Thermal Paste Specification Comparison

[Yongkit]

ARCTIC MX-6 (8 g) - Ultimate Performance Thermal Paste for CPU, Consoles, Graphics Cards, Laptops, Very High Hhermal Conductivity, Long Durability, Non-Conductive, Non-Capacitive https://amzn.asia/d/g4NiwXr

8g - S$22.68

 

[matique]

Managed to do some testing with the new Kryonaut Extreme paste:

13900K Stock Cinebench R23 Run : 85 degrees / 311W - 40587 points

13900K Tuned LLC Level 3 (DC 1.1 / AC 0.28) Stock clocks Cinebench R23 Run : 77 degrees / 266W - 40251 points

In comparison my 12900K with TFX at 5.2Ghz all core ran at 92 degrees and scored 29426 points in Cinebench R23.

However, 2 factors that changed was I had to change the contact frame on my Z690 Apex as for some reason I just couldn't get proper contact so I installed the Feng Zao contact frame in Fibreglass material and that drastically improved my temps as above. I think that could also have been a reason why my 12900k ran warm when overclocked with the Intel stock ILM but right now this sits at around 15 degrees cooler than my 12900K with tuned LLC on both CPUs! And 13900k runs all core 5.5Ghz all-core for P-cores and 4.3Ghz for E-cores. 12900k in comparison ran 5.2Ghz all-core for P-Cores and 4.1Ghz for E-cores.

The other factor is that I changed my coolant brand to some newer one that's apparently better than EK Cryofuel. Seems less foamy and much much easier to bleed.

Pretty happy so far with this paste together with the contact frame mod. Everything runs nice and cool now and with tuned LLC, 13900k runs really nice and stable. Looped Cinebench R23 and R15 for 30 mins each to test stability. 0 crashes.

Glad the contact frame is working out for ya man. The fibreglass is easy to use, bottoms out properly so don't need to worry about torque and is pretty cheap. Nice tuning too to get the power draw to be lower while maintaining stock perf!

 

[elmariachi]

Glad the contact frame is working out for ya man. The fibreglass is easy to use, bottoms out properly so don't need to worry about torque and is pretty cheap. Nice tuning too to get the power draw to be lower while maintaining stock perf!

Yeah man thanks alot for letting me know where to get it. The Intel ILM is horrible when I removed my block first time. Only the left half had contact and temps were very much higher. I'm glad that even your memory tuning works still properly with the contact frame. That was my biggest worry. The fiberglass contact frame I feel you can get away with slightly higher toque as the material is not as rigid as the other ones which are aluminium.

One way I found to optimize the torque for the bolts is to screw them evenly first till you have resistance equal on all 4 bolts, then tighten each side till the board spins slightly under the resistance of the screwdriver on the bolt. My lowest and highest P-core delta in Cinebench R23 is only 7 degrees. My 12900k was much higher closer to 12 degrees between hottest and coolest P-cores on stock Intel ILM.

 

[elmariachi]

Quick Cinebench R23 Run with Tuned LLC.
Temps and Power Draw in HWinfo.

 

[Encrypted11]

How well does Samsung B work on RPL?
Leaving the i9's better IMC and Hynix A/M kits aside.

 

[elmariachi]

How well does Samsung B work on RPL?
Leaving the i9's better IMC and Hynix A/M kits aside.

Most of the direction now is towards Hynix A Die. Even any M Die would run better than Samsung. Samsung runs hotter and can't clock that high. My current M Die at 6800C32 runs 1.5V and temps are under 40 during gaming. But the Samsung's would run no issue on most boards today with the newest bios.

Also, if you're on z690, better compatibility would be like 6400 kits M Die cause I've not seen many Z690 4 dimm boards running A die reliably as most of the kits start from 6600-7800 and those boards have difficulty running it. Z790 is a totally different story though even a hero is validated for 7600 but your IMC is also very important to run those speeds. So you have to evaluate your CPU IMC, Z690 or Z790 and make a decision.

 

[Koenig168]

8g of MX-6 for S$16-72. No need to think, just grab one!

https://www.amazon.sg/dp/B09VDLH5M6?psc=1&smid=ARPIJN329XQ0D&ref_=chk_typ_imgToDp

 

[Encrypted11]

Most of the direction now is towards Hynix A Die. Even any M Die would run better than Samsung. Samsung runs hotter and can't clock that high. My current M Die at 6800C32 runs 1.5V and temps are under 40 during gaming. But the Samsung's would run no issue on most boards today with the newest bios.

Also, if you're on z690, better compatibility would be like 6400 kits M Die cause I've not seen many Z690 4 dimm boards running A die reliably as most of the kits start from 6600-7800 and those boards have difficulty running it. Z790 is a totally different story though even a hero is validated for 7600 but your IMC is also very important to run those speeds. So you have to evaluate your CPU IMC, Z690 or Z790 and make a decision.

