What's a good therman grease? Are thermal pads any good?

Falkentyne said:
Well ? Been a month + and you never posted back. Did it destroy the CPU? :)
Click to expand...
Okay so I finally tested the paste against thermal grizzly kryonaut paste on an MSI RTX 3060 Ti dual fan card in a mining rig. Since I didn't control the ambient temp in the garage, I used the card right next to it as a reference point.

Usually the delta between the card tested vs. the card next to it is around 12-16c due to the crappy cooler on this MSI card not having a super beefy heatsink. This has always been the case and no matter how many times I remounted or play around with the pads/coolers the delta is always 12-16c in the morning and throughout the evening using the generic 30g tube of paste.

So taking thermal pad thickness out of the equation I have a large jar of thermal putty that I replaced all the thermal pads on the card so the GPU core will always get a perfect mount and I tested this on both pastes.

The delta between the two cards with thermal grizzly kryonaut paste is now 10-12c. So in my quick and probably flawed way of testing, Thermal Grizzly Kryonaut is about 2-4c better than the Protronix Series 9 Extreme Performance paste that you can get for $9 for a 30g tube for this specific RTX 3060 Ti. I would put this paste somewhere along the lines of AS5-level of performance but you get a ton of it for $9.

The Protronix Series 9 thermal compound paste is 62% more thermally conductive than the Series 7. The high density formula is suitable for extreme environments and applications that demand the very best performance. Exceptional long term stability, odorless, low oil content, non-curing, non-volatile, non-corrosive, non-toxic, flame retardant, not electrically conductive •
Operating Temperature: -50/340? •
Thermal Conductivity: > 5.15W /m-K •
Thermal Resistance: < 0.004? -in²/W • 45% metal oxide compounds Large 30G Syringe
Click to expand...

Edit: Added some graphics from HIVEOS stats.

PASTES.jpg
 
Last edited:
legcramp said:
Okay so I finally tested the paste against thermal grizzly kryonaut paste on an MSI RTX 3060 Ti dual fan card in a mining rig. Since I didn't control the ambient temp in the garage, I used the card right next to it as a reference point.

Usually the delta between the card tested vs. the card next to it is around 12-16c due to the crappy cooler on this MSI card not having a super beefy heatsink. This has always been the case and no matter how many times I remounted or play around with the pads/coolers the delta is always 12-16c in the morning and throughout the evening using the generic 30g tube of paste.

So taking thermal pad thickness out of the equation I have a large jar of thermal putty that I replaced all the thermal pads on the card so the GPU core will always get a perfect mount and I tested this on both pastes.

The delta between the two cards with thermal grizzly kryonaut paste is now 10-12c. So in my quick and probably flawed way of testing, Thermal Grizzly Kryonaut is about 2-4c better than the Protronix Series 9 Extreme Performance paste that you can get for $9 for a 30g tube for this specific RTX 3060 Ti. I would put this paste somewhere along the lines of AS5-level of performance but you get a ton of it for $9.
Click to expand...
And 2 - 4c is exactly what we are paying for ;)

I've been using Thermalright TF8 for a couple of years. Love it. Decided to try Gelid GC Extreme again (bought a new tube). That stuff used to be great but.....its been shown in the past couple of years to be wildly inconsistent if not outright seemingly a different/worse product. I bought a new tube of TF8 and I'm back to normal.
 
chameleoneel said:
You are claiming 53C full load on air, 11900k with Adaptive boost switched on?

What is the "full load".

Is that with the unlimited turbo or with the standard turbo time limit of about 1 minute (and then it clocks down a lot).

How about Cinibench R23 multi-core or even encoding a video in handbrake?
Click to expand...
I've run CB on this, and it holds 5.1Ghz on all 8 cores during a run if I remember correctly. I typically only care about what I actually use my machine for though. I mostly watch the stats for my use cases, because I think that's a bit more intelligent to focus on. The way mine is configured, it boosts up to 5.1Ghz on all 8 cores and holds it, never dropping at any point according to my hwinfo records. There's just a lot of thermal headroom with my IC Graphite pad and Noctua NH-D15, 30 or 40C worth in my use cases. If it's not a limiting factor, I move on and don't wring my hands over it endlessly. Performance is fantastic in daily use, unquestionably the fastest machine I've ever used so it's a mission accomplished.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Deadjasper said:
Got a box running at 45c at idle. Me thinks it might be the cheap Chink no name thermal grease I used. The last name brand grease I bought was Thermal Grizzly and it worked great but the next time I ordered it I must have gotten a bad batch because it was impossible to spread out, all it did was roll up into a ball.
Click to expand...
Amazon sellers were know to have shipped counterfeit Kryonaut paste.
 
