c-attack

Mistake at the factory.  These days almost every AIO has its own SATA or Molex power delivery.  Once upon a time smaller AIOs would draw their power directly from a motherboard fan header and so board makers started changing chassis fan headers into W_Pump or AIO_pump headers.  That came a bit late and with RGB everything needed more power than that.  So these days that single wire lead from the AIO to the motherboard is an optional safety feature.  When connected to CPU Fan, the motherboard will immediately throw up a warning (CPU boot error) if the pump does not talk back on start up.  This is a motherboard safety feature designed to prevent you from booting the PC with no CPU cooling method in place.  For an air cooled system, that becomes your CPU fans on the air box and if the fan fails to start, it sets it off.
 
However, this is not required and the BIOS will let you turn off this feature (no need for the connector at all) and the AIO has its own warning systems if the pump does not start.  It will flash red and the fans will get loud very fast.  Since CPU/OPT generally make poor case control headers, most users should connect the AIO tachometer lead to CPU fan as that will satisfy both motherboard and general safety needs.  

For quick static or effects that do not move from fan to fan, the lighting link or murals presets work just fine.  Where a lot of people are caught off guard is they get a bunch of LL or QL ring fans, these have some cool presets not found on other fans, but then they can't run ping, pong, infinity, etc. through the entire case.  If you do want to use the special QL/LL presets, set it up on each controller and then use complimentary lighting elsewhere.  
 
This brings fan planning per controller into play.  If you really want a wave or moving pattern to go up the front, over the top, and to the back, then you need to put those on one controller in sequential order from 1-6.  With all these new 10 fan cases, you likely have to make a choice between 3 front + 3 top as a group, leaving 3 side + 1 rear on the other or do the front/side 6 fans together and then top + rear as a group of 4.  No right answer here.  Depends on your viewing angle and preference.  You do want to keep the radiator fans on the Commander Core PWM side to make sure it works when CUE is not running, but PWM and RGB are separate so you can split the PWM and RGB function if needed for aesthetic purposes.  

No need for that.  The fan controller in the AIO is running the fans.  The BIOS and Windows do not affect that, other than for CUE general operation.
 
Standard first step if this happens with any fan on any controller is to pop the fan off it's power connector.  This is the twin fan connectors coming out of the pump.  Simply disconnect and reconnect.  
 
However, since both fans are doing this, it most likely is controller based.  Go to the CUE settings and check for a firmware update for the H115i XT.  It will says none available, but then "force firmware update" (or the French language version of that) will appear. When you click it the firmware will reload.  Fans and pump should max out for 5-10 seconds as this completes.  
 
If that does not work, you'll need to start looking at any effect Signal RGB might have on things.  While it's not intended to control fan speed, it may get locked up with CUE and cause the controller to misbehave.  If there is a real physical problem with the H115i XT, you'll need to demonstrate to Support Signal RGB is not impacting operations.  

On DDR5 most of the difference is going to be in the heat sink design, visual appearance, and physical module height.  Dominators have a slightly larger heat sink and with everything else being equal will run slightly cooler than the same exact IC under a Vengeance heat sink.  They are slightly taller as well and that could be an issue in compact systems.
 
However, with most all DDR5 right now IC change fairly quickly and the same stuff is being used under the hood for the same kits at the same frequency and capacity.  Since you are going for higher RAM capacity and stretching the RAM out to it's last rung of stability is not part of the plan, these subtle differences in temperature or binning are not going to be important.  You should choose the physical style you prefer, then look for the frequency and capacity you need.  If you are still undecided at that point, look to the primary timings to compare two kits at the same frequency and also the voltage.  1.45v is about the max daily level for most boards and kits, so if you have two kits with the same frequency and timings, but one is 0.05v less on voltage, it likely will run a bit cooler and/or have a bit more headroom.  

It likely works and those are not particularly aggressive timings, so I think plug and play XMP will be just fine.  You also have an "enhanced XMP" option on the Hero that will tweak a few tertiary read values and has a dramatic effect on throughput.  That should be easy as well.  However, be aware that if you have any issues related to the motherboard or CPU, companies tend to use your "non-compliance" with the QVL list as an offensive way to deny support, even when there is no possible connection.  That's just and FYI and it is most likely you never have problems with anything.

Are you getting a "pump fail" error from CUE after install?  If the pump was not circulating water, the small bit trapped in the block would slowly heat to 65C over several minutes and put the CPU temperature at 5-30C above that depending on voltage state.  This is what you would expect with a dead pump, but it should also trigger a CPU boot error from the BIOS and the pump fail notification in CUE and likely a (!) on the LCD top.
 
