Ok now I know why I ordered Kryonaut via Express shipping today. 
This is my test for the previous configuration, All Core 50, 1.300V, AVX 3
Every test was done 2 times, same conditions (of course not time).
Here the last version of Prime95 (https://www.guru3d.com/files-details/prime95-download.html). I've used the first one, the most power hungry/heat load, even if unrealistic.
@Wahaha360, I have also tested it under Blender and IntelBurnTest, but I will not post here, we decide to go for Prime95, AIDA, Cinebench).
Here you are, I guess we summarize in a later post??
Prime95, 1 HOUR, Test 1 and 2
AIDA 1 HOUR, Test 1 and 2
Cinebench R15, 5 times IN A ROW test (basically pressed RUN for 5 times when done, every time), Test 1 and 2
Every test was done 2 times, same conditions (of course not time
).Here the last version of Prime95 (https://www.guru3d.com/files-details/prime95-download.html). I've used the first one, the most power hungry/heat load, even if unrealistic.
@Wahaha360, I have also tested it under Blender and IntelBurnTest, but I will not post here, we decide to go for Prime95, AIDA, Cinebench).
Here you are, I guess we summarize in a later post??
Prime95, 1 HOUR, Test 1 and 2
AIDA 1 HOUR, Test 1 and 2
Cinebench R15, 5 times IN A ROW test (basically pressed RUN for 5 times when done, every time), Test 1 and 2
Second test, All Core 47, 1.275V, AVX 0
Again, every test was done 2 times.
BIOS pics for the configuration, and the screenshot of each test!
Prime95 1 HOUR, Test 1 and 2
AIDA 1 HOUR, Test 1 and 2
Cinebench R15, 5 times IN A ROW test (basically pressed RUN for 5 times when done, every time), Test 1 and 2
Again, every test was done 2 times.
BIOS pics for the configuration, and the screenshot of each test!
Prime95 1 HOUR, Test 1 and 2
AIDA 1 HOUR, Test 1 and 2
Cinebench R15, 5 times IN A ROW test (basically pressed RUN for 5 times when done, every time), Test 1 and 2
@Wahaha360
Last test, did only 1 time, because the results are really consistent.
Loaded DEFAULT settings on BIOS, DISABLED MCE and ENABLED XMP I.
This is to use the CPU at STOCK, as planned by Intel, with Turbo Boost etc.
Fan Profile STANDARD, AIO pump control disabled (Speed at 100%)
Here pics of the BIOS settings and after the results.
Prime95 1 HOUR
AIDA 1 HOUR
Cinebench R15, 5 times IN A ROW test
Last test, did only 1 time, because the results are really consistent.
Loaded DEFAULT settings on BIOS, DISABLED MCE and ENABLED XMP I.
This is to use the CPU at STOCK, as planned by Intel, with Turbo Boost etc.
Fan Profile STANDARD, AIO pump control disabled (Speed at 100%)
Here pics of the BIOS settings and after the results.
Prime95 1 HOUR
AIDA 1 HOUR
Cinebench R15, 5 times IN A ROW test
Ahahahah, don't worry! I've preferred to anticipate these because I have time to test the system until the 11th.
After I will be busier!
@Wahaha360 here the LS120 Test for the CPU Overclock at 5Ghz all core, AVX 3, 1.300V Vcore.
This high Vcore was necessary to run stable Cinebench R15.
Some personal thoughts:
- 99 Degrees under Prime95, SMALLEST sFFT option, is related to the high core voltage chosen. Prime can run with a 1.250V or lower with no problem and less heat!
- Personally, these are stress test, for stability mainly. A single 120 Rad, in this case, the HW Labs LS120, IMO is really good to cool this CPU under normal use or heavy loads as well! ( I'll share also The IntelBurnTest, RealBench 30min and Blender 25min, to show the temps).
- For a 120 rad, the sweet spot for OC the 9900K in my opinion is to have under AVX load, 4.5/4.6GHz for the CPU frequency, and IF you really want, limits the power to 205/210W, keeping the MCE enabled to force the CPU to stay with All core at 5GHz, but I have no idea under long games session.
