The Corsair User Forums

The Corsair User Forums (https://forum.corsair.com/forums/index.php)
-   Overclocking and Benchmarking (https://forum.corsair.com/forums/forumdisplay.php?f=131)
-   -   5200-5300 scores in CineBench R20 (https://forum.corsair.com/forums/showthread.php?t=193871)

NuclearLight 02-08-2020 01:10 PM

5200-5300 scores in CineBench R20
 
My current system, an Intel i9-9900KS, overclocked to 5.1 GHz (and sometimes 5.2 GHz) is scoring in the 5200 to 5300 range on Cinebench R20.

You can see a video of the Cinebench run here:


The temperatures get high, over 90C sometimes.

I thought I had the fans set at 100% in the BIOS on my Corsair H150i but they seem very quiet and never throttle faster. Is there a BIOS setting on my ROG motherboard I am missing?

c-attack 02-08-2020 02:37 PM

You only have BIOS fan control over the H150 if you plugged the fans directly into the motherboard. More likely you have them on the pump fan controller and those speeds are managed in iCUE.

I haven't run R20 in a while, but that is going to be all CPU/voltage heat versus the long duration type where fan speed matters.

NuclearLight 02-08-2020 09:56 PM

Quote:

Originally Posted by c-attack (Post 1034418)
You only have BIOS fan control over the H150 if you plugged the fans directly into the motherboard. More likely you have them on the pump fan controller and those speeds are managed in iCUE.

I haven't run R20 in a while, but that is going to be all CPU/voltage heat versus the long duration type where fan speed matters.

Thank you, I will double check that as soon as I can.

And here is a screenshot showing the temperatures.

https://i.ibb.co/2ZkhXnc/Cinebench-R20-5000-MHz.png

I read the "FAQ" post about the water-cooled radiators and I saw a mention there, but didn't make the connection until seeing your post!

Thanks again.

NuclearLight 02-08-2020 10:12 PM

Quote:

Originally Posted by c-attack (Post 1034418)
You only have BIOS fan control over the H150 if you plugged the fans directly into the motherboard.

All 3 of the fans are kind of bundled together and connected to one SATA power cable. There is a 3-pin connector of which only one pin is apparently a data delivery line for fan speed control. This is connected to the CPU Fan Header.

So my guess is all 3 fans are spun at the same speed ... but from ... where? Any idea?

DevBiker 02-08-2020 11:20 PM

As for your CPU temps - we'd need to see your vCore before we could even begin to speak about it intelligently, if that's what you are asking about.

c-attack 02-09-2020 09:52 AM

I did pull up and give R20 a run yesterday. Now I remember why I don't run it. You 9900 guys kill my poor little 8700K@5.2 and I can barely break 4000 pts in MC. However, it did hit a peak CPU package temp of 67C which for my delidded CPU under a custom water cooling system is a lot. In fact, it is more than when I run most stress tests at 100%. R20 is a high CPU temp benchmark, although a fairly short one.

Your scores are excellent and jealousy inducing, so the CPU is not throttling. As suggested above, this is all about voltage and the cooling really doesn't matter on a run that lasts less than 30 seconds. Are you running the factory defaults for the KS, which would be all cores at 50x100 and likely Auto voltage? If so, there is all your temperature. None of these motherboards will allow you to run a max load stress test out of the box and be cool. They are voltage heavy to make sure every CPU can boot and run. You'll want to find an overclock guide for the 9900K family and likely an Asus specific one. A lot of these motherboard settings are the same, but brand specific. Asus I can help with.

NuclearLight 02-09-2020 01:27 PM

Quote:

Originally Posted by DevBiker (Post 1034478)
As for your CPU temps - we'd need to see your vCore before we could even begin to speak about it intelligently, if that's what you are asking about.

Ok here they are. First boot and first run of the morning. System was nice and cool at start. Temps are usually higher.

https://i.ibb.co/Bth7Kmd/voltage-and-temps.png

I know the voltages are on the high side. This is for 5.1 GHz.

I'd just like to get the fans on the H150i to spin faster. They are hardly moving. The airflow over so much surface area of the fins could DEFINITELY lower the temperatures, even for a short test such as this. I know there is an asymptote out there, but I am nowhere near that on the heat curve.