housecat said:
I've run CB on this, and it holds 5.1Ghz on all 8 cores during a run if I remember correctly. I typically only care about what I actually use my machine for though. I mostly watch the stats for my use cases, because I think that's a bit more intelligent to focus on. The way mine is configured, it boosts up to 5.1Ghz on all 8 cores and holds it, never dropping at any point according to my hwinfo records. There's just a lot of thermal headroom with my IC Graphite pad and Noctua NH-D15, 30 or 40C worth in my use cases. If it's not a limiting factor, I move on and don't wring my hands over it endlessly. Performance is fantastic in daily use, unquestionably the fastest machine I've ever used so it's a mission accomplished.
Click to expand...
Well that's why I was asking what you meant by "full load". If we we are talking about Cinibench R23 or even Handbrake, I would contest 53C.

But yes, 11900k is very fast!
 
Last edited:
chameleoneel said:
Well that's why I was asking what you meant by "full load". If we we are talking about Cinibench R23 or even Handbrake, I would contest 53C.

But yes, 11900k is very fast!
Click to expand...
You're correct for sure, I don't recall as I pieced this together and ran all those typical benchmarks back in early May. Temps are higher than mid to low 50s that I see in my gaming. But I will note, my measurement isn't just while gaming, I use 8 thread Steam Link software encoding from my desktop PC to my TV. That's how I measured 53C. It's doing the encoding and gameplay. I consider that pretty decent as that's relatively intensive.

11900Ks are derided everywhere actually, people are totally brainwashed that this isn't the fastest single core CPU on the market. A buddy of mine has an Apple M1 and thought from all the hype it was so fast. We both ran https://web.basemark.com/ in MS Edge and I crushed him, of course. It's pretty obvious in latest titles like MS Flight Simulator or other cutting edge AAA titles that Intel still has the gaming lead. But mostly due to reviews that for some reason worry about power usage over actual performance the perception is twisted. If I were worried about power consumption, I'd buy a NUC or a tablet. 99% of the time my machine is on, it's idling. I don't care if it's 500 watts when I'm gaming. This is a software dev and gaming PC. One of the few proper benchmarks with MCE left on default settings (enabled)- https://www.eurogamer.net/articles/digitalfoundry-2021-intel-core-i9-11900k-i5-11600k-review?page=4 I liked my Ryzen chips too, I've had 4 of them. My 5900X was great in multicore and the massive cache results in amazing performance in older titles like CSGo. I don't know why everything has to be a fight between tribes. And as far as the price, I don't really care about that either. When I'm 100% happy with a system as I am now, I tend to use them for 5+ years so cost is irrelevant. So that's my battle cry for the Intel 11th gen on 14nm. I try to upgrade based on RAM technology. Low latency DDR5 20ns or so, when it arrives will be leaps and bounds over what I have now. Today they'll probably be about equal.
 
housecat said:
You're correct for sure, I don't recall as I pieced this together and ran all those typical benchmarks back in early May. Temps are higher than mid to low 50s that I see in my gaming. But I will note, my measurement isn't just while gaming, I use 8 thread Steam Link software encoding from my desktop PC to my TV. That's how I measured 53C. It's doing the encoding and gameplay. I consider that pretty decent as that's relatively intensive.

11900Ks are derided everywhere actually, people are totally brainwashed that this isn't the fastest single core CPU on the market. A buddy of mine has an Apple M1 and thought from all the hype it was so fast. We both ran https://web.basemark.com/ in MS Edge and I crushed him, of course. It's pretty obvious in latest titles like MS Flight Simulator or other cutting edge AAA titles that Intel still has the gaming lead. But mostly due to reviews that for some reason worry about power usage over actual performance the perception is twisted. If I were worried about power consumption, I'd buy a NUC or a tablet. 99% of the time my machine is on, it's idling. I don't care if it's 500 watts when I'm gaming. This is a software dev and gaming PC. One of the few proper benchmarks with MCE left on default settings (enabled)- https://www.eurogamer.net/articles/digitalfoundry-2021-intel-core-i9-11900k-i5-11600k-review?page=4 I liked my Ryzen chips too, I've had 4 of them. My 5900X was great in multicore and the massive cache results in amazing performance in older titles like CSGo. I don't know why everything has to be a fight between tribes. And as far as the price, I don't really care about that either. When I'm 100% happy with a system as I am now, I tend to use them for 5+ years so cost is irrelevant. So that's my battle cry for the Intel 11th gen on 14nm. I try to upgrade based on RAM technology. Low latency DDR5 20ns or so, when it arrives will be leaps and bounds over what I have now. Today they'll probably be about equal.
Click to expand...
One of the issues with the Rocket Lake reviews is that the bios code wasn't tuned well. Intel has released several updates since release and performance has crept up, especially in gaming. But most reviewers haven't revisited Rocket Lake to check that out. Also, 10 series seems to perform better on the 500 series motherboards.
 
