Watercooling schema

[Solisius]

Hi!

 

What do you think of this schema?

This is my first watercooling... I'm quite sure this may be overcomplicated as I know the whole loop can only have few degrees differences but I'm sure this may be be better not to chain all waterblocks.

 

Also, I know watercooling SSD may be overkill, but, it's cheap, and I'll just watercool the one that could be intensively used with multiple VMs with complex simulation. What I am uncertain though is the fact that I'll use is a Samsung 990. This SSD is designed to dissipate heat from behind. The corsair waterblock for SSDs only have water on copper on the upper side of the SSD. What do you think about that, can you picture workarounds?

 

The config:

Mobo: ASUS ROG MAXIMUS Z790 HERO
CPU: i9 13900K with EK waterblock for the mobo + socket
GPU: MSI 4090 Suprim X with EK both side waterblock
SSD: Samsung 990

 

Thank you very much!
EDIT: Fixed probes locations

 

 

 

 

 

[Blue4130]

12 hours ago, Solisius said:

Hi!

 

What do you think of this schema?

This is my first watercooling... I'm quite sure this may be overcomplicated as I know the whole loop can only have few degrees differences but I'm sure this may be be better not to chain all waterblocks.

 

Also, I know watercooling SSD may be overkill, but, it's cheap, and I'll just watercool the one that could be intensively used with multiple VMs with complex simulation. What I am uncertain though is the fact that I'll use is a Samsung 990. This SSD is designed to dissipate heat from behind. The corsair waterblock for SSDs only have water on copper on the upper side of the SSD. What do you think about that, can you picture workarounds?

 

The config:

Mobo: ASUS ROG MAXIMUS Z790 HERO
CPU: i9 13900K with EK waterblock for the mobo + socket
GPU: MSI 4090 Suprim X with EK both side waterblock
SSD: Samsung 990

 

Thank you very much!
EDIT: Fixed probes locations

 

 

 

 

 

Seems excessive. Majorly overkill...

[thrasher_565]

there are ssd that have the chips on one side. but the controller cant go below a serten temp or there will be problems.

there has be no proof that cooling a ssd will make it last longer.. its just a guess cooler better right? the cost to cool it is already a wast of moeny and can be spent on replacing it if it dies... but you can cool it for looks thow 

 

my build will do this and some. also 8x 560...

 

 

 

[Solisius]

7 hours ago, Blue4130 said:

Seems excessive. Majorly overkill...

Hi!

Well, in the first gen of Samsung SSDs there was some thermal throttling. However, now that mobos have integrated heat spreaders and that thermal management with latest gen seems to be really great...  Maybe you're right, and I'm really only wasting time with this idea of watercooling SSD.

 

But maybe you're also talking about the radiators here, you think I'm wasting radiators here?

[Solisius]

7 hours ago, thrasher_565 said:

there are ssd that have the chips on one side. but the controller cant go below a serten temp or there will be problems.

there has be no proof that cooling a ssd will make it last longer.. its just a guess cooler better right? the cost to cool it is already a wast of moeny and can be spent on replacing it if it dies... but you can cool it for looks thow 

 

my build will do this and some. also 8x 560...

 

 

 

Hi! Very nice build hehe!

 

Hum, I see, about the controller being too cold, do you think it can happen in a normal watercooling loop? Like, I thought this kind of things may happen with below zero extreme cooling stuff but not in watercooling, am I right?

 

About my build, I feel I'd like it to be useful in some way, if it's not at all and may even cause unexpected behaviors... Maybe I should pass.

 

Specifically about the Samsung 990 Pro (this is the SSD I have), Samsung designed a copper coated heat spreader on the back for the NAND and DRAM. The controller has a nickel heat spreader that radiate on top of the SSD, so V-NAND and DRAM won't really be in contact with the watercooled copper plate, only the controller will. V-NAND and DRAM will be cooled from the metal backplate included in the Corsair Hydro X Series XM2.

[Blue4130]

49 minutes ago, Solisius said:

Hi!

Well, in the first gen of Samsung SSDs there was some thermal throttling. However, now that mobos have integrated heat spreaders and that thermal management with latest gen seems to be really great...  Maybe you're right, and I'm really only wasting time with this idea of watercooling SSD.

 

But maybe you're also talking about the radiators here, you think I'm wasting radiators here?

I mean the radiators. You have three 480's, a 360 and a 140 where most people would have a 360 and OK temps...

[Solisius]

46 minutes ago, Blue4130 said:

I mean the radiators. You have three 480's, a 360 and a 140 where most people would have a 360 and OK temps...

I see, thank you very much for the input. Well, I was looking for OC but with low noise, so I'll use PWM fans (and pumps).

For CPU (this 13900K is great for winter :P) + GPU, I estimated max TDP of 1100W, you can be OK with a 360?

