am not concerned that my CPU is going to overheat etc. Maybe I am just unlucky in having a hot CPU, who knows. The PC is quieter and much nicer to live with than with the old air cooling system, so at least there is that.

If anyone can think of anything further, by all means share.

Thnx for reading.

 

[Kangaxx]

I tightened it gradually, 2 opposite corners at a time. Once they were all pretty tight, I moved onto fully tightening them individually. It was not possible to go tighter!

I have tried the sensor test again. I ran it again this evening after the rebuild, and the max temp seen was 77 degrees on core 2 (seems to be the hottest) and core 0 maxed out at 56 degrees. As such, core 2 has the biggest movement, and core 0 has the smallest.

I know a strange issue has been found with core heat difference, but out of curiosity, what RPM are the rad fans running at, when on load?

The Radiator fans are all fixed speed, under-volted. Currently, the Rad fans are both running at 1278rpm (approx.), with the remaining chassis fans running at 1060rpm ish.

On a separate note, i have found the invoice for the mobo/CPU/RAM bundle, and it was sold as a single unit, so I have no idea if the CPU was retail or OEM. I would be amazed if they wasted money on a retail CPU, when it would be shipped OEM, so to speak, so that would mean its out of warranty, but I think I will raise a note with them anyway, see what they say.

 

[Kangaxx]

OcUK have agreed that the temps are not what they should be, and after reviewing the evidence (this thread) have agreed to RMA the chip and send out a new one.

It does mean I will be without a PC for a few days, but I guess it's worth it in the long run.

 

[Kangaxx]

This to me screams software faulty, more specifically BIOS.. Or... A bad thermal diode in the CPU.

Looking at the temps core 0 seems to be at the temperatures it should be at.

Try updating or re-flashing the BIOS and see if that clears it.

Already tried that. I flashed the motherboard to the latest BIOS, and the idle temps stayed the same, matching between BIOS and windows, so i suspect the software is correct. As this issue has always been preset, in spite of numerous formats, software re-installations, I highly doubt its a software problem. I agree, core 0 is the "proper" temperature that most people seem to get with these clocks on these types of cooling. The fact the other 3 are sky high would appear to be some kind of IHS problem, or sensor issue on the chip itself.

Ignoring the weird core 0, it's still running hotter than it did when it was air cooled so I can't see how the cpu is at fault. When you take the cpu out to return it, check the base of that block to see if it's warped/bowed. It might be a idea to do the bucket test again as well. This time through the whole loop though just to rule out a blockage somewhere.

I have reviewed the tests I did on the old k2 cooler, and it would appear i made a mistake there. The idle temps, and the load temps are cooler on water. The problem is I never ran the intel burn test whilst the K2 was on the chip, so I have no direct comparisons. What I can tell you, with BOINC running at 100% utilization on all 4 cores pushed the temps to 90 degrees on air. On water, they barely push 75. When I RMA the chip, and the replacement comes back, and the issue is the same, I will eat humble pie, but I really can't see how the loop is the issue.

CertifiedDan - I could try this, however, the block is mounted as tightly as it will go. I cannot turn the lower knurls any more if I wanted to, so I don't think it's a poor mounting. The fact I have re-mounted the block 3 times now, each time getting the same idle and load temps (bar a few degrees), I really don't think mounting is the issue. My idle temps are good (between 28 and 32 degrees, room temp. dependant), its just the load temps that appear to be the issue.

While it will be inconvenient, I will RMA the CPU and see how that goes. If the issues persist, I will probably cry and then look at other things

 

[Kangaxx]

Not being funny but... have you checked to see if they pump is even running?

Take the whole loop out and push water through the lot to see if its flowing enough.

See Post #38. I have tested the pump. It is working fine. I know what happens when the pump doesn't work, as I managed to accidentally unplug the molex cable on the last rebuild when re-attaching the side panel. The idle temps start at approx 55 degrees, and gradually increase to the Tjmax. When the pump is working, as it is now, the idle temps are fine.

Please stop bashing the pump! It is working. I can feel it vibrating, and can see the fluid moving in the window of the Res.

 

[Kangaxx]

As the CPU is currently back with OcUK, waiting for another one to be shipped out, I have taken the loop apart again, and checked everything you have mentioned.

There are no blockages in the block (I did check on the last mounting, just in case), and as far as I can tell, there are no blockages in the Rad either. The flow rate is the same, even with the block and Rad present in the loop.

I did flush the radiator before installing it (used distilled water. In the distant past it was vinegar that I used, but that appears to not be necessary anymore ) I didn't see much, if any, flux/crap come out, but I did rinse it through a good 3 times.

On a side note, I may have another way of verifying if the loop is ok or not, as I have another mobo/CPU coming for a NAS box, so I can whack it on this loop and see if the cooling is up to scratch. Not conclusive by any means, granted, but it may show if the loop is doing something it shouldn't be.

I am sorry for my last snappy reply. It had been a crap day, not an excuse I know, but there you go. I know you guys are trying to help, and I do appreciate all the suggestions.

When I have anything further, I will update the thread.

 

[Kangaxx]

Very odd. At least your loop sounds like its clear of restrictions. Hopefully they pick a high clocker out of the box of cpu's for you

As it was a pre-clocked bundle, they should do. Only thing that is concerning me is that they don't appear to sell the 3570k as an OEM CPU anymore. The RMA is in "awaiting replacement" status, so who knows how long it'll take to get a replacement. Have to wait and see.

 

[Kangaxx]

While it's all apart take a straight edge and check the base of the block to make sure it's not warped.

I'll try and do this tonight. I've still got an ancient netbook that's still working, so i'll take some pictures and upload them later on.

 

[Kangaxx]

Block doesn't look warped. See below