Jump to content
Corsair Community

H80i V2 coldplate curiosity


rg500g
 Share

Recommended Posts

I'm messing around with overclocking a 3600X and running stress tests with P95. It's more fun than some games. It looks like an all core clock of 4.175 GHz is feasible on 1.32 volts per Ryzen Master. I also note that my sustained CPU temp is 77C, a 55C jump from ambient. Given the thermal mass of the coolant, maxxing the fans isn't useful until I hit 35C coolant temp. I then see CPU temp creep down about 4 degrees over the next few minutes. The fan influence could be purely coincidental, but it repeats.

 

Given what Ryzen Master is saying, the CPU total socket power is about 105 watts. I'm now getting curious as to just how much heat the coldplate can pull out of the CPU. I see CPU temp rise to 77-78C in a minute or two and coolant is 32.xC and slowly rising as expected. I tweaked the CPU for some more speed and it crested 80C which was more than I was willing to accept so I backed off of some things. My case is not an issue as I pulled all the panels off of the Core V1 during the test, essentially making it an open air chassis. I'm satisfied with what I've got, but am still curious regarding what should be expected of this AIO. Some of the reviews on the Internet show CPU temp deltas 10C below mine, which has me wondering.

Link to comment
Share on other sites

The CPU temp to coolant temp differential is going to be massively influenced by voltage and to a small degree specific characteristics of your CPU. Some 3600X chips will run cooler or warmer at 1.32v on the same cooler, hardware, test bench, etc. However, this makes comparing or predicting the expected differential fairly difficult except when it is clearly out of range. Most CPUs (Intel or AMD) will be pushing a 50C+ CPU temp to coolant differential near the limit. It is also an illustration that the voltage + CPU is the limiting factor. If you dial up 1.40v for Vcore, you won't have to wait minutes to see if it gets too hot. It will hit the limit or beyond at the first second of load.

 

The conductivity of the cold plate is a fixed value and presumably you won't be taking apart your CPU to make physical improvements. That leaves the voltage (obviously) and the TIM as the only things you can work with to reduce the differential. I am not a huge proponent of TIM miracles and any of the more repeatable brands should produce similar results. If you are looking to take back another 1-1.5C, there are professional comparison tests out there. Just be aware since the Ryzen is so efficient and relatively low TDP (105W), the results are not as dramatic as if you were using a piggy 10900K at 250W.

 

The secondary factor in this is coolant temperature and that is the cooler's responsibility. The theoretical low is room temp, but you will never get there. That usually requires large external radiators and with an AIO, the temp sensor is in the CPU block area so it would have to transmitting zero heat to get past the sensor. Most users will see an idle coolant temp about 4-7C above their room temp. This is both fan speed, case design, and location specific. The rear exhaust is going to be warmer than the lowest front intake spot, just on local temp alone.

 

 

 

 

If you really want to get down to basic numbers, you can try the following:

 

1) After a non-load period of time, cold boot, etc., run the fans at a fixed rate as fast you can tolerate for 3-5 minutes. This will remove fan speed as a variable, or at least as much as you can. This is your minimum possible coolant temp with minimal load.

 

2) Run a fixed load stress test with no warm up period. I like CPU-Z's Bench Tab stress test for this because it moderate and flat. Other linpack tests can work as well. You don't want the dynamic load type created by AIDA or OCCT that is meant to attack stability. We just want watts.

 

3) Note the coolant temperature and press start. You will instantly get a CPU temp lift to a high level. That is your true CPU temp to coolant differential (as least for this load type). That differential should hold throughout the duration of the fixed load test. It will hold true at any fan or pump settings and should be repeatable every time. The only difference is room/coolant temp will change, but the differential should not. This now allows you to isolate fan/pump settings and it gives you a maximum coolant temp threshold in relation to what you will allow for the CPU. If your differential is +34C and your limit is 79C, then you know 45C is your max coolant temp regardless of room or case conditions.

 

4) During the load test, you will see coolant temp creep up at relatively slow rate -- maybe +1C every 30 seconds or so. This will be additive to CPU temp and they should change together. That is the effect of the fans and pump. +1C to coolant equals +-1C to CPU. The maximum coolant temp and delta will vary by load. Other sources of heat (like the GPU) will add to the total and thus the max CPU temp and that is one of the trickier bits to predict or manage. However, if you know you maximum coolant temp, then you have a target to avoid.

Link to comment
Share on other sites

OK, in my setup the CPU vs coolant differential is 43C after 2 minutes of sustained constant load. As the load continues CPU temp gets ahead of coolant for obvious reasons - the coolant's thermal mass. Creep becomes miniscule when fans go 100% and coolant holds around 34-34.5C. I'm not curious enough to test heat soak and leave it for a half hour.

 

So, given what I'm seeing, using a specific CPU, open air case, etc., the H80iV2 can keep my 3600X under 80C for at least 15 minutes' sustained stress of Prime95, small FFT, 12 threads. Per Ryzen Master the AIO is dealing with the full TDP, 95 watts.

 

Now I'm wondering if this is the best I can expect regardless of cooling equipment employed. This hobby is starting to feel like stereo - the relentless pursuit of the next 'thing' that will make the setup 'better'. Leads to tubs of parts in the basement...

Link to comment
Share on other sites

A cold plate is a cold plate, despite any marketing. Not much ground to gain there. Of course, your rate of heat dissipation will change with a larger radiator surface area or to a lesser extent more fan speed. The coolant/CPU differential will be still be the same, but you can reduce the coolant temperature and that is the CPU temp savings. These are small gains as well. It's XX watts per radiator and fan speed speed calculation. From a quick look, it seems like a 240mm radiator is about the largest the V1 can take. Will it be massively better? No. You might take back a couple of degrees, but you also might be able to do so with slightly less fan speed. That is major gain when going up radiator sizes -- slightly better cooling with less noise.
Link to comment
Share on other sites

The V1 is stuck with a 140mm max radiator. The V21 has much more room for a 240. I'm really pushing what I can get out of a cube this small. If I went with air cooling I might get more heat out of the CPU in the short term, but that heat will have to be pushed out the case fast or else heat soak sets in and I'm done.

 

The two 80mm fans in the back can only do so much. Admittedly, sustaining the worst load P95 can exert on all threads for 15 minutes straight is a ludicrous exercise with no real world application, but I'm bored and curious. I think I've tapped out the H80iV2's capabilities. A collateral product has been the best voltage for a 4.1 GHz configuration, a .075 negative offset that is stable at idle but doesn't push 1.4 volts or over at peak. I would not have derived that had I not gotten curious as to what the AIO could do. The Asus 'Five Way Optimization' overclocker can push the CPU to over 4.2 GHZ but the voltage is excessive and it had the CPU get above 80C.

 

I'm now pondering just what an air cooler with max height of 140mm can do. Hopefully I can leave well enough alone.

Link to comment
Share on other sites

 Share

×
×
  • Create New...