[SOLVED] AIO questions/issues - Fractal Celsius+ S36 Prisma

[S13ontap]

So I've been trying to track down the cause of some issues that I've been having with the system I built. I have the aforementioned AIO, but I feel like it's running hotter than it should, or not cooling as well as it should... Does anybody know how to troubleshoot this kind of stuff? Maybe I just have something wonky, like being plugged into the wrong port - AIO in cpufan vs pump. Wrong settings for PWM vs auto on the pump/bios settings. Does anyone know how to test if pump is functioning properly? why it maxes the fans all when I switch it in off auto fan control to PWM in the bios... etc....

There is not a lot on these things in the way of issues or reviews, I'm wondering if someone here gas some general knowledge bombs they can drop before I toss this thing in a lake....

 

[Phaaze88]

I feel like it's running hotter than it should, or not cooling as well as it should...

Well, there could be various causes for this... more info would be needed.

Does anybody know how to troubleshoot this kind of stuff?

A)Feel the cpu block, the tubing, and the radiator. The heat should be felt throughout most of the unit.
B)Consider where, and against what, the unit is currently installed in the chassis. Liquid coolers are just hybrid air coolers, and are even more dependent on good chassis airflow than air coolers are.
Restricted airflow means cooling is more dependent on the liquid flow rate and radiator size.

C)Vertical and horizontal mounted radiators. Vertically mounted rads should be done with the tubes entering from the bottom; this allows the air pockets to get pushed into the rad and get trapped there. Not doing so introduces the pump to said pockets occasionally.
Horizontal top mount does not have this discrepancy.

Maybe I just have something wonky, like being plugged into the wrong port - AIO in cpufan vs pump. Wrong settings for PWM vs auto on the pump/bios settings. Does anyone know how to test if pump is functioning properly? why it maxes the fans all when I switch it in off auto fan control to PWM in the bios...

Just last month I installed a Kraken G12 + Celsius S36 onto my 1080Ti. All that was in the instruction manual...

 

[S13ontap]

not sure if this pic link will work here

Pics - https://1drv.ms/u/s!Amfqx2Kgr6gAgvcjOY1Rn6ndmz_qLQ?e=rcW2k9

System:
MSI MPG Z490
i7 10700k
GeForce RTX 2080Ti
XPG SPECTRIX S40G RGB NVMe M.2
Multiple SSDs
Gskill TridentZ ddr4 32gb
Fractal Celsius+ Prisma s360 AIO
EVGA supernova 750B2 PSU (should have lots of headroom for the new GPU and CPU - even with OC)

Fans:
(all fractal)
2 140mm front fans pulling in
1 140mm bottom fan pulling in
1 rear 140mm fan exhausting rear
3 120mm fans exhausting through top through fractal AIO rad
(working with old case, so balance isnt perfect yet. Swapping bottom 140mm for 120mm soon

rad is horizontal top mount

(Issues started when I swapped in the M.2 drive, the vertical GPU mount and the fractal aio for my corisair. Since, I've went back to standard GPU mount. but I think I've ruled it out as an issue anyway.)

one side of the tubing is warm, the other side hot.
Where I'm seeing those temps is in prime95 v.26.6 smallfft - doesnt hit that high at stock clock. but simply enabling extreme performance through MSI will cause it to run so high it gets to 1 or 0 from tjmax. pretty much throttles every core.

It didn't used to. and no changes. I'm gonna try and redo the paste today. see if it makes a differance.

As for the auto/PWM, I always had it on auto, and it was fine at first (fine-ish. i should say)
I've been having problems with stuttering and frame loss in games like warzone/metro/control etc. They hang up for a few seconds or worse, and my CPU response time goes to <Mod Edit>. then it catches up.

Haven't been able to see whats happening by monitoring while gaming. but stress tests show those numbers in hwinfo

I feel like heat on the pump, and heat on the one tube make sense. And cooler temps on the return tube also make sense.... but why the throttling and high temps with a simple 'click turbo' boost.....
Pretty sure it didn't do it before.

All new stuff though. so I've been ruling things out one by one.... removed vertical gpu mount again, redid paste a few days ago, cause I didn't like the preapplied stuff (I thought)

Only other thing is the m.2 drive mounted under the cpu. she runs hot too, but I understand that's normal....

Tempted to throw my corisair aio back in and see if it fixes the problem....

 

[Phaaze88]

Hmm...
That's a bit of a tight fit up there... the tubing isn't being pinched, is it? Plus you've essentially blocked off the far right fan.

Issues started when I swapped in the M.2 drive...