It also should not only do this when CUE loads.  A dead pump is dead and doesn't respond from the moment the power goes on.  Another user recently reported a similar condition, with it appearing to work fine until CUE loads.  The LCD top can complicate things.  Try the standard remedy of power down, PSU OFF, the disconnect the wide power cable on the Commander Core then reconnect.  Then pop the LCD top off and reconnect.  My best guess is the handshake on CUE load is failing (physical or software) and this causes the system to lock up.  The power cycle forces the components to do a full re-check which might help the physical.  If this is a software issue, I don't have any suggestions.

Yes, ignore.  I don't think it should be prompting the Wizard when you use a thermal probe since that is the native control source for the XT, but it's just an overly helpful set-up assistant.  You don't need it's help right now.  

That is not abnormal for Cinebench R20 or R23 on Kaby/Raptor Lake CPUs when you run the test in "standard" motherboard configuration.  Most motherboards will either boost or remove the limits entirely unless you actively put them back.  That makes running anything with AVX or other similar instructions a high volt affair and high temps will result.  Fan speed and orientation likely have no impact on this.  When your temps hit the mark at the end of the first pass, it has not been long enough for the cooler to lose much ground in the watts in vs watts dissipated battle.  Those 30-60 second tests are all voltage and conductivity.  
 
Presumably you don't need to run CPU renders all day.  If you do, then you definitely need to tune the BIOS for the job.  I would run a different stress test without AVX to provide a comparison point.  Try the "bench" test in CPU-Z.  It is linear in load and should provide a steady CPU load temp to watch.  If you see CPU temps jumping around with this, then that suggests you don't have good physical contact or bad TIM application.  Most likely it's going to be steady and a good 15C cooler than Cinebench.  Also take note of the load Vcore reading while running either CPU-Z or R20/23.  That value gives a good indication of why things are they way they are.  

Any other monitoring software active like HWinfo or AIDA?  Those full spectrum suites that poll the Commander Core at the same time as CUE can lick each other up. Common results are loss of control and disappearing UI elements. However, just the pump speed is very selective. 
 
I do not think this is a fatal error and you clearly do not have the red (!) in CUE indicative of a pure hardware failure. So is the problem the usb cable for the lcd top?  Or the wide communication power cable?  I can’t tell from this end. I think you need to hang on for support and from their end sending you a new LCD top seems a likely next step. 

As long as you are clear it does not matter whether your CPU temp is 30, 50, or 70C -- the fans react to the coolant temperature.  They can only remove heat from the liquid and so they don't need to speed up or down until the liquid temp changes.  If your coolant temp does not increase, the fans don't increase, and it is not affecting CPU temperature until that liquid temp starts to move up.  

The elite Cappelix models come with a Commander Core. It has independent RGB headers so it’s much easier to mix fans. You can combine the LL and ML on the same controller. It can’t make the ML do ring effects on a ring it doesn’t have, but you the UI will properly set up for whatever combination you use. 
 
Also be aware the current ML fan is the ML-elite. It has 8 center hub LEDs and that is what comes with the AIO. However, the prior model was the ML-Pro with 4 LEDs. You likely will prefer another 8 led fan, whether it is the ML-Elite, SP-Elite, or just to be released AF-Elite RGB. All can do the rear exhaust job. 
 
 

I think you are going to be fine using the Com Core paired with the LCD top splitter on one port and the Com XT on the other. Most Z790 appear to have remedied the compromises made on Z690 went they first introduced pci-e 4. What you can’t do is piggy back the Commander Core onto the Com XT “usb passthrough”. Those passthrough ports typically work for simple devices and neither the Com Core or LCD top fit that description. 
 
If there is an issue, using a usb 3->2 conversion cable is an option to create another path. Most users wind up needing a powered usb 2 hub when going beyond two devices. 

1. USB connections - Certain x570/B550 boards and their predecessor x470/B450 models with bios updates to make them compatible with the subsequent cpu generation have trouble with multiple devices stacked in the same usb line. This was a problem from the moment those boards came out and most users need a powered USB hub to help manage multiple connections. These are inexpensive but have a SATA connector for power and a circuit board to manage typically 4-5 usb 2 ports. The traditional unpowered “splitter” does not provide a steady connection and may cause frequent erratic connection. 
 
2. “SP fans” -> You need to be really careful with Corsair names. The SP-RGB fan came out in 2016 and is discontinued. You don’t have that so make sure your SP-Pro or SP-Elite is set to “8 LED fan series” in Lighting Setup. If you set it to SP fan it may not light or show a strange white pattern. 
 