The 5.3GHz CFL-S 8700K still running strong for me.
The same old OCed CPU with DVFS + OCed RAM.
W11 seems harsher for memory stability.

If anything that’s spurring upgrades for me it’s the gleam outlook on tech side, possibly delays in the next few gens from huge job cuts hurting R&D, due to export control against CN market hitting big tech balance sheets if not I’ll side towards holding longer like @hlots123 sifu.

CFL’s very lacking IO / transfer speeds vs current platforms is also making me consider an upgrade.

 

[elmariachi]

The 5.3GHz CFL-S 8700K still running strong for me.
The same old OCed CPU with DVFS + OCed RAM.
W11 seems harsher for memory stability.

If anything that’s spurring upgrades for me it’s the gleam outlook on tech side, possibly delays in the next few gens from huge job cuts hurting R&D, due to export control against CN market hitting big tech balance sheets if not I’ll side towards holding longer like @hlots123 sifu.

CFL’s very lacking IO / transfer speeds vs current platforms is also making me consider an upgrade.

More than anything DDR5 itself is just worse off to getting it stable as it is extremely temperamental as far as temps go. You can run XMP at low voltage and still be unstable if your case have bad flow. To make matters worse, some RAM kits like G.skill do not have PMIC thermal pads which makes it even worse.

I am done with my 13900K upgrade and my A-Die. I am gonna skip 14th Gen after going through first adopter massive issues with Z690+12900K+DDR5. Time to take a step back and breathe. Even 4000 series can go fly kite. I am pretty happy with my 3080 Ti right now and 4000 series cards are just insanely priced. Time to enjoy my rig and piss people off more. @hlots123 been MIA for a while hope all good with him. He really got me into memory overclocking lol!

 

[Mach3.2]

So what did you do to get reported this time

Also, the SPR is pretty broken if your aim is on point or if you throw in rotational aim assist knn can vomit blood playing against a lobby of competent controller players abusing single shot weapons..

 

[elmariachi]

So what did you do to get reported this time

Also, the SPR is pretty broken if your aim is on point or if you throw in rotational aim assist knn can vomit blood playing against a lobby of competent controller players abusing single shot weapons..

Because I was the only guy on the server with 4ms ping and these faggots were all like 50+ms. They obviously do not have a clue what ping is and what it does.

SP-R 208 pretty broken at long range. I need 2 shots to kill or if the guy is wounded 1 shot is enough. But almost always 2 shots at long range. But maybe they didn't intend to make marksman rifles too strong that it makes sniper rifles with scopes irrelevant. Mid range the SP-R 208 is perfect that's why I pair it with my STB-556 Assault Rifle for rushing into rooms and corners. But yeah, fk aim assist on controllers. You go watch YouTube it's pretty fkin strong man. I also can get XIM for PC but no standard. My mouse aim is pretty good for a 41 yr old.

 

[Mach3.2]

Because I was the only guy on the server with 4ns ping and these faggots were all like 50+ns. They obviously do not have a clue what ping is and what it does.

SP-R 208 pretty broken at long range. I need 2 shots to kill or if the guy is wounded 1 shot is enough. But almost always 2 shots at long range. But maybe they didn't intend to make marksman rifles too strong that it makes sniper rifles with scopes irrelevant. Mid range the SP-R 208 is perfect that's why I pair it with my STB-556 Assault Rifle for rushing into rooms and corners. But yeah, fk aim assist on controllers. You go watch YouTube it's pretty fkin strong man. I also can get XIM for PC but no standard. My mouse aim is pretty good for a 41 yr old.

Nowadays kids just scream hacks when they get dunked on, no one recognises that they are getting outplayed anymore

You know it's busted when high level players using kbm is switching to controller lol....
Still pretty satisfying when you can get the 1 hit on mouse with the SPR, **** isn't easy esp when my aim is all over the place

 

[elmariachi]

Nowadays kids just scream hacks when they get dunked on, no one recognises that they are getting outplayed anymore

You know it's busted when high level players using kbm is switching to controller lol....
Still pretty satisfying when you can get the 1 hit on mouse with the SPR, **** isn't easy esp when my aim is all over the place

Yeah controllers are seriously meta. The aim assist on it is no joke. But I just simply can't play on them. I'd still stick to kbm and this mw2022 is weird you can get a few games with good kills and wins and the next few games after that will all completely go dogshit. Don't know if you experienced it. But it's the stupid SBMM algorithm testing you.

 

[elmariachi]

Managed to run a higher LLC and undervolted my CPU slightly while running the new feature of overclock on most Asus Z690 & Z790 bios. Basically, the 2 best cores will run at +1 bin over your stock all core frequency. So I have 2 cores at 5.6Ghz and the other 6 cores at stock 5.5Ghz. E-cores bumped to 4.5Ghz over stock 4.3Ghz.