That Adata XPG in the pic? Yeah, I could see that being a problem for 3 reasons(compounded):
1)It's in the path of the gpu exhaust - typical; not normally an issue.
2)You're using an AIO instead of an air cooler on the cpu. One of the cons of that is no direct airflow for the VRMs and other nearby devices that an air cooler would normally provide.
3)The M.2 drive has LEDs on it. That's going to give it higher operative thermals.

I'd keep an eye on that drive's thermals. It's normal for the drive's controller to run warm - they actually perform better warm.
It's another story for the NAND chips themselves, however.

I feel like heat on the pump, and heat on the one tube make sense. And cooler temps on the return tube also make sense.... but why the throttling and high temps with a simple 'click turbo' boost.....

Yes, the heat should be spread throughout most of the unit normally.
As for one click turbos, they crank more Vcore than necessary by default; it's guaranteed stable though... but that's no good if it's running too hot.

Tempted to throw my corisair aio back in and see if it fixes the problem....

The advantage of the Celsius is that it does not require 3rd party software to operate.
The downside is that pump and fan control is shared; they can't be adjusted individually. Whether auto or PWM, the pump can't be run 100% without the fans doing the same.


There may be more, but it's slipped my mind for the time being.

 

[S13ontap]

The advantage of the Celsius is that it does not require 3rd party software to operate.
The downside is that pump and fan control is shared; they can't be adjusted individually. Whether auto or PWM, the pump can't be run 100% without the fans doing the same.

So that is what I'm looking into right now. I want to see if I can plug the pump into the pump header instead of cpufan, then plug the fans into the cpufan. So i can run the pump at 100, and the fans wil lbe controlled by cpu temps.

Opinions would be great, is my reasoning sound?
It doesn't seem to run very hard in auto fan curve (default) it sits around 1000 rpm and goes to like maybe 1500 max. even in testing under load. In PWM mode, selected on the pump manually, and on the bios, it runs wide open. Which i feel like, is what I want for the pump. but not the fans.
Thoughts?

Hmm...
That's a bit of a tight fit up there... the tubing isn't being pinched, is it? Plus you've essentially blocked off the far right fan.

More pics. - https://1drv.ms/u/s!Amfqx2Kgr6gAgvcornnf5dgVE1bjaw?e=fuF0JH

There's a bit more space than it seems, but it's not perfect. My corisair 2 fan rad seemed fine, but didnt come into the front as much.
worth noting that it powered by usb2, and then it plugged into cpufan seperately on it's wiring. So presumably not the same setup as the celsius
I know my JRAINBOW plugs are able to be separated on the fan hub (for lack of a better term) mounted on the rad, and I also think that the fans are as well, but I'd have to y them together or use sysfans or something....

Thoughts?

 

[S13ontap]

That Adata XPG in the pic? Yeah, I could see that being a problem for 3 reasons(compounded):
1)It's in the path of the gpu exhaust - typical; not normally an issue.
2)You're using an AIO instead of an air cooler on the cpu. One of the cons of that is no direct airflow for the VRMs and other nearby devices that an air cooler would normally provide.
3)The M.2 drive has LEDs on it. That's going to give it higher operative thermals.

I'd keep an eye on that drive's thermals. It's normal for the drive's controller to run warm - they actually perform better warm.
It's another story for the NAND chips themselves, however.

avg temp in hwinfo is 49degrees
probably more or less idle.
does that sound normal? Looks normal to me, but undoubtedly contributing to heat in the area.
I think the vertical GPU mount actually helps here... to allow airflow around the card, over the PCH and m.2 drive, out the rear exhaust

81 degrees avg when running diskmark64

I know.... argb is like cocaine.

 

[Karadjgne]

Try this. Lower the rpm on the bottom intake, a good 4-500 below the front intakes. This'll still put a good amount of air in, but without as much static pressure which will allow the front intakes to pick up the flow, without redirecting it from hitting the gpu area. Pull the rear exhaust and block off the port.

Sounds redundant lowering flow, but it's really the opposite. Whatll happen is the intakes now get air to the gpu. The gpus warmer exhaust will go straight up. Without the rear exhaust, you get a pure chimney effect, the airflow directed straight up into the waiting fans of the aio. With the rear exhaust, you end up stealing flow from the aio fan near the rear and with the front already compromised you actually loose out on aio cooling.

The 10700k is intrinsically the same thing as a 9900k only better. But it still has most of its thermal properties. Where it gets Trixie is in the mobo. Both MSI and Gigabyte have ignored Intel suggested specs. Instead of a 2 second burst, with a 56 second max turbo at Intel spec'd power, Msi and Gigabyte have changed the bios to 'factory optimized default' which includes a 4064A max current, 99999 second turbo. So instead of short burst, limited turbo time, lowering temps and power usage over time, your motherboard cranks the turbo beyond max, and holds it the indefinitely. For all intents and purposes it's an MSI Factory applied OC.

