Custom loop temp validation

[CrackinX]

So I just built a new system, 9900k paired with an rtx 2080, oced to 5ghz on 1.32 vcore. I'm running a custom loop on the cpu and gpu, an ek 120 rad between the gpu/cpu and an ek 280 pe before it goes back to the res. I know this cpu runs hot as hell, but doing the blender classroom test I hold steady right around 85c just looking for some input as I came off a 6700k which didn't run as hot as this. Are these load Temps OK or are they high for a custom loop? 

[Skiiwee29]

Sounds normal to me. 

[CrackinX]

Ok cool, i was a little nervous when I ran prime 95 it hovers right around thermal throttle, but I do understand that those are unrealistic loads. Thank you. 

[Benjeh]

29 minutes ago, CrackinX said:

So I just built a new system, 9900k paired with an rtx 2080, oced to 5ghz on 1.32 vcore. I'm running a custom loop on the cpu and gpu, an ek 120 rad between the gpu/cpu and an ek 280 pe before it goes back to the res. I know this cpu runs hot as hell, but doing the blender classroom test I hold steady right around 85c just looking for some input as I came off a 6700k which didn't run as hot as this. Are these load Temps OK or are they high for a custom loop? 

Comparing temps from a 6700k to a 9900k is apples to oranges, also loop order doesn't matter, at all. 85c with that amount of rad space sounds ok, personally i hate heat so both my cpu and gpu share a 420mm and a 360mm rad combo. at 1.55v my cpu hits 86c during stress tests. at 1.6v my cpu hits 92c during stress tests but my cpu is only capable of 4.7ghz so isn't worth going about that.

[CrackinX]

Yeah, I would have liked to have been able to do a second 240 or 280 rad but I very much ran out of space. I had to cram it in there pretty good as it is. 

[Benjeh]

Not bad man, not bad. 

[CrackinX]

I'm also going to try and dial back the voltage a bit, I actually landed at 1.32 trying to fix realbench stress test constantly finding instability, until I did some reading and found that it's very common and bugged, I've had no issues with any other testing. So I'm thinking I might be stable at a lower voltage. 

[Benjeh]

11 hours ago, CrackinX said:

I'm also going to try and dial back the voltage a bit, I actually landed at 1.32 trying to fix realbench stress test constantly finding instability, until I did some reading and found that it's very common and bugged, I've had no issues with any other testing. So I'm thinking I might be stable at a lower voltage. 

Depends what you've been using to test with? If my PC passes an IBT AVX maximum test then 9/10 its fully stable.

 

[CrackinX]

I will look into that, I've done semi short runs with prime 95 max heat, it does stay stable but puts me into the mid 90s so I don't run that very long, I do a custom fixed fft run and it handles better, aida64, blender classroom render, I've done a few other random tests as well. 

[CrackinX]

occt seems underwhelming on the stress side lol

 

[CrackinX]

I'm also unsure why it's dropping to 4.5 ghz when I have an avx offset of zero.

 

Edit: I think it is due to the ring down ratio, let's turn it off to see. 

[CrackinX]

On 5/1/2019 at 9:24 AM, Benjeh said:

Depends what you've been using to test with? If my PC passes an IBT AVX maximum test then 9/10 its fully stable.

 

Thanks so much for the info on this, I'm pretty sure I have a stable oc at this point, managed to finish an hour run of occt medium with no crash. Prime 95 on a balanced load with avx crashes out individual cores but I think it's from Temps, it slays a few cores well into thermal throttle. I'll be doing more stress tests tomorrow and trying to tweak things into a lower temp range. 

[Benjeh]

Honestly, if IBT AVX MAX passes like 20 runs, you're gold, the result should be a 3.xxxxxxxxx result anything else shows instability. Prime crashing cores means unstable, personally I'm not a fan of prime, it needs to run for a long time as it runs through passes of different sizes, people think like 3-4 hours is long enough and think they're stable then run into random blue screens because it hasn't ran long enough.

[LmnSour]

Honestly, rule of thumb is 120mm rad is bare minimum per GPU/CPU.  240mm per GPU and CPU is about where water starts to perform.

[CrackinX]

Yeah I have a single 120 rad and a 280 rad in my loop. I do want to try to swap the 120 out for a 240, but might have to go external if I want that to happen. I've re-seated my waterblock once and it didn't help any with Temps but it just seems off on some of the higher load stress tests, like an avx run of prime kills it to 100c in under a minute,and that's at lower vcore than most of what I'm reading people land on to be stable. I'm going to be doing a tare down and swapping out the mobo for something with a vastly better vrm set (mostly because I just like to burn through money) so maybe a good clean off and another re-seat on a new mobo will help. I am honestly not thrilled with the clearance for the cpu block thumb screws on the maximus xi hero, so I'm not overly confident that I have a good even seat. Hoping the Aorus master I just ordered will have better hand room around the vrm.