Other possibilities involve a defect in the lighting chain. The Lighting Node Core is serial so if LED #6 fails on fan #1, then the lighting will come to a stop right there. A partial signal can cause weird flickering effects, but that does not seem to be the case here. Assuming top right fan is “#1” on the LNCore, swap it with fan 4 and see if lighting return to the first three fans. 
 
The usb issue is not related to the lighting. Even if you unplug the sun for the LNCore, the SATA power will produce the hardware lighting effect across all fans. You cannot selectively apply lighting in HW Mode so that eliminates usb connectivity as a cause for this. Corsair fans should always have lighting in HW Mode unless you save a static black pattern to the controller. 

Disconnect fan #2 from the H100i pro fan control wire and then immediately reconnect. You can do this with the power on. Sometimes a reconnect is all it takes. 
 
If this does not work, switch the two fans on the twin fan cables from the pump. If the new fan on #1 does not spin, you know the issue is on the fan side. If #2 is still 0 rpm then that suggests a problem on the AIO side. 
 
Contact Corsair Support, but better to have this isolated a fan or AIO issue first. Warranty is 5 years. If it is the fan, you could use any PWM fan available until a replacement arrives. 

Hardware Lighting setting only apply when CUE is not running. When CUE is active the software controlled “Lighting Effects” setting is in control. For RAM, the effect is immediately written to the modules when selected in HW Lighting. You will see a short delay before the selected effect appears in “preview”. That is when the effect is transferred to the module. 
 
So while you are setting the effect in the software, you are in software mode and the HW Lighting is just a projected preview. You must actually quit the software from the task bar to drop cue into HW Mode. If you intend to not to run CUE, make sure to deselect the “start with windows” option in the settings. 

If you go to the CUE settings (gear), you can disable the various integrations.  If you wish to selectively disable some aspects, go to "software and games" and then block the relevant parts.  If you do not want CUE to do use any of the plugins from Asus, Nvidia, or games, you can disable it all with one click from the Settings>General tab by deselecting "plugins" about half way down the list.  

For most users you want one in the rear exhaust area.  That can be inside the case or run out the back where you can't see it.  The exhaust air temp is the best indicator of your current internal case air temp and that is typically what you need your case fans to regulate.  The second probe does not need to be used for control purposes, but you can use it somewhere near the front intake to get a contrast data point vs the exhaust.
 
Where this gets a bit trickier is if you are running an AIO or other water cooling system with a front intake radiator.  In those circumstances, the coolant temp is going to control the front fans and the top and rear really should respond to what they do in order to keep balance of intake vs exhaust.  If you have a Corsair AIO, then H1xx Temp is already a control option on the Commander XT.  If you have another brand, you can use one of those temp probes right behind the front intake radiator.  Exhaust air temp from the radiator has a direct relationship with the coolant temp (small offset), so it will track +-1 with it.  You can then use that for control variable on the rest of the case or do a mix and match with the above.  
 
Aside from that, you can use the 2nd wherever you want for informational purposes,  

Yes, you need to make or buy an adapter to make it compatible with the Lighting Node Core “fan” port. Then in CUE you tell the software it is another device based on the number of LEDs you need (8 series LED, HD=12 LEDs etc). If you use the Lighting Node Core you are limited to using fan models as the building blocks in 4, 8, 12, or 16 led units. LED ports like on the Commander Pro or Lighting Node Pro have more options. EK often use odd numbers but you can deactivate any extra LEDs in the lighting effect. 

6 fans or less and you only need the Commander Core from the AIO kit. 7-12 fans means you’ll will combine that with the included Lighting Node Core (RGB) and PWM repeater hub (speed control). More than 12 and you need a different solution and additional gear. 
 
Commander Core - needs 1 SATA for power and 1 usb 2 connection to the motherboard for communication. 
 
Lighting Node Core - 1 SATA and 1 usb 2.0. 
 
PWM hub - 1 SATA for power and 1 PWM lead back to the Commander Core PWM side to make all those fans CUE programmable. 
 
See also: 
 

Yes, it will suck air in through the fan and then blow it out the back mesh.  PSU fan is likely to kick on when total load is 500W on the new model and there is also a secondary temp temp trigger you are not likely to encounter.  I don't think you will have too much trouble getting to 500W with a 4080 when gaming but it certainly does depend on the exact GPU model and what you're doing with it.  I would expect the fan to run when gaming.  