This is going to result in maximum cpu outputs, both in speeds and temps. Instead of hitting 224w for just a minute, then dropping to 125w, it's holding the full or higher wattage, depending on boost levels.

View: https://youtu.be/qQ_AETO7Fn4

You think you are stock, you really aren't, you'll be pushing @ 100w± higher than 'Intel stock' on that mobo, and under p95 full thread torture, that's a lot of additional power, all from the mobo, not the cpu.

Yes, you have should plug the pump into the pump header and run it maxed constant, fractal doesn't use a 'variable speed geometry' pump. Cpu fans should be on cpu header. You won't need SpeedFan or other, bios/msi software is quite capable of governing the cpu fans.

 

[S13ontap]

SO....
Thanks for the advice so far guys,

Been making some MAJOR headway.
Separating the pump and fans did it. From there, adjusting fan curve in bios got me into the low 80's for heat in smallfft. Now I'm just adjusting voltage down to get that dialed in. Set LLC to 1 instead of auto and dropped the voltage quite a bit.
But pump is now running full tilt. With the default setup for the AIO, as per their instructions, it was pumping at half speed

I also dialed in the fan rpm curves for the case, and set them to read from the PCH instead of CPU

At this point, now that Isee that decrease in temps, and I'm decent. I'm gonna start lowering voltage until i get under 80 in prime for 15 mins.

Although I've heard these new 10th gen buggers run hotter. Is 80 and under still the threshold I should be shooting for when testing for thermal compliance? or is there a hotter expected operating temp? say 85?

 

[Phaaze88]

85C, is the common suggested recommendation - although, I have set a thermal limit of 90C in the bios.

 

[Karadjgne]

Kinda depends on your usage. P95 is extreme case, but for somebody whom renders on a quite regular scale then extreme is more the norm, so below 80's would be better, below 70's better still. For someone who plays more simple games like roblox or minecraft or even CSGO, you'll be lucky to tax that cpu hard enough to get it out of the 50's, so having a max of 80 or even 90 doesn't really matter, you ain't getting anywhere close.

So you'll have to see where you stand. If gaming puts you in high 70's low 80s consistantly, might want to tinker some more, but low 70's is kosher. Depending on your comfort level.

 

[S13ontap]

Coming up on 77 max on only one core at 15 mins of smallfft. 74 degrees avg
Looks like it's sorted. Dialing down the voltage (sorta did this backwards) now until i get 50x on all cores as low as i can get the volts.
Then i can finally dial the clock up a bit and see what it can do.

Feels sorted guys. Hopefully I don't jinx it.

 

[S13ontap]

View: https://youtu.be/qQ_AETO7Fn4

Is there such a thing as voltage too low??
My buddy that does OC <Mod Edit> said this thing should be somewhere between 1.175-1.225 with LLC1

I'm booted, and browsing running P95 at 1.14 right now. Avg temps sub 70. some cores have hit 74 max.
I'd love to get this below 70 max everywhere. But Should I be worried about voltage so low?
Is the LLC setting just boosting me to a required voltage? Voltage in HWINFO is hovering at 1.144, but some cores occasionally jump to 1.164

 

[Phaaze88]

Is there such a thing as voltage too low??
My buddy that does OC <Mod Edit> said this thing should be somewhere between 1.175-1.225 with LLC1

I'm booted, and browsing running P95 at 1.14 right now. Avg temps sub 70. some cores have hit 74 max.
I'd love to get this below 70 max everywhere. But Should I be worried about voltage so low?
Is the LLC setting just boosting me to a required voltage? Voltage in HWINFO is hovering at 1.144, but some cores occasionally jump to 1.164

Something I have a bit of a problem with. Prime 95, small fft, AVX off, is more of a cooler stability test. The load this program puts on the cpu is too steady for measuring Vcore stability.
You should be running an app like Cinebench R20 'infinite loop' instead; constantly fluctuating workload = best for testing Vcore.

 

[Karadjgne]

P95 w/o AVX is temps only. It used to be stability too, up until @ 3rd-4th gen, but the newer editions of p95 used a different way of doing things. Small fft is still the same consecutive 100% load, but it's not as finicky and doesn't pull a bluescreen very often. Much more stable.

Voltage too low? No. Any cpu needs a certain set of voltages to run stable. Anything below those minimums creates instability. But, it's particular to every cpu, they are truly unique. Just in 3rd gen, there were certain batches that ran higher volts but lower temps, or lower volts and higher temps or low volts low temps etc.

