Overclock.net banner
3081 - 3091 of 3091 Posts

·
OC...D
Joined
·
28,239 Posts
Cheers, works fine at 4.7GHz 1.180v static, really strange, why even include adaptive as a setting if it only works when overvolting...
my next question is, What are your cooling solutions like? as when i ran tests with said settings i was having strange temps like 65° to 95° core temps, now my setup shouldn't be having trouble as its got 2x 480mm XT45's and an EK motherboard block (R6EE) and I've repasted twice to see the same cores 30 degrees hotter than the coldest cores.
Safe to say i've no idea what i'm even doing anymore :)
If you mean a monoblock, getting a really good fit on the CPU is always a challenge - this may be the cause. Also, use a good TIM and the "pea" method for application. Cooling an 18 core and getting a low delta-T between cores caused a lot of guys here to essentially Bin waterblocks. 😲
 

·
Registered
Joined
·
3 Posts
If you mean a monoblock, getting a really good fit on the CPU is always a challenge - this may be the cause. Also, use a good TIM and the "pea" method for application. Cooling an 18 core and getting a low delta-T between cores caused a lot of guys here to essentially Bin waterblocks. 😲
Yeah thats the one, covers the VRM too. a right pain to take off just for remounting though.
I used the paste that came with it of which i presume is TG Hydronaut applied originally by the "X" method then the big pea, just not sure how my temps are so horrid, what paste would be best for this use case?
 

·
Old and Decrepit
Joined
·
2,314 Posts
Cheers, works fine at 4.7GHz 1.180v static, really strange, why even include adaptive as a setting if it only works when overvolting...
my next question is, What are your cooling solutions like? as when i ran tests with said settings i was having strange temps like 65° to 95° core temps, now my setup shouldn't be having trouble as its got 2x 480mm XT45's and an EK motherboard block (R6EE) and I've repasted twice to see the same cores 30 degrees hotter than the coldest cores.
Safe to say i've no idea what i'm even doing anymore :)
The mono block is your problem. Ive tried several of them and they all suck. They cool the VRMS great but suck at cooling the CPU. I tried the exact same one you are referring to. Tried a million ways from sunday and it just doesnt cut it. More for looks than anything. They are OK for stock clocks but a huge fail for OC. Get an Optimus Sig V2 or I'm running an EK magnitude with a flat cold plate that isnt exactly flat. Heat Killer IV pro isnt too shabby either but falls behind the afore mentioned by 2-3C. The magnitude with the flat cold plate just happens to fit the curvature of my particular IHS better. One thing I learned through the process of testing many blocks is the the IHS on all of them are slightly different, even the same SKU. Replace that POS with a quality block and put the original VRM heatsink back on and all will be golden. I got 10C+ higher temps with it on the CPU but it did a damn fine job on the VRMS as does the stock cooler. The infineon VRMS on the R6EE are supposedly rated for 100C+ but running mine with the stock cooler the fan never cuts on. Heavy AVX workload for hours and they never pass about 54C. With my fans and rads liquid temp doesnt pass 30C and hottest core on AVX workload with no offset at 4.7 is 75C and 80-82C at 4.8. Only time I passed 90C was with the monojunk and running 5.1GHz.
EK rep even said at stock they projected 6C higher on the monoblock than a velocity that isnt a good choice for an HEDT but the monoblock uses the same cooling engine.

I get about a 10C spread across the cores except one lone core, core 0 that likes to run quite a bit cooler than the rest. I tune and bench off of the hottest, not the average so when you see me post a temp its the hottest core.

My loop is overkill but Im also cooling a pair of 2080Tis with it.
 

·
Registered
Joined
·
1,180 Posts
Cheers, works fine at 4.7GHz 1.180v static, really strange, why even include adaptive as a setting if it only works when overvolting...
my next question is, What are your cooling solutions like? as when i ran tests with said settings i was having strange temps like 65° to 95° core temps, now my setup shouldn't be having trouble as its got 2x 480mm XT45's and an EK motherboard block (R6EE) and I've repasted twice to see the same cores 30 degrees hotter than the coldest cores.
Safe to say i've no idea what i'm even doing anymore :)
The reason adaptive is still there on x299 boards is because it works on any syklake-x cpu BUT the current 10 series. Why get rid of that functionality when it works perfectly on 7 and 9 series?

I gave up on attempting adaptive on 10 series long ago, keeping it locked manual has added about 50-60 watts idle vs stock on my 10980xe. This would be my 4.7 @1.850 v 1.850VIN LLC5. Basically I ended up copying JPM and Justin. Actually I didn't, I did find 1.80 worked for a while, but maybe it was adding the 128gb that forced me to go 1.85 on the cores. either way, llc5 and 1.850VIN is thanks to the talk here.
 