I use the thermal grease from Startech and have never had any issues at all. It's easy to apply and also easy to remove the fan months later / clean and re-apply. It's pretty inexpensive at like $6.99 a tube.
 
GeneO said:
If you run your CPU hot a lot, I found thermal Grizzly pumps out and dries up and your temperatures start to rise. It is very good other than this.
Noctua NT-H2 paste is very good. It is what I am using now. I have an aio and my idle temeperatures are < 1c above ambient room temperature.
Click to expand...

housecat said:
The paste crowd has been split 50/50 for years on this. Some use pastes formulated for sub-ambient cooling like a lot of Thermal Grizzly pastes and maybe gain a celsius or two, but they do dry up. At risk of angering everyone, I believe it's just improper use based on a misreading of benchmarks and bad reviewers. Most people, who won't be posting here about it, have moved to just using Noctua's provided paste (which is great) or something like Arctic MX5. Even Arctic's lower end stuff is pretty good. I favor just using pastes that are formulated for ambient temp cooling if I'm doing ambient temp cooling. Nothing wrong with using the "better" pastes though, go for it. Just not enough gain to matter to me.

I'm getting average paste performance out of a IC Graphite thermal pad, maybe somewhere around Arctic Ceramique, and that's working well for me. My system holds 5.1Ghz on 8 cores during gaming and never blips, and runs around 53C at full load last I looked just last night. I'm not sure what my limiting factor is on my cooling, I'm guessing nothing. This particular chip runs at high voltage. An extra degree or two might extend my chip's life, but I doubt it. Limiting factor being just 14nm physics considering I run a NH-D15, but I don't really need it to be any better. Maxing out Intel Adaptive Boost as it's advertised is all I was shooting for with my build.
Click to expand...
Indeed, the Noctua NT-H1 and H2 paste are what I recommend for people whom want something which will last a long time and also isn't the most expensive.
 
chameleoneel said:
One of the issues with the Rocket Lake reviews is that the bios code wasn't tuned well. Intel has released several updates since release and performance has crept up, especially in gaming. But most reviewers haven't revisited Rocket Lake to check that out. Also, 10 series seems to perform better on the 500 series motherboards.
Click to expand...
Completely agree. Needs revisiting. You seem well informed, where are you on reddit? ;) I don't have a reddit account anymore, so I'm joking.
When they do the Alder Lake reviews though, if they disable MCE again (not only is it default on, and statistically few change BIOS options at all, as well- no one buying one would disable it), it may again still look bad and a "pointless gen". I wanted it for the Skylake bug fixes alone (L0 bug and parity bug). But also the TB4/USB4 support, PCIE4 (on Intel at least), and just overall IO updates like latest Bluetooth 5.2 support as I use BT constantly. For me a PC is more than just the performance specs, while I try to hit late-stage RAM evolutions and my next upgrade will be tail-end DDR5, I am interested in IO improvements. While I'd get some of that with 10th gen on Z590, I wouldn't get two M.2 slots on mITX, which I also wanted. I enjoy playing with a computer but don't like total teardowns, stressful because I'm picky about having everything right, so the extra $200+ USD over 10th gen was worth it to me. Newer chip means a teardown gets pushed out that much longer. But yup, if you ask reddit, there's nothing slower than an 11900K. Total garbage for a lot of money. I wouldn't be surprised if someone there with a Ryzen 1600X feels their system is faster than mine. It's 14nm, has to be worthless.

