RealBench temperature way too high

[Apollo190]

1 hour ago, For Science! said:

Paste looks great, the point is to fill gaps, not introduce another layer. So where you can see the letters, it has good metal-to-metal contact, which is what you want.

 

I think you're at the point where the question of need/want come into play, are you willing to spend another couple hundred or so USD (or more) for perhaps 0.1~0.2 GHz performance? If so, it may be worth a try, certainly a bigger AIO or the Icegiant will give you another couple degrees headroom, although whether that will actually let you go up another step in the OC depends on your luck.

 

You can look down here for silicon lottery of the 10900K, it sounds like you're slightly on the losing side of what the chip is capable of on average. Personally, I think I'd recommend you to know that you've done all that you can, and so just sit back, relax, and enjoy the performance that the PC is capable of.

 

https://siliconlottery.com/pages/statistics

I forgot to mention that this vcore voltage is with avx offset set on 0, if that makes any difference. Is the hwinfo vcore reading same as "All Core Die Sense Vcore" at silicon lottery page? It shows 4.9GHz avx using only 1.170v "All Core Die Sense Vcore", which seems really unbelievable to me , it's incredibly low.
To be honest I'm not sure if 100-200MHz more is worth it, especially in gaming, but on the other hand I don't like having such high temps all the time either.

Have you checked out the sub-zero cooler i linked in my last message? Would that help me get much higher overclocks? Or is that just a gimmick?

 

17 minutes ago, Mister Woof said:

Realbemnch stresses all of your components (which makes it more realistic) so it's possible it's dumping more GPU heat from that 3080 which is raising your cpu temp

I thought about that too, it's possible, but many reviewers said it shouldn't affect the cpu temps so I don't know. It's true that blender and cinebench don't overheat as high. But prime95 does (after 15-20 minutes), in fact the system shuts down afterwards.

[Mark Kaine]

3 hours ago, Apollo190 said:

Did I put enough paste there

Imo it's way too much, thermically not ideal. see how everything gathers on the outside..? does that look even to you? (because it doesn't to me)

 

it's not gonna kill anything, and probably better to put too much than not enough, but I used like not even half of this amount and it's sitting "comfortably" at 74C doing prime... 

 

of course I don't do this anymore as it's a useless exercise putting strain on my hardware unnecessarily while also using an absurd amount of energy for no reason whatsoever, but yes, that's way too much paste in my opinion. 

 

 

[For Science!]

46 minutes ago, Apollo190 said:

I forgot to mention that this vcore voltage is with avx offset set on 0, if that makes any difference. Is the hwinfo vcore reading same as "All Core Die Sense Vcore" at silicon lottery page? It shows 4.9GHz avx using only 1.170v "All Core Die Sense Vcore", which seems really unbelievable to me , it's incredibly low.
To be honest I'm not sure if 100-200MHz more is worth it, especially in gaming, but on the other hand I don't like having such high temps all the time either.

Have you checked out the sub-zero cooler i linked in my last message? Would that help me get much higher overclocks? Or is that just a gimmick?

 

I thought about that too, it's possible, but many reviewers said it shouldn't affect the cpu temps so I don't know. It's true that blender and cinebench don't overheat as high. But prime95 does (after 15-20 minutes), in fact the system shuts down afterwards.

I think you can get higher OC with subambient cooling, but in my opinion it makes little sense to have to consume more power to cool something so that it can consume even more power.

 

In my opinion, it is not too much paste and reducing the amount of thermal paste will not improve thermals, and so is not worth doing.

[Mark Kaine]

1 minute ago, For Science! said:

I think you can get higher OC with subambient cooling, but in my opinion it makes little sense to have to consume more power to cool something so that it can consume even more power.

I agree, makes no sense, but it's fun as hell! 

 

 

[Bombastinator]

4 hours ago, Apollo190 said:

Did I put enough paste there? It looks really thin on the cpu now, I can even see the letters on the cpu. The fans were already running at their max speed, 1500rpm and 1200rpm. 

 

I set my voltage to offset mode but couldn't put more than -10mV because system kept crashing on idle, but was fine under load. So I decided to decrease llc to level 1 instead, and put +30mV, so now I'm getting between 1.288v and 1.296v under heavy load. That's the lowest I could go in order to have no issues. I could lower it to 1.279 without system crashing, but then I'd occasionally get errors in hwinfo.

 

What do you suggest me to do from here? Should I get an aio, icegiant prosiphon elite or maybe this https://landing.coolermaster.com/pages/ml360-sub-zero-cryo-clocking/ ? Although I don't know if the last 2 would be compatible with my motherboard and case, they probably should be.