Your coolant temp was 34C before initiating the test, so climbing to 38C is not much of a jump. With the fans maxed you might expect a +7-8C coolant increase with a steady 300W load. Your cooler is not having trouble expelling the heat and you can’t have a zero degree coolant rise regardless of fan speed. 
 
The coolant temp to cpu temp delta is about 60C. That’s typical for any of the new CPUs right out of the box running synthetic tests with the usual built in “motherboard optimization”. Most board makers tweak their values to score well on review testing and also to make the worst cpu boot and run stable. If you maximally press it in this mode without refining your voltage and power targets, you will get high temps regardless of cooling method. A cpu that hits 95C plus 2 seconds after the test starts is not cooler limited. It’s voltage limited as suggested above. If you really want to run synthetic benchmarks, you’ll need to fine tune your voltage and power settings in the BIOS. Cooling is still conductive. If you create more heat than can be conducted through the processor, the AIO side doesn’t matter. Its job is heat transportation and disposal after the cpu. Either way fan and pump speed settings are only a small part of your number (34-39C). The larger part is voltage. 
 
34C idle in a 21C room is a bit higher than expected, but 34C is likely the internal case temperature as well. You may be able to check that versus a motherboard temp sensor or other data. You may or may not be able to change that. Case design, hardware layout, and even where you stick the case in the room is going to have an effect. If you stick it in a corner under a desk, heat has no where to go and you will continue to increase the internal temp as long as it’s running. Newer all glass cases will run warmer than something with mesh all around. 

The lighting on most Corsair controllers is serial.  If you have a dead LED or break in the circuit at LED #4 on fan 2, then the lighting will stop there.  Sometimes if the path is not completely severed you get some weird intermittent effect or changes in behavior from one boot to the next, but it is the same issue.  So assuming the bottom fan in your picture is fan #1, there appears to be a problem on fan #2.  You can check this by swapping fans 2 and 3 on the RGB Lighting Hub.  If #2 is the problem, the top and bottom fans will now be lit and 2 will remain the same.  You also can check by moving #2 to the #1 position and it will knock out the rest of the fans.
 
If you are still in your 2 year warranty period on the LL, contact Corsair Support.  Otherwise you appear to need 1 new LL fan.  

On serial controllers like the Commander Pro, Lighting Node Pro, and Lighting Node Core, you can only designate one fan type.  It's pretty simple math.  Pick fan type with Y number of LEDs.  Tell it Z number of fans.  CUE produces Y x Z number of LEDs in groups of Y.  So you get 8-8-8-8 or 16-16-16-16-16, etc.  This can work for a QL string with 1 LL fan on the end because the front of a QL is the same as QL.  In this instance you would tell CUE "QL fan x 4".  It generates 4 QL in the UI.  The front and back sides of the QL are rendered at the same time, so the only difference is the last LL doesn't have a back.  This is one of the few combinations that work with no drawbacks.
 
The Commander Core and Commander XT have independent RGB ports and drop the serial configuration.  This allows CUE to understand configs like 8-16-16-8-34.  Now it won't turn a center hub fan into a ring fan or vice versa, but you don't have to figure out your own greatest common multiple for the fan group.  Even though the Lighting Node Core would work for the 3 QL + LL, since you need the Commander XT for PWM fan control there is no reason not to use it's RGB side as well and save a USB connection.  As you have seen, those become the sticking point.  Also, if you need to break the 6 fans on the Com XT or C-Pro into smaller lighting groups, it is as simple as a click to make fans 1-2-3 do Pattern A and 4-5-6 do Pattern B.  The only reason to deliberately use more controllers than you need is if you were setting up the system to run without CUE active.  When in Hardware Mode (no software), each controller acts as its own indivisible group and all effects are uniform or 1-6.  

It probably needs to be the other way.  Can you post a screen shot what you are doing in the hardware lighting section for the RAM?
 
After you set the effect in Hardware Lighting, go down to the task bar and click on the ^ arrow.  Then left click on the CUE icon and select quit.  That should drop the RAM into the hardware lighting mode.  

Correct. You could program multiple complex lighting patterns and save them to the KB memory. Then set CUE not to run on startup. Freely use AIDA with no complications. However, with no other internal Corsair devices for cue and AIDA to fight over, I don’t think there is an issue with running it either. Nevertheless, if you are a minimalist then you have the option. 
 
Obviously the above poster is having a hard time and not happy to be at the tail end of the holiday crush. This is not the likely outcome of your experience, but if you have any doubts make sure you buy from a vendor with an amicable return policy. The K100 also carries a 2 year warranty.