Here the results:
Prime95 29.5 Build 9 1 HOUR
AIDA 1 HOUR
Cinebench R15, 5 TIME IN A ROW Test
And here, FYI, other test:
IntelBurnTest, temps taken at the 9TH loop
Blender 25MIN, Benchmark Scene
RealBench StressTest 30MIN
This high Vcore was necessary to run stable Cinebench R15.
Some personal thoughts:
- 99 Degrees under Prime95, SMALLEST sFFT option, is related to the high core voltage chosen. Prime can run with a 1.250V or lower with no problem and less heat!
- Personally, these are stress test, for stability mainly. A single 120 Rad, in this case, the HW Labs LS120, IMO is really good to cool this CPU under normal use or heavy loads as well! ( I'll share also The IntelBurnTest, RealBench 30min and Blender 25min, to show the temps).
- For a 120 rad, the sweet spot for OC the 9900K in my opinion is to have under AVX load, 4.5/4.6GHz for the CPU frequency, and IF you really want, limits the power to 205/210W, keeping the MCE enabled to force the CPU to stay with All core at 5GHz, but I have no idea under long games session.
Here the results:
Prime95 29.5 Build 9 1 HOUR
AIDA 1 HOUR
Cinebench R15, 5 TIME IN A ROW Test
And here, FYI, other test:
IntelBurnTest, temps taken at the 9TH loop
Blender 25MIN, Benchmark Scene
RealBench StressTest 30MIN
@fabio
Our TX240 results are off by 1C, but I think I know the problem, I have better ambient temperature control for the new few weeks, so I will resume testing.
I got the LS120, took a photo comparing the fin density on the radiator too.
* Hardware Labs LS120 on the left, Koolance HX-CU420V on the right
I got the right dust filter for testing too.
Our TX240 results are off by 1C, but I think I know the problem, I have better ambient temperature control for the new few weeks, so I will resume testing.
I got the LS120, took a photo comparing the fin density on the radiator too.
* Hardware Labs LS120 on the left, Koolance HX-CU420V on the right
I got the right dust filter for testing too.
That's some high FPI on the Koolance! I'd be super curious to see how noise results compare between the two and a decent pressure optimize fan (even just qualitatively). At a glance I'd imagine the Koolance out performing the hardware labs, but only if you can get enough airflow through it.
That's cool!!! Keep me updated! I can continuing to test if you need!@fabio
Our TX240 results are off by 1C, but I think I know the problem, I have better ambient temperature control for the new few weeks, so I will resume testing.
I got the LS120, took a photo comparing the fin density on the radiator too.
* Hardware Labs LS120 on the left, Koolance HX-CU420V on the right
I got the right dust filter for testing too.
I've always wanted to know the answer to the FPI question, and it seems like no one has ever had a definitive answer until now. So glad you're doing these tests! I can't wait to try this with the Der8auers OC frame on a 9900k, might be able to shave off a few degrees Celsius more
There are radiator roundups around, this one for example. Results are as you'd expect: at lower fan speeds (<1000RPM), high FPI rads perform near the bottom, while at higher speeds (>1800RPM) they perform at the top.
Bad news - on my way back from EU, Lufthansa may have lost my checked luggage = my components and prototype are gone.
They were suppose to find my luggage, it's been 48hours, chances of finding luggage after 48hours becomes slim.
I have the HWL LS120 and Koolance HX-CU420V radiator, but not much else.
FML
They were suppose to find my luggage, it's been 48hours, chances of finding luggage after 48hours becomes slim.
I have the HWL LS120 and Koolance HX-CU420V radiator, but not much else.
FML
Last edited:
They had one job! Have them pay for them to make another prototype and components. I keep a GPS luggage tracker in my suitcase because it's bound to happen sometime.
Last edited:
I read this and the article from Anandtech did back in 2015(? having a hard time finding it right now) they all seem to compare rad width and FPI as the same variable, where I'm more interested in only FPI with the same pump speed. These results in this link are normalized from a data set including higher and lower GPM than what I assume we are looking at with the 48mm pump has been proposed in the Fr33Flow threads. The fewer variables, the easier it is to pinpoint the bottleneck.
Similar threads