Thin paste is good not bad unless it’s because there isn’t enough paste.  The buildup on the edges suggests not enough paste wasn’t a problem. 

[Apollo190]

I think I'm going to delid the cpu with rockitcool delidding tool. Is it better to do liquid metal + IHS + paste, or direct die + paste? Thinking about it logically, direct die + paste should be equal or better, becuse all the IHS does is transfer cpu die heat to the coldplate. I also feel much more comfortable using paste because it's not conductive. Is there any reason not to do direct die cooling? I can't do liquid metal with direct die because I'll probably switch to kraken z73 which has a copper coldplate, not nickel. However, I saw a video on youtube where some guy couldn't get a proper contact between arctic lf2 and cpu die, and he was using rockitcool's direct die frame. Could it have the same issue with kraken z73?

[Bombastinator]

4 hours ago, Apollo190 said:

I think I'm going to delid the cpu with rockitcool delidding tool. Is it better to do liquid metal + IHS + paste, or direct die + paste? Thinking about it logically, direct die + paste should be equal or better, becuse all the IHS does is transfer cpu die heat to the coldplate. I also feel much more comfortable using paste because it's not conductive. Is there any reason not to do direct die cooling? I can't do liquid metal with direct die because I'll probably switch to kraken z73 which has a copper coldplate, not nickel. However, I saw a video on youtube where some guy couldn't get a proper contact between arctic lf2 and cpu die, and he was using rockitcool's direct die frame. Could it have the same issue with kraken z73?

My memory is there are limitations to liquid metal.  It is conductive which causes some issues and it reacts with some but not all metals which also causes some issues.  Both need to be addressed before liquid metal can be used.  Bright side of liquid metal is that if it does work it works really really well.

[For Science!]

4 hours ago, Apollo190 said:

I think I'm going to delid the cpu with rockitcool delidding tool. Is it better to do liquid metal + IHS + paste, or direct die + paste? Thinking about it logically, direct die + paste should be equal or better, becuse all the IHS does is transfer cpu die heat to the coldplate. I also feel much more comfortable using paste because it's not conductive. Is there any reason not to do direct die cooling? I can't do liquid metal with direct die because I'll probably switch to kraken z73 which has a copper coldplate, not nickel. However, I saw a video on youtube where some guy couldn't get a proper contact between arctic lf2 and cpu die, and he was using rockitcool's direct die frame. Could it have the same issue with kraken z73?

This is not going to be so easy with a soldered cpu, so the gains may be minimal.  Only worth it if you’re going to do direct die + liquid metal in my opinion, also with a nickel plated cooler.

[Apollo190]

54 minutes ago, For Science! said:

This is not going to be so easy with a soldered cpu, so the gains may be minimal.  Only worth it if you’re going to do direct die + liquid metal in my opinion, also with a nickel plated cooler.

der8auer said he got about 7C difference with liquid metal under IHS. Do you think I'd see similar results with direct die + kryonaut?

I saw gamer nexus post this some time ago: "If you're direct die mounting and eliminating the IHS, it's best to use a good thermal paste (not LM) for long-term viability and for peak performance. LM is good, but delta is minimal vs. paste in direct die. What most people don't understand is that the biggest improvement from delidding and re-lidding w/ liquid metal isn't from the liquid metal itself, but rather the removal of a layer of silicone adhesive and thus an air gap die-to-IHS. LM helps a lot, but isn't everything" .

Don't know how accurate that is however.

[Bombastinator]

13 minutes ago, Apollo190 said:

der8auer said he got about 7C difference with liquid metal under IHS. Do you think I'd see similar results with direct die + kryonaut?

I saw gamer nexus post this some time ago: "If you're direct die mounting and eliminating the IHS, it's best to use a good thermal paste (not LM) for long-term viability and for peak performance. LM is good, but delta is minimal vs. paste in direct die. What most people don't understand is that the biggest improvement from delidding and re-lidding w/ liquid metal isn't from the liquid metal itself, but rather the removal of a layer of silicone adhesive and thus an air gap die-to-IHS. LM helps a lot, but isn't everything" .

Don't know how accurate that is however.

To be fair der8auer is one of the best overclockers in the world and a consummate pro’s pro. Comparing what you would get to what det8auer can get may not be realistic. 

[For Science!]

11 hours ago, Apollo190 said:

der8auer said he got about 7C difference with liquid metal under IHS. Do you think I'd see similar results with direct die + kryonaut?