Just so you know, I've done some crazy experimental stuff, such as 2-stage 2-phase passive thermosiphon cooling:

https://i.ibb.co/WDzFFWJ/20200103-182736.jpg

I've coupled heat exchangers to two Corsair H100s coupled to two cooling loops; one liquid, one vapor.

I just "missed something" when I hooked up your H150i this time around. I am a "tinkerer" which means I am bad at following directions.

And this is the case I built for the i9-9900KS. I made it big enough to house a quad-radiator at the top if I need it, but I started with the Corsair H150i.

https://i.ibb.co/v4H0M9R/case-size.jpg

You can see it is pretty big. And sturdy as can be.

DevBiker 02-09-2020 01:51 PM

Well, that's showing us the VID - what the core is asking for in terms of voltage. We need vCore - the voltage it is actually getting. We'd also need to see what those are under the load that's generating the heat. If you are using auto-overclocking settings, those voltages are going to be super-high. And that is going to be the problem, not the cooler.

High processor temps are usually indicative of too much voltage, not a cooler failure. Voltage generates heat at the pins. It's not immediately transferred to the cold plate - and there is some inefficiency with that anyway due to the TIM that Intel uses between the cores and the heatspreader. Yes, the 9000 series has soldered TIM but from the reviews I've seen, it's only marginally better than the paste TIM that they used previously. If Cinebench is using AVX instructions, that's gonna throw up some additional heat as well. No manner of cooling, short of LN2, is going to change that.

c-attack 02-09-2020 05:59 PM

You can create a custom curve any time you need with the + button. Within it are fixed RPM and fixed % settings. However, none of that will do much for a test that lasts 30 seconds. The most your coolant is going to go up is 1-2C and that means the most you can lower it is by 1-2C. Maybe if you were keeping the fans artificially low you might make up some ground, but this is not a heat exchanger or external chiller, so the coolant temp won't go any lower than the case ambient. It doesn't matter for a multi-thread R20 run. It may matter more for actually use and that seems to be what you should look at for temps vs this short high intensity benchmark --- unless you are about to start some CPU rendering projects. If the VID is 1.37 then I am going to assume your actual Vcore is elevated as well. Voltage=Temp for this. 0.05-0.10v less will reduce temps far more than any cooler settings.

NuclearLight 03-25-2020 02:51 PM

Quote:

Originally Posted by c-attack (Post 1034565)
You can create a custom curve any time you need with the + button. Within it are fixed RPM and fixed % settings. However, none of that will do much for a test that lasts 30 seconds. The most your coolant is going to go up is 1-2C and that means the most you can lower it is by 1-2C. Maybe if you were keeping the fans artificially low you might make up some ground, but this is not a heat exchanger or external chiller, so the coolant temp won't go any lower than the case ambient. It doesn't matter for a multi-thread R20 run. It may matter more for actually use and that seems to be what you should look at for temps vs this short high intensity benchmark --- unless you are about to start some CPU rendering projects. If the VID is 1.37 then I am going to assume your actual Vcore is elevated as well. Voltage=Temp for this. 0.05-0.10v less will reduce temps far more than any cooler settings.

OK finally! Sorry for the long delay between posts. I did not bookmark this forum's URL and I had experienced another hardware complication as well.

The only way I could get all 3 fans of the Corsair H150i to spin faster than the default "quite mode" of about 400 RPM was to do the following:

1. Locate a strange cable with a Micro-USB Hub connector on one side and a 10-pin monstrosity on the other end.
2. Attach the MicroUSB side to the Corsair H150i and the 10-pin connector (with many unused connectors) to the motherboard.
3. I had to download the CORSAIR ICue software from your site.
4. The "Performance" tab on the ICue program allowed me to specify the max setting, which spun the fans as high as 1500 RPMs and the pump speed even faster!

This lowered the temperatures from 97C to about 92C when running all 8 cores using a stressful prime number "sieving" application which was looking for 10-million digit prime number candidates.

So all's well that ends well. The computer is running much cooler now, even under the harshest tests on all 8 physical cores on the Intel i9-9900KS.

https://i.ibb.co/fG25hzL/20200121-175629.jpg


All times are GMT -4. The time now is 12:29 AM.

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2020, vBulletin Solutions, Inc.