The thing about OC is its a 3 way ordeal. Highest speeds for lowest stable voltage resulting in lowest temps. It's generally accepted knowledge that because every cpu is different, and responds differently, and has different needs, that Intel will set stock voltages high enough to cover All the cpus. Which is what you get using Auto. So Intel might set 1.25v, but there's nothing saying your cpu might only need 1.08v, or 1.114v etc. Only way to find out is lower the voltages in baby steps, and test until you glitch. As voltages go down, so do running temps.

 

[Phaaze88]

Slipped my mind earlier: Instead of plugging my Celsius S36 fans into the motherboard, I replaced them - they were reminding me of the old IPPCs I used to use - with NF-S12As, plugged them into the on-radiator fan headers, turned the knob to PWM mode, set 100% speed in the bios, so both pump and fans are running 100%.
This is actually working out for me because of the 3x NF-A14s in the front doing most of the work.

 

[Karadjgne]

It works, and you can live with the results. All that matters in anything.

 

[Phaaze88]

I thought it might have been a terrible idea at first - like, it should've been obvious to use the NF-F12s instead, but a LTT video and another article that was in dutch, both showed me that the difference between the 2 is within margin of error on thinner, less restrictive rads and oriented as push-pull. By themselves, the F12 pulls ahead a little.
The S12As ran quieter, so I bought those instead.

 

[S13ontap]

Sweet, I'll download cinebench and check it as I lower down

 

[Karadjgne]

Difference in design. S series are high airflow, low SP fans, so will work ok on a aio if speeds are higher. They really suck at low speeds as the minimal SP isn't strong enough to push the cfm through, push/pull helps Alot there. F12's are directed flow fans, the exhaust is almost totally perpendicular to the fan, not the more usual triangular shaped cone. So even mediocre SP is plenty as there's relatively little resistance. Far better at slower speeds than the S series, but the Series has better cfm at higher speeds. Trade off on opposite ends of the curve.

The A series is stronger in the middle range than both, but weaker in the extreme ends than both.
So F12's will be better at @ 500-800rpm, the A's at @ 700-1100rpm, the S's at 1000+ with slight overlaps.

Most ppl only see max values at max speeds, because that's all that's printed, very few realize that they almost never run fans at max, so what might look great on paper, can easily turn out like (explitive) in reality. Same as removing a fan and blocking the port with cardboard can actually improve airflow.

 

[Phaaze88]

They really suck at low speeds as the minimal SP isn't strong enough to push the cfm through, push/pull helps Alot there. F12's are directed flow fans, the exhaust is almost totally perpendicular to the fan, not the more usual triangular shaped cone. So even mediocre SP is plenty as there's relatively little resistance. Far better at slower speeds than the S series, but the Series has better cfm at higher speeds. Trade off on opposite ends of the curve.

That's exactly the conclusion those 2 samples showed. I was going to run the fans at max anyway - the S12s do just fine with less noise over the F12s.

The As are obviously a jack of all, master of none kind of fan, but I didn't need those, and they're not available in Chromax black anyway.

Most ppl only see max values at max speeds, because that's all that's printed, very few realize that they almost never run fans at max, so what might look great on paper, can easily turn out like (explitive) in reality. Same as removing a fan and blocking the port with cardboard can actually improve airflow.

Can't do much about that, when literally 2 companies among the popular ones post the curves for their fans; be quiet and Noctua.
Even knowing that CFM, SP, and noise levels don't scale linearly, we're still left with mostly inaccurate second guessing.

 

[S13ontap]

Here's a question. How come My multiplier is set to 51 on all cores, making my adjusted frequency 5100mhz
but hwinfo and all of my sensor programs all read that it's only at 47 or 4700mhz?

Someone was telling me it was AVX or something. I changed that from auto to 0, and same thing. so i changed it back.

Anyone know what would cause this? It won't load bios past 51. So I suspect I'm actually at 51, but nothing reads that. Although it used to when i was first tooling with stuff. before any overclocking in bios, in MSI dragon center monitor it was reading 5100 on some cores

Had to reset cmos battery boosting to 5.2
Haven't really messed with anything in bios. shouldnt have changed any settings.
would some software override it? nothing is running.

 

[Karadjgne]

You can set all cores to max out at 5100. But there's other settings like turbo max, which says that 1 core will max out at 5100, 2-3 cores will max at 4900, 4-6 cores will max out at 4800 and all cores max at 4700. But when you set 5100, it's not Per core, it's Actual core. You sat core 1 max is 5100, core 2 max is 5100 etc. And that's exactly what happens. Cpu can decide Which core sees 5100, which 2-3 can only reach 4900 etc. Similarly, you can set turbo max to 5600, but if you leave the multipliers at 5100, the cpu only gets to 5100, won't try to push 5600. The cap works in both directions.