I saw gamer nexus post this some time ago: "If you're direct die mounting and eliminating the IHS, it's best to use a good thermal paste (not LM) for long-term viability and for peak performance. LM is good, but delta is minimal vs. paste in direct die. What most people don't understand is that the biggest improvement from delidding and re-lidding w/ liquid metal isn't from the liquid metal itself, but rather the removal of a layer of silicone adhesive and thus an air gap die-to-IHS. LM helps a lot, but isn't everything" .

Don't know how accurate that is however.

Wasn't this for the 9XXXK series CPU, where it was Intel's first time soldering a CPU (in a while) and it was a semi-mess because it was too thick (both the silicon and the solder). IIRC the 10-XXX series they did a better job I thought. I only half-agree with GN on the matter of whether it is the silicone that makes the bigger difference as I always re-paste (i.e. apply silicone back) after delidding, and still observe great temperature improvements (for pasted CPUs, such as the 7700K, 8700K) and so you can still have great contact as long as you apply sufficient pressure during the re-lidding procedure. 

[Apollo190]

3 hours ago, For Science! said:

Wasn't this for the 9XXXK series CPU, where it was Intel's first time soldering a CPU (in a while) and it was a semi-mess because it was too thick (both the silicon and the solder). IIRC the 10-XXX series they did a better job I thought. I only half-agree with GN on the matter of whether it is the silicone that makes the bigger difference as I always re-paste (i.e. apply silicone back) after delidding, and still observe great temperature improvements (for pasted CPUs, such as the 7700K, 8700K) and so you can still have great contact as long as you apply sufficient pressure during the re-lidding procedure. 

It could've been the same for 9XXXK, but he did it with 10900k as well.

I also found this comparison between conductonaut and kryonaut (on stock IHS it seems). If the difference is that small on direct die as well, I don't know if it's worth risking damaging the cooler or worse frying the components if it leaks.

I could also try to find the new kryonaut extreme, as it seems to have a bit better thermal conductivity (14,2 W/mK, while kryonaut has 12,5 W/mk). But it would probably perform almost the same.

[For Science!]

4 minutes ago, Apollo190 said:

It could've been the same for 9XXXK, but he did it with 10900k as well.

I also found this comparison between conductonaut and kryonaut (on stock IHS it seems). If the difference is that small on direct die as well, I don't know if it's worth risking damaging the cooler or worse frying the components if it leaks.

 

LM shines when there is a area of highly concentrated heat. So I am not surprised (and knew already) that LM on top of an IHS brings negligible improvements as the bottleneck for thermal transfer is down below the IHS. This is also why direct-die GPU cooling benefits very little from even liquid metal, as the die is sufficiently large enough for the heat to transfer with a good paste like kryonaut.

 

Things start getting iffy with Intel CPUs with their small monolithic dies and that's why it makes sense to use LM for this application. Although whether going through the hassle of a delid + direct die cooling for an already soldered CPU can be debated. Similar logic should and can be applied to Ryzen chips, but nobody seems to be bothering with delidding those, so in my opinion, there was little rationale to delid the 9th gen intel CPUs, and even less with the 10th gen.

[Apollo190]

On 12/7/2020 at 2:57 PM, For Science! said:

LM shines when there is a area of highly concentrated heat. So I am not surprised (and knew already) that LM on top of an IHS brings negligible improvements as the bottleneck for thermal transfer is down below the IHS. This is also why direct-die GPU cooling benefits very little from even liquid metal, as the die is sufficiently large enough for the heat to transfer with a good paste like kryonaut.

 

Things start getting iffy with Intel CPUs with their small monolithic dies and that's why it makes sense to use LM for this application. Although whether going through the hassle of a delid + direct die cooling for an already soldered CPU can be debated. Similar logic should and can be applied to Ryzen chips, but nobody seems to be bothering with delidding those, so in my opinion, there was little rationale to delid the 9th gen intel CPUs, and even less with the 10th gen.

The kraken z73 finally arrived yesterday. I was blown away by how much cooler the cpu runs now. I was able to run realbench for 1 hour now, peak temperature was 77C while average was 73-74. Considering I was throttling at 100C with drp4 within 5 minutes, this could be well over 23C difference. I was also able to lower the load voltage during realbench stress test from 1.288v to 1.243 and remain stable (I guess the cooler the cpu is the less voltage it needs). At 1.234v however, I started getting whea errors in hwinfo, so 1.243v should be the limit for 4.9GHz here. Is that a normal result? I mean compared to other 10900k chips out there? I feel like it should be able to go lower at 4.9GHz, somewhere around 1.2v, or is that possible only for the lucky golden sample chips?

