[SOLVED] H150i not so great results

[icyulkn]

I feel like I'm not getting my money's worth. I have a 9900k and getting not so great 97C in RealBench at 1.285V @ 5GZ following this guide https://www.gigabyte.com/FileUpload/...le/525/946.pdf. Idle hovers at around 41C. One of the fans rpm is at 1430 all the time. What do you guys normally get?

 

[Phaaze88]

Link broken.
Can you post full PC specs?

 

[icyulkn]

i9900k, corsair vengeance pro 64 gb 3200 MHZ (3800MHZ overclock), MSI 1080 ti Gaming X 11G, Aorus Master Z390, Is this OK?

https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/525/946.pdf

 

[Phaaze88]

Chassis?
Number of fans?
Airflow setup?
Psu make and model?

I'll take a look at the guide in the meantime.

 

[icyulkn]

Corsair 570x, 3+3 fans, pull, PSU I can't get to but it's 750 W + 80 GOLD (it's a better model). I have the performance mode set to "Extreme " on all fans and the pump in iCUE. Thanks for your help!

 

[Phaaze88]

I see...
I feel like that guide has left some things out and was vague on others...
LLC: Turbo is high enough - heck, even High is plenty. Stay away from Extreme - that's better suited for LN2 - that just cranks too much voltage through the cpu after the Vdroop.
AVX: It told you about the uncore, but not the AVX Offset. That should've been set to 3 also. You can work on getting it closer to 0 after you find a stable voltage for 5.0ghz.
TJmax: That should stay at 100C.
Prime95: Small FFTs - Ok, but the guide should've told you to disable the AVX instructions below in the same window. Find your stable non-AVX frequency first, then do the AVX.

Corsair 570x, 3+3 fans, pull

Ok, so the AIO is front intake...
The top 3 fans are exhaust?
The rear fan is also exhaust?

Is the radiator mounted with the tubing entering from the bottom?

 

[icyulkn]

It's at the top. I do have it set to Turbo so I'll try cranking that down as well.

 

[Phaaze88]

Is it possible to 180 it, or would that be too much of a stretch?
The reason is: AIOs are not topped off with fluid, so there will be some air present. There is a particular situation encountered with front mounted rads that isn't present with the top mounted ones.
With the rad positioned with the tubing entering from the bottom, the pump will push the air into the rad which then rises to the other end where it gets trapped. You actually want it there.
But with the tubing the other way, it tends to get trapped in the tubing, going back and forth from the rad to the pump, which you never want to be exposed to air while running; it'll wear out faster that way.

 

[Karadjgne]

So 3x intake at front. 3x exhaust on aio up top. Rear fan is next to useless as it steals air from the aio.

While I agree mostly with Phaase88, I'd do the AVX backwards. Set AVX offset for 0, run Prime95 small fft with all 3 AVX checked disable for ½ hour. After that, if temps are decent, then run same test with 1st box, AVX unchecked (you'll never see AVX2 or AVX-512 as those are professional and specialized applications) and compare temps. Set AVX offset to 2 or 3 whichever responds to being closest but gets better fps. If the offset creates instability, forget about it and leave it at 0, games don't use high enough amounts to make really any difference but a few temporary °C.

LLC doesn't need to be anything higher than @ 50-66 or medium-high, otherwise it adds too much voltage that's wasted on nothing but heat. You set vcore for 1.275 but with turbo/extreme its realistically running at closer to 1.4v if not higher. Which does nothing but cook your VRM's and possibly creating instability.

 

[icyulkn]

I can 180 it at a later date sure. Couldn't hurt right? I'll mess with the LLC and report. Thanks for your help.

 

[Phaaze88]

I can 180 it at a later date sure. Couldn't hurt right?

Not at all. It'll help the unit last longer if it fits that way.