Though Intel does recommend MCE off if you enable ABT, in regular desktop usage I'm recording 5.3GHz on 7 of 8 cores simultaneously constantly with it on so I'm resisting. My 5.1Ghz all-core lock wouldn't change at all as they did ABT within 125W. But I do think ABT on + MCE off will make it look pretty good at this point. Be interesting to see if reviewers try to ignore RKL with the Alder Lake launch, most venues are only benching 10th gen. I haven't bothered comparing ABT on with MCE on vs off because I'm sure it'll kill near-constant the 5.3GHz desktop clocks that I see (upon activity, like messing with Spotify and Edge, etc).
 
housecat said:
Completely agree. Needs revisiting. You seem well informed, where are you on reddit? ;) I don't have a reddit account anymore, so I'm joking.
When they do the Alder Lake reviews though, if they disable MCE again (not only is it default on, and statistically few change BIOS options at all, as well- no one buying one would disable it), it may again still look bad and a "pointless gen". I wanted it for the Skylake bug fixes alone (L0 bug and parity bug). But also the TB4/USB4 support, PCIE4 (on Intel at least), and just overall IO updates like latest Bluetooth 5.2 support as I use BT constantly. For me a PC is more than just the performance specs, while I try to hit late-stage RAM evolutions and my next upgrade will be tail-end DDR5, I am interested in IO improvements. While I'd get some of that with 10th gen on Z590, I wouldn't get two M.2 slots on mITX, which I also wanted. I enjoy playing with a computer but don't like total teardowns, stressful because I'm picky about having everything right, so the extra $200+ USD over 10th gen was worth it to me. Newer chip means a teardown gets pushed out that much longer. But yup, if you ask reddit, there's nothing slower than an 11900K. Total garbage for a lot of money. I wouldn't be surprised if someone there with a Ryzen 1600X feels their system is faster than mine. It's 14nm, has to be worthless.

Though Intel does recommend MCE off if you enable ABT, in regular desktop usage I'm recording 5.3GHz on 7 of 8 cores simultaneously constantly with it on so I'm resisting. My 5.1Ghz all-core lock wouldn't change at all as they did ABT within 125W. But I do think ABT on + MCE off will make it look pretty good at this point. Be interesting to see if reviewers try to ignore RKL with the Alder Lake launch, most venues are only benching 10th gen. I haven't bothered comparing ABT on with MCE on vs off because I'm sure it'll kill near-constant the 5.3GHz desktop clocks that I see (upon activity, like messing with Spotify and Edge, etc).
Click to expand...
MSI is the only brand where their newer motherboards with 2 NVME slots offer both slots to 10th gen. They wired slot 1 to the CPU and to the chipset, so that 10th gen can use it. Their ITX Z590 MEG Unify and B560 Mortar MATX board both have this setup. And so do most of their other ITX and MATX boards. MSI was late to market with some of them, such as the Unify, Mortar, and B560I Gaming Edge ITX. Around June or something. Most of them weren't in the initial wave of reviews.
 
chameleoneel said:
MSI is the only brand where their newer motherboards with 2 NVME slots offer both slots to 10th gen. They wired slot 1 to the CPU and to the chipset, so that 10th gen can use it. Their ITX Z590 MEG Unify and B560 Mortar MATX board both have this setup. And so do most of their other ITX and MATX boards. MSI was late to market with some of them, such as the Unify, Mortar, and B560I Gaming Edge ITX. Around June or something. Most of them weren't in the initial wave of reviews.
Click to expand...
Didn't know that existed. Nice option for someone who wants a 10850K or 10900K and Z590. I thought hard on that configuration, who doesn't want the 10 cores? Especially since this was a 2nd machine for me alongside a 5900X, and it ended up replacing it after playing with both of them. My Z590 board was so expensive that the savings going with a 10850K no longer made that route appealing, especially losing an M.2. Just went for the whole thing.
 
DaeviousMax said:
That’s interesting, I still have my original spread of kryonaut on my 2700x built in 2018 and my recorded temps have not gone up since. That makes it 3 years old at this point…I wonder if maybe you don’t have enough mounting force on your cpu. Maybe the tighter the gap the more friction there is to prevent the paste from pumping out under heat?
Click to expand...
My mounting is spring loaded and very good even pressure and if I remove the CPU block then there is only a thin film left on each side. Could be that the excellent mount makes it more susceptible to pump out effect as there is very little paste left between the CPU and the cold plate after mounting (most gets squeezed out). I mostly measure in a repeatable way though with very little margin of error (around +/- 0.5 degrees). E.g. when I first installed kryonaut my 5900x was aprox. 32.5 degrees above ambient while after 4 months it was aprox. 35 degrees above ambient as average at stable state during a 3 minute recorded period in cinebench r20 all core load.
 
You must log in or register to reply here.
Back
Top