[Bombastinator]

35 minutes ago, Apollo190 said:

The kraken z73 finally arrived yesterday. I was blown away by how much cooler the cpu runs now. I was able to run realbench for 1 hour now, peak temperature was 77C while average was 73-74. Considering I was throttling at 100C with drp4 within 5 minutes, this could be well over 23C difference. I was also able to lower the load voltage during realbench stress test from 1.288v to 1.243 and remain stable (I guess the cooler the cpu is the less voltage it needs). At 1.234v however, I started getting whea errors in hwinfo, so 1.243v should be the limit for 4.9GHz here. Is that a normal result? I mean compared to other 10900k chips out there? I feel like it should be able to go lower at 4.9GHz, somewhere around 1.2v, or is that possible only for the lucky golden sample chips?

The astoundingly high delta change implies either a mount problem with the earlier cooler or a case heat problems you solved.  Possibly both.

[Apollo190]

30 minutes ago, Bombastinator said:

The astoundingly high delta change implies either a mount problem with the earlier cooler or a case heat problems you solved.  Possibly both.

I posted pictures of the old cooler in my previous posts here, the contact seemed to be good. I installed the new aio on the top as an exhaust, maybe the case needed more exhaust before and now that's helping as well, I'm not sure, but I'm happy with the results  . Also I tested with the dust filter installed on top, and without it, and didn't notice any difference in temperature, so I'll keep it on, because when pc is off the dust will fall inside if there's no filter. Just want to figure out if this voltage is normal now, hoping I got a decent chip.

[Bombastinator]

22 minutes ago, Apollo190 said:

I posted pictures of the old cooler in my previous posts here, the contact seemed to be good. I installed the new aio on the top as an exhaust, maybe the case needed more exhaust before and now that's helping as well, I'm not sure, but I'm happy with the results  . Also I tested with the dust filter installed on top, and without it, and didn't notice any difference in temperature, so I'll keep it on, because when pc is off the dust will fall inside if there's no filter. Just want to figure out if this voltage is normal now, hoping I got a decent chip.

Mount problem doesn’t always mean paste, I does a lot though.  There may be some sort of problem with that air cooler.  30° delta change is crazy.  Should have been more like 5° maybe.  It’s looking like some other problem got solved.

[WoodenMarker]

On 12/12/2020 at 12:10 PM, Bombastinator said:

Mount problem doesn’t always mean paste, I does a lot though.  There may be some sort of problem with that air cooler.  30° delta change is crazy.  Should have been more like 5° maybe.  It’s looking like some other problem got solved.

20-30c delta isn't abnormal if one of the coolers being compared is no longer able to keep temps in check since temps only stop rising at shutdown. 

Here are some examples of lighter loads that aren't even near thermal shutdown showing a 10c+ drop between the DRP4 and more powerful coolers. A heavier load like the one in this thread can easily result in a higher temp delta.

https://www.kitguru.net/components/silas-newman/be-quiet-dark-rock-slim-review/5/

https://www.kitguru.net/components/cooling/james-dawson/arctic-freezer-7-x-cpu-cooler-review/5/

https://www.computerbase.de/2020-11/pure-loop-test-aio-be-quiet/2/#diagramm-differenz-cpu-zu-raumtemperatur-ueber-schalldruckpegel-oc

[Bombastinator]

2 hours ago, WoodenMarker said:

20-30c delta isn't abnormal if one of the coolers being compared is no longer able to keep temps in check since temps only stop rising at shutdown. 

Here are some examples of lighter loads that aren't even near thermal shutdown showing a 10c+ drop between the DRP4 and more powerful coolers. A heavier load like the one in this thread can easily result in a higher temp delta.

https://www.kitguru.net/components/silas-newman/be-quiet-dark-rock-slim-review/5/

https://www.kitguru.net/components/cooling/james-dawson/arctic-freezer-7-x-cpu-cooler-review/5/

https://www.computerbase.de/2020-11/pure-loop-test-aio-be-quiet/2/#diagramm-differenz-cpu-zu-raumtemperatur-ueber-schalldruckpegel-oc

Sure but those are overwhelmed coolers.  The cooler that was overwhelmed is well over 200w.  About the same size as the cooler that worked.  If one was overwhelmed the other would have been too.  Unless it was anot performing to spec.  There are times when even big air won’t cut it.  They start happening as one approaches 250w or more. 
So the air cooler was most likely damaged or something.  A 240mm rad and big air are about the same.  

