EK Custom Loop high temps

[00newboldc]

 

Apologies if this has been covered before but I’ve just completed my first full custom loop and was wondering if my temps are a little higher than they should be for a custom loop. My setup is as follows:
Ryzen 7 1700 (small oc 3.7ghz, stock voltage)
Gigabyte x370 Gaming K7
16GB DDR4 3200mhz Team Group dark pro
EVGA 1080 SC edition
Seasonic Focus 750w
Phanteks Evolv ATX tempered glass case

My cooling setup is:
EK DDC Pump
EK GPU waterblock
EK cpu & motherboard block
12mm hardline tubing, with ek fittings
1no. 240mm black ice nemesis radiator in the top of the case
1no. 240mm black ice nemesis radiator in the front of the case
3no. 120mm EK Vardar fans as intake in the front of the case
2no. 120mm EK Vardar fans exhaust top of the case
1no. 140mm Corsair mag lev fan as exhaust in rear of case

Currently my static temps for cpu are hovering around 35-40 and are going up to 60 - 65 under load. I know this isn’t an issue but would have thought a custom loop would have offerd better temps. Also after the computer has been under load the temps struggle to return back to static temps. I was wondering if this was normal and if anyone can provide and advice? Spent around £500-£600 on the custom loop and I want to make sure I’m getting the full potential out of it! Many thanks in advance.

[Semper]

What are your ambient temps?

If your pump is variable speed, knock the speed up a bit, it's possible that you don't have enough flow to keep things cool.

[00newboldc]

Thanks for responding, the pump is an EK ddc pump with pwm so I have it on a curve at the moment but it’s runs at full speed when under load. I’m not sure on the ambient temps, but they won’t be that high as it’s winter in the UK and I never have the heating on in my office to help with the the cooling!

[Mick Naughty]

So only the cpu is running that way?

Have you ran the fans at max speed?

I assume the system has been bled properly and the block makes contact as it should.  Correct inlet used on the mono block as well?

[Wild Penquin]

What are your GPU temps (at load) ?

 

Just testing my loop; 1 x 360mm radiator, GPU block (RX Vega 64) and i4790k. All are EK Phoenix MLC series (the pump is the cheap SPC kind integrated into the rad).

 

If folding ~180W+ (it varies a bit depeding on work unit) and burning CPU (burnK7) at 8 x 100% processes, and pump and fans at 100% -> GPU temp plateau at ~45°C and CPU plateau ~55-65°C (there's some delta between cores, and I've delidded the CPU - there's a possibility I've not applied the TIM optimally between the die and the IHS, since core temperatures varies a bit; but it was like this even before delidding, so it could be just this individual CPU).

 

I never get these kind of tempeartures at real-world workloads, this is an extreme example.

 

Our systems are not directly comparable; I'm not sure how hot do compare the i7 to the Ryzens thermal output, but I believe the Vega puts out more W of heat than the 1080 SC, and you have 2 x 240mm radiators, and I have a 360mm rad. I'd still carefully conclude, the temps you have are are fine, if they were measured under synthetic, 100% pegging the CPU benchmarks. If they are real-world temperatures, something could be wrong. In other words: what do you mean "under load"? Was the GPU also under load (I presume having 180W of extra heat in addition to the CPU in a loop is different to not having it).

 

My first bet would be a bad contact at the CPU and / or GPU heatblock. In case you have good temperatures at other while the other is bad, points towards this issue (that's why I asked your GPU temps).

 

Hope this helps!

Edited February 13, 2019 by Wild Penquin
Heavily re-wording trying to make my thoughs more clear... but the gist is the same

[Wild Penquin]

Hmm,

 

I believe a picture tells more than some sentences, and gives a rough idea about a certain system.

 

Without the GPU folding on my system, CPU temps decrease almost 10°C by over 5°C (while folding).

 

See the screenshot and draw your own conclusions. Of course this is just one example, and any conclusions can not be generalized to other systems.

 

(Rad exhaust is from an extra temperature sensor at the radiator exhaust).

EDIT: realized the screenshot does not have time legend; well, 1 pixel = 1 second. Vertical lines are 30pixels = 30 seconds apart.

Edited February 13, 2019 by Wild Penquin
timescale!

[00newboldc]

7 hours ago, Mick Naughty said:

So only the cpu is running that way?

Have you ran the fans at max speed?

I assume the system has been bled properly and the block makes contact as it should.  Correct inlet used on the mono block as well?

Thanks for the reply, yes I’ve tried running fans at max speed, doesn’t make a massive difference. Removing the side panels and lid of the case (phanteks evolv atx TG) does make a small difference. As far as I’m aware I’ve bled the radiators correctly, first time I’ve done a custom loop so there’s a chance I may not have done it properly!

[00newboldc]

5 hours ago, Wild Penquin said:

Hmm,

 

I believe a picture tells more than some sentences, and gives a rough idea about a certain system.

 

Without the GPU folding on my system, CPU temps decrease almost 10°C by over 5°C (while folding).

 

See the screenshot and draw your own conclusions. Of course this is just one example, and any conclusions can not be generalized to other systems.

 

(Rad exhaust is from an extra temperature sensor at the radiator exhaust).

EDIT: realized the screenshot does not have time legend; well, 1 pixel = 1 second. Vertical lines are 30pixels = 30 seconds apart.

Many thanks for the detailed feedback! To be honest I have not taken a lot of notice of the gpu temps as they always seem to be lower than than the cpu temps. In terms of testing for cpu ive just been using cinebench for the cpu and uningine benchmark for gpu and I’ve been monitoring them temps with hwmonitor. What would you recommend for properly testing cpu and gpu and monitoring temps?

[Wild Penquin]

I believe Cinebench is too short to reach a plateau in temperatures.

 

4 hours ago, 00newboldc said:

What would you recommend for properly testing cpu and gpu and monitoring temps?

It doesn't matter what you use, as long as it is synthtetic test (i.e. worst-case scenario, pegs all cores at 100%; I think your real core count of threads is enough, actually using HT number of threads might be less taxing on the CPU). You also need to run it so long that the temperature reaches a plateau (as in: look at my graph. The temperature is "going down" to reach the plateau, since my fans are software controlled and were not at 100% at the start of stress. But my example is a bit special case; usually this is reversed as fans are usually set to 100% before starting a test - or at least any sane user would do that, but maybe I'm not sane )

 

I believe a lot of Windows users use something that calculates Pi indefinitely (any numbers of threads - until the user stops it of course). I don't use Windows so I can not give a definitive recommendation, you will need to Google for it.

[For Science!]

15 hours ago, 00newboldc said:

 

Are you P-state overclocking? This can drastically reduce your idle temperatures. And Load temperatures will depend on your Vcore voltage and how aggressive or not your LLC is.

 

Personally I recommend HWinfo64 to monitor your temperatures, RealBench for a whole system stress test, P95 blend for CPU only stress test, and Furmark for GPU only stress test.