You'd also either need to disable turbo and physically set each core, or bump turbo max to accommodate the full 5100, so when turbo is active 1 core sees 5100, 2-3 sees 5100 etc.

Using HWInfo, you see only what actual speed is, the turbo max is set lower than what core max is, so gets limited.

AVX offset only applies when the cpu recognise the use of AVX, so a setting of 2 will drop the clock speed 200MHz, but only if the testing program is using AVX.

 

[S13ontap]

I'll read the above when I get a sec and reply.
Had to set to allcore instead of per core to get MAX TURBO LIMIT and TURBO ATTENUATION to shut off and allow more than 4700mhz
I'd love to learn how to set it per core and get those limits turned off.

For now, fairly stable at 4.9 with a 4.8 ring. -Passes realbench for 1 hr fine, but locked up playing Warzone, and Deliver us the moon yesterday, so I upped the voltage a bit. HAven't tested gaming yet, but made a LLC change, and checked temps quick. running 5 degrees cooler now. So more room for Volts now

LLC4 - that was a temp gamechanger - now I have more room to push it
Vcore set to 1.175v but HWINFO has me sitting at 1.18v with max of 1.20v
2nd core seems to be a hot spot for me by a few degrees, and I still haven't tried to reapply paste. But with LLC4 set now it isn't hitting 80 anymore. peaks at 75 in thermal compliance testing So more room now to try and get 5ghz stable

Thoughts? I should have probably posted this in OCing. Seems like the AIO is good meow.

 

[S13ontap]

You can set all cores to max out at 5100. But there's other settings like turbo max, which says that 1 core will max out at 5100, 2-3 cores will max at 4900, 4-6 cores will max out at 4800 and all cores max at 4700. But when you set 5100, it's not Per core, it's Actual core. You sat core 1 max is 5100, core 2 max is 5100 etc. And that's exactly what happens. Cpu can decide Which core sees 5100, which 2-3 can only reach 4900 etc. Similarly, you can set turbo max to 5600, but if you leave the multipliers at 5100, the cpu only gets to 5100, won't try to push 5600. The cap works in both directions.

So where do I find the turbo max settings? sounds like you're right. But it seems to be set to 4700. I had all cores set to 5100 per core, and I only got 4700 in HWINFO. ever. Nothing ever boosted past there

You'd also either need to disable turbo and physically set each core, or bump turbo max to accommodate the full 5100, so when turbo is active 1 core sees 5100, 2-3 sees 5100 etc.

Pretty sure this is what I had done, as per above, does it sound like i did it wrong?

AVX offset only applies when the cpu recognise the use of AVX, so a setting of 2 will drop the clock speed 200MHz, but only if the testing program is using AVX.

Using P95 V2.26 - from what i understand it does not need to have AVX disabled, it doesn't factor it in, or test with it. Is that accurate?

 

[Karadjgne]

LLC is line level calibration. The cpu doesn't really work at a constant rate, there's very slight dips when it's demand is next to nothing, then it starts a new line, and has a gap, new line etc. It's those gaps in demand that have limited voltage, are the vdroop. But the cpu has to start work before it can demand enough voltage to cover the work. That's where you get instability, not enough supplied voltage to cover what the cpu needs, before it demands it. That's why LLC was created. It's a pre-emptive voltage, an addition to raise the vdroop up high enough to bridge the gap.

BUT, it's an added voltage, so also applies to the top of the demand too. For instance, cpu hits the gap in need, demand drops to 0.9v, starts a new line and needs 1.2v, but will demand 1.4v just in case. Being at 0.9v, for that instant its unstable, bluescreen. LLC can add 0.3v to the droop, maintaining the 1.2v necessary and the cpu is happy. But it also adds 0.3v to the demand, so the cpu gets 1.7v and your temps are not happy at all.

LLC should be used judiciously, for most there's never a need for 6 or extreme or 100% or you end up with a cpu using 1.7v or more. Not good for the cpu at all, and ppl can't get stable because they keep dropping vcore to avoid the 1.7v they just saw in Hwmonitor. Most OC will never need more than 3-4, medium, 55-66%.

Theres vcore, vid, cpu voltage and other names used, but what's important is figuring out if that's actual cpu need, cpu demand, cpu supplied and whether it's after LLC adjustment or before. And it's different with every software, every bios. Gotta learn your particular bios settings. Getting all that in balance makes for a successful OC at voltages the cpu can live with and temps you can deal with.