·
Registered
Joined
·
1,180 Posts
The mono block is your problem. Ive tried several of them and they all suck. They cool the VRMS great but suck at cooling the CPU. I tried the exact same one you are referring to. Tried a million ways from sunday and it just doesnt cut it. More for looks than anything. They are OK for stock clocks but a huge fail for OC. Get an Optimus Sig V2 or I'm running an EK magnitude with a flat cold plate that isnt exactly flat. Heat Killer IV pro isnt too shabby either but falls behind the afore mentioned by 2-3C. The magnitude with the flat cold plate just happens to fit the curvature of my particular IHS better. One thing I learned through the process of testing many blocks is the the IHS on all of them are slightly different, even the same SKU. Replace that POS with a quality block and put the original VRM heatsink back on and all will be golden. I got 10C+ higher temps with it on the CPU but it did a damn fine job on the VRMS as does the stock cooler. The infineon VRMS on the R6EE are supposedly rated for 100C+ but running mine with the stock cooler the fan never cuts on. Heavy AVX workload for hours and they never pass about 54C. With my fans and rads liquid temp doesnt pass 30C and hottest core on AVX workload with no offset at 4.7 is 75C and 80-82C at 4.8. Only time I passed 90C was with the monojunk and running 5.1GHz.
EK rep even said at stock they projected 6C higher on the monoblock than a velocity that isnt a good choice for an HEDT but the monoblock uses the same cooling engine.

I get about a 10C spread across the cores except one lone core, core 0 that likes to run quite a bit cooler than the rest. I tune and bench off of the hottest, not the average so when you see me post a temp its the hottest core.

My loop is overkill but Im also cooling a pair of 2080Tis with it.
If you want max performance, stop using monoblocks. Full stop. I did it once on my X99, also my first major watercooled build. EK for the x99 strix. Not only was it difficult to maintain due to all the pieces and orings, but the screw heads were very malleable, even with basic hand turning, i would strip the heads, really hate those EK screws... To this day, they still use them. Since then, I've learned to add a tiny bit of silicone grease to the threads so they don't lock into the surface area.
 

·
Registered
Joined
·
60 Posts
If you want to maxed out AVX-512 too then mixed between static and adaptive is the way to go.

1.15v is already too much for AVX-512 it goes very hot under stress test while 1.15v will only net you 4.6 - 4.7 at best.

So for cores running 4.6 set static while for those running 4.7 and above set adaptive to optimize everything. This is I believe what Asus board is doing with their AI overclocking algorithm and I couldn’t agree more.
 

·
OC...D
Joined
·
28,239 Posts
Yeah thats the one, covers the VRM too. a right pain to take off just for remounting though.
I used the paste that came with it of which i presume is TG Hydronaut applied originally by the "X" method then the big pea, just not sure how my temps are so horrid, what paste would be best for this use case?
Hydronaut is ..meh. Kryonaut is what they should have supplied. That said, I find Swiftech's "TIM Mate TIM-2" to be better or at least as good in thermal conductivity, never dries out and spreads easily. I've had it on CPUS here for years and it always remains fluid. (well except at sub zero temps).
If you are gonna stick with the monoblock, either do any necessary massaging to the standoffs to get a better contact patch on the CPU (I had to do this with the monoblock no longer on my RE5-10BE, I even used pressure sensitive contact film), or back down to an OC where no core sustains temps above 80C during normal use. Where "normal use" is not some stress test power virus. Pushing 400+W through the chip and saying the temps are too high does not reflect reality unless that is how you use the machine all the time. Too many folks limit their OC based on temps seen during benchmarks, stress tests or other artificial loads. If during normal use you see sustained temps above 80C that is outside the durability comfort zone for this SKU. Short, peak temps well above 100C at voltages at or above 1.3V (and LLC>1.9V) ... frankly that's Flash Bulb territory :)

Edit, The Optimus Sig V2 is an excellent block, expensive tho. Their Foundation block works nearly as well (I'm talking a degree or less in my hands) and costs much, much less. Both are designed with the "jet plate" type inlet nozzle. A design flaw IMO. Same with the EK blocks.
 

·
Registered
Joined
·
60 Posts
Actually if you are using Unreal Engine 5, during compiling shaders the peak temp go upwards near to prime95 (still few degrees below but close enough for real world app IMO). UE5 is the next gen game engine and I believe it is realistic enough.
 

·
OC...D
Joined
·
28,239 Posts
Compiling, encoding, compute, etc, these are all daily uses that can pull >2xTDP (easily). And that is another "signal". Sustained loads over 2xTDP will cause wear/degradation.
 

·
Registered
Joined
·
60 Posts
Well yeah it surprise me to see that but it does means prime95 load is actually not that far from real world scenario (still depends on usage of course).
 

·
OC...D
Joined
·
28,239 Posts
No not really. 🖖
 
3081 - 3091 of 3091 Posts
Top