This seems to run even cooler than my stock configuration with tuned LLC3. Right now running LLC7 with 1.25V manual Vcore and managed to stabilize ring at 5.0Ghz. Games max temps 67 degrees and Cinebench R23 is only 74 degrees. Score is 41,832. Did the CPU only bench on Timespy and got 26,896 points.

 

[Yongkit]

Managed to run a higher LLC and undervolted my CPU slightly while running the new feature of overclock on most Asus Z690 & Z790 bios. Basically, the 2 best cores will run at +1 bin over your stock all core frequency. So I have 2 cores at 5.6Ghz and the other 6 cores at stock 5.5Ghz. E-cores bumped to 4.5Ghz over stock 4.3Ghz.

This seems to run even cooler than my stock configuration with tuned LLC3. Right now running LLC7 with 1.25V manual Vcore and managed to stabilize ring at 5.0Ghz. Games max temps 67 degrees and Cinebench R23 is only 74 degrees. Score is 41,832. Did the CPU only bench on Timespy and got 26,896 points.

May I know if you have consider to try Aida64 FPU for temp reading?

 

[elmariachi]

May I know if you have consider to try Aida64 FPU for temp reading?

I haven't tried that. Usually I don't run those power viruses that loads the CPU unrealistically. I just used Cinebench R23 30 mins testing to check for stability and temps and my best test of all, is just gaming. I don't use my pc for anything else and stability for me is Cinebench R23 and Gaming.

The reason why I don't like running those heavy CPU load tests is that it puts an unrealistic load on your CPU that even encoding does not do that for example. If you're just a home user, I think Cinebench R23 and Gaming is a good enough test as long as you're stable. Testing consistently all these power viruses will only degrade your CPU over time. Take SinJulia and Mandel tests for example the system is overloaded during testing.

 

[Yongkit]

I haven't tried that. Usually I don't run those power viruses that loads the CPU unrealistically. I just used Cinebench R23 30 mins testing to check for stability and temps and my best test of all, is just gaming. I don't use my pc for anything else and stability for me is Cinebench R23 and Gaming.

The reason why I don't like running those heavy CPU load tests is that it puts an unrealistic load on your CPU that even encoding does not do that for example. If you're just a home user, I think Cinebench R23 and Gaming is a good enough test as long as you're stable. Testing consistently all these power viruses will only degrade your CPU.

It's true you are right.

I use the FPU to test overall cooling system capabilities.

Because game also have different load on CPU and some game like ace combat 7 is more cpu intensive load than other game work harder than r23 I test.

 

[elmariachi]

It's true you are right.

I use the FPU to test overall cooling system capabilities.

Because game also have different load on CPU and some game like ace combat 7 is more cpu intensive load than other game work harder than r23 I test.

I'm just abit more careful with this CPU as I've got a good one. But Cinebench R23 30 mins test is more than enough to saturate your cooling system. You don't need the CPU test. Ram testing however, you need to use TM5 1usmus config + Kahru 6000% (minimum) to test that your voltages for the CPU is enough. Initially I had a clock watchdog timeout bsod after setting my ram overclock and realise that I needed slightly more Vcore to run 6800Mhz and 5.0Ghz ring bus.

With the above, I think it's more than enough to determine the stability of your system for general use for home including intensive games. Unless you're looking to use your PC to control a nuclear plant, I wouldn't bother with absolute stability that needs your PC to be more than 200% stable.

 

[Yongkit]

I'm just abit more careful with this CPU as I've got a good one. But Cinebench R23 30 mins test is more than enough to saturate your cooling system. You don't need the CPU test. Ram testing however, you need to use TM5 1usmus config + Kahru 6000% (minimum) to test that your voltages for the CPU is enough. Initially I had a clock watchdog timeout bsod after setting my ram overclock and realise that I needed slightly more Vcore to run 6800Mhz and 5.0Ghz ring bus.

With the above, I think it's more than enough to determine the stability of your system for general use for home including intensive games. Unless you're looking to use your PC to control a nuclear plant, I wouldn't bother with absolute stability that needs your PC to be more than 200% stable.

Very valuable insight

After all my test I realized I can just use 65w eco mode for gaming at 4k since I do not fully utilitze the all core performance now.

 

[elmariachi]

Had abit of time to tweak my setup abit more. Managed 5.6 all core with 2 best cores 5.7Ghz. Bumped E-cores to 4.6Ghz.

This would be the max I can push on my 13th gen and cooling. Might just use this for benchmark profile as games it hits 75 degrees. Earlier one with 5.5Ghz all core with 2 cores at 5.6Ghz only 66 degrees max. Max temps 87 degrees when running Cinebench R23 with 316W. Broke 27k for Timespy CPU score.

Honestly love this Kryonaut Extreme paste. Temps are incredibly consistent.