Edited December 14, 2020 by Bombastinator

[WoodenMarker]

2 minutes ago, Bombastinator said:

Sure but those are overwhelmed coolers.  The cooler that was overwhelmed is well over 200w.  About the same size as the cooler that worked.  If one was overwhelmed the other would have been too.  Unless it was anot performing to spec.  There are times when even big air won’t cut it.  They start happening as one approaches 250w or more. 
So the air cooler was most likely damaged or something.  A 240mm rad and big air are about the same.  

There are better and worse performing 240mm's. The best performing 240mm AIO's easily outperform the DRP4 and the Kraken Z73 is a 360mm. 

[Bombastinator]

13 minutes ago, WoodenMarker said:

There are better and worse performing 240mm's. The best performing 240mm AIO's easily outperform the DRP4 and the Kraken Z73 is a 360mm. 

Lotta lotta heat then.  I always had the impression that the effective difference between a low performing 240 and a high performing 240 was mostly about reliability and noise.  As far as actual heat the answer went while there were differences they weren’t particularly large.  That a 360mm rad is required though surprises me. 

[WoodenMarker]

47 minutes ago, Bombastinator said:

Lotta lotta heat then.  I always had the impression that the effective difference between a low performing 240 and a high performing 240 was mostly about reliability and noise.  As far as actual heat the answer went while there were differences they weren’t particularly large.  That a 360mm rad is required though surprises me. 

The upper and lower end of 240mm's seems to trade blows with lower end 360mm's and mid range 120mm's respectively. 

[Bombastinator]

17 minutes ago, WoodenMarker said:

The upper and lower end of 240mm's seems to trade blows with lower end 360mm's and mid range 120mm's respectively. 

My understanding is that a big air cooler can crush a 120mm AIO. 

[WoodenMarker]

11 minutes ago, Bombastinator said:

My understanding is that a big air cooler can crush a 120mm AIO. 

High performance 120mm AIO's like the H80i / GT, Liquid Freezer 120, or Seidon 120XL regularly trade blows with the best performing air coolers. 

https://www.bit-tech.net/reviews/tech/cooling/cooler-master-masterliquid-pro-280-review/2/

https://www.cowcotland.com/articles/2141-4/ventirad-noctua-nh-d15s.html

https://www.guru3d.com/articles_pages/be_quiet_silent_loop_360_review,9.html

https://www.geeknetic.es/Review/986/Cooler-Master-Hyper-612-v2.html

https://lab501.ro/racire/aio-round-2015-9-sisteme-de-racire-one-testate/13

https://www.chiphell.com/portal.php?mod=view&aid=13786&page=7

https://www.tweaktown.com/reviews/9399/scythe-mugen-5-argb-plus-cpu-cooler/index.html#Test-System-Setup-Thermal-Tests-and-Noise-Results

https://www.clubedohardware.com.br/artigos/refrigeração/teste-do-cooler-blizzard-t2-da-cooler-master-r36315/?nbcpage=6

[Bombastinator]

2 minutes ago, WoodenMarker said:

High performance 120mm AIO's like the H80i / GT, Liquid Freezer 120, or Seidon 120XL regularly trade blows with the best performing air coolers. 

https://www.bit-tech.net/reviews/tech/cooling/cooler-master-masterliquid-pro-280-review/2/

https://www.cowcotland.com/articles/2141-4/ventirad-noctua-nh-d15s.html

https://www.guru3d.com/articles_pages/be_quiet_silent_loop_360_review,9.html

https://www.geeknetic.es/Review/986/Cooler-Master-Hyper-612-v2.html

https://lab501.ro/racire/aio-round-2015-9-sisteme-de-racire-one-testate/13

https://www.chiphell.com/portal.php?mod=view&aid=13786&page=7

https://www.tweaktown.com/reviews/9399/scythe-mugen-5-argb-plus-cpu-cooler/index.html#Test-System-Setup-Thermal-Tests-and-Noise-Results

https://www.clubedohardware.com.br/artigos/refrigeração/teste-do-cooler-blizzard-t2-da-cooler-master-r36315/?nbcpage=6

Looking at your first link, that’s not a 120 it’s a 240.  Second link is in French, but the graphs are pretty clear.  They’re also saying a D15s is less powerful than a darkrock3. This strikes me as unlikely. There is a 120 that apparently keeps up with the dark rock3.  I’m not sure what the difference is between a dark rock 3 and a darkrock pro4 is. These graphs are quite different from others I’ve seen.  They’re also 4 years old.

third link uses a 4970k @4.6 as a test bench which strikes me as a bit odd.  It’s also too old to have a darkrockpro4 in it, though it does have some noctuas.  There are 360 rads getting beat by 120 rads in that list.  These are very different that the stuff I’ve seen lately.   Apparently manufacture of coolers has changed a lot in only a few years or something.