 

[icyulkn]

Is it possible to 180 it, or would that be too much of a stretch?
The reason is: AIOs are not topped off with fluid, so there will be some air present. There is a particular situation encountered with front mounted rads that isn't present with the top mounted ones.
With the rad positioned with the tubing entering from the bottom, the pump will push the air into the rad which then rises to the other end where it gets trapped. You actually want it there.
But with the tubing the other way, it tends to get trapped in the tubing, going back and forth from the rad to the pump, which you never want to be exposed to air while running; it'll wear out faster that way.

So 3x intake at front. 3x exhaust on aio up top. Rear fan is next to useless as it steals air from the aio.

While I agree mostly with Phaase88, I'd do the AVX backwards. Set AVX offset for 0, run Prime95 small fft with all 3 AVX checked disable for ½ hour. After that, if temps are decent, then run same test with 1st box, AVX unchecked (you'll never see AVX2 or AVX-512 as those are professional and specialized applications) and compare temps. Set AVX offset to 2 or 3 whichever responds to being closest but gets better fps. If the offset creates instability, forget about it and leave it at 0, games don't use high enough amounts to make really any difference but a few temporary °C.

LLC doesn't need to be anything higher than @ 50-66 or medium-high, otherwise it adds too much voltage that's wasted on nothing but heat. You set vcore for 1.275 but with turbo/extreme its realistically running at closer to 1.4v if not higher. Which does nothing but cook your VRM's and possibly creating instability.

I'm giving up hope trying to get it to medium or high LLC. I tried AUTO too. I am trying RealBench and keep getting a BSOD. I know you said Prime but I wanted to try RealBench. Should that be any different? Getting higher temps too.

 

[Phaaze88]

There is a difference, actually.
Prime95 stress testing runs a constant, steady load, and is ideal for testing cooler thermal stability. It is not as effective for checking voltage stability.
Asus Realbench and Cinebench R20 are fluctuating workloads, and ideal for testing OC voltage stability, but not for thermal stability.
Get it?
You're getting the BSOD in Realbench because the current voltage setting isn't actually stable.

 

[Gmoney06ss]

Prime is excellent for temperature compliance. As it only gits the cpu really hard. Realbench taxes the entire system, so it may be issues elsewhere, other than the overclock. It's a better stability compliance test. Try to un overclock your ram and just focus on cpu until stability is confirmed, and then move to ram.

 

[icyulkn]

I'll un overclock the ram and try again.

 

[Phaaze88]

@Gmoney06ss is right about Asus Realbench stressing the entire system.
For voltage testing the cpu only, run Cinebench R20.
Open R20, click on File > Preferences > Minimum test duration. Enter an impossible number, and click Ok.
Then run it for at least 1 hour - obviously, the longer you can do it, the better.

 

[Karadjgne]

RealBench or C20 do not work hand in hand with Prime95. You must breach that gap in your head. It's pointless to have a stable pc, if Prime says it's getting 95°C to get it and just as pointless if Prime says the temps are an awesome 57°C, yet RealBench says the OC is unstable.

So you need to tweak the OC, not only to satisfy Prime95 and temps, but also to satisfy RealBench/C20 stability. By getting the best combination of both, you'll have the best OC outcome, regardless of whether that's a certain MHz lower or higher than what you decided you wanted.

 

[icyulkn]

I changed some things in iCUE and now I'm getting 90 C in RealBench @5 GHZ. That's VERY comforting. I had to tweak the fan curve a tad (set to 100% always), it could have just been an iCUE error, but I dunno really. I think even when I set it to EXTREME setting in iCUE something doesn't happen to the fans correctly, or the pump. My room is cooler today too so that might of helped. I didn't have it on quiet mode I'm pretty sure of it. I can hopefully get it lower when I mess with the RAM.

I can't find the h150i anywhere now except at astronomical prices. I paid a hefty $199 for mine at ecollegepc.com. Wish they had the Thermaltake Floe Triple Riing RGB. Guess I'll live. Thanks for your help and let me know what you think about the Floe.