[Blue4130]

49 minutes ago, Solisius said:

I see, thank you very much for the input. Well, I was looking for OC but with low noise, so I'll use PWM fans (and pumps).

For CPU (this 13900K is great for winter :P) + GPU, I estimated max TDP of 1100W, you can be OK with a 360?

Don't let my words steer you away, I love over built systems. I think you are 300 to 500 watts over estimating max power however.

 

Curios, what kind of case are you planning on putting so many rads into?

[Solisius]

40 minutes ago, Blue4130 said:

Don't let my words steer you away, I love over built systems. I think you are 300 to 500 watts over estimating max power however.

 

Curios, what kind of case are you planning on putting so many rads into?

Oh ok, maybe I should get rid of the 140 rad but keep the fan there for airflow purpose because I need air for RAM and some SSDs (actually I'm not even sure I need an extractor there if I keep extractors on top of the case...). This would simplify the loop without loosing perf, what do you think?

 

Well, I was charmed by the Lian Li V3000 Plus black xD.

[Solisius]

Weeeeell, a simplified version that should be enough, cheaper and simpler. The only doubt I have is about having a good airflow for the RAM and the SSDs not watercooled!

 

Also, I hope that in idle usage, I can run the loop without fans on. Not sure if it's possible...

 

 

 

[thrasher_565]

4 hours ago, Solisius said:

Hi! Very nice build hehe!

 

Hum, I see, about the controller being too cold, do you think it can happen in a normal watercooling loop? Like, I thought this kind of things may happen with below zero extreme cooling stuff but not in watercooling, am I right?

 

About my build, I feel I'd like it to be useful in some way, if it's not at all and may even cause unexpected behaviors... Maybe I should pass.

 

Specifically about the Samsung 990 Pro (this is the SSD I have), Samsung designed a copper coated heat spreader on the back for the NAND and DRAM. The controller has a nickel heat spreader that radiate on top of the SSD, so V-NAND and DRAM won't really be in contact with the watercooled copper plate, only the controller will. V-NAND and DRAM will be cooled from the metal backplate included in the Corsair Hydro X Series XM2.

i have no idea all i no is some part on the ssd dose not like being too cold or it complanes. linus did a video on this. i think 40c was a sweet spot if i remember?

 

my ssd are only 500gb and are thee for looks so i could care less if there slower or not.. and i when out of my way to get one with all the chips on the top side.  but the ssd water blocks cost like $75 each...

Edited August 5, 2023 by thrasher_565

[Solisius]

14 minutes ago, thrasher_565 said:

i have no idea all i no is some part on the ssd dose not like being too cold or it complanes. linus did a video on this. i think 40c was a sweet spot if i remember?

 

my ssd are only 500gb and are thee for looks so i could care less if there slower or not.. and i when out of my way to get one with all the chips on the top side.  but the ssd water blocks cost like $75 each...

The Corsair waterblock costs like 35 USD in my country. I could even get it at 27 USD if I had been fast enough on Amazon.

 

However, I don't want performance issue on my side. I have checked some videos about watercooling SSDs Linus made, and he said as long as you don't do subzero, you should be fine with longevity increased (or just no benefits if you're unlucky). But I think he made several videos about watercooling SSDs, can you retrieve one about performance issues with watercooling? For a Samsung 2To 990 Pro I think this kind of make sense in the end as this SSD is a beast xD.

 

Thank you very much for your help!

[thrasher_565]

10 hours ago, Solisius said:

The Corsair waterblock costs like 35 USD in my country. I could even get it at 27 USD if I had been fast enough on Amazon.

 

However, I don't want performance issue on my side. I have checked some videos about watercooling SSDs Linus made, and he said as long as you don't do subzero, you should be fine with longevity increased (or just no benefits if you're unlucky). But I think he made several videos about watercooling SSDs, can you retrieve one about performance issues with watercooling? For a Samsung 2To 990 Pro I think this kind of make sense in the end as this SSD is a beast xD.

 

Thank you very much for your help!

i think this is it?

 

 

[Solisius]

Oh yeah, thanks! In this one he had his little tiny waterblock for 7 usd and it was working like a chill charm on his side!

[Solisius]

Last update for those who may be interested...

According to this info:

I've reworked airflow!

 

I also have used thin radiators in pull because according to this, only push+pull really benefit from thick rads.

https://www.ekwb.com/blog/radiators-part-3-surface-thickness/

 

What do you think?

[thrasher_565]

the fittings are normally not in the center of the rad. might be better to make the bottom rad go to the res and it be last.

 

so pump,cpu,gpu,ssd,top rad,front rad, bottem rad, pump/res

[Solisius]

Hum yes. The idea was to distribute rads so that the loop gets few degrees cooler before entering the GPU. In practice though, I know it only makes very little difference because a loop tends to equalize temp (but still there is like 4/5° diff). So... Does it worth the complexity... ? Maybe not, you're right!

[thrasher_565]

1 hour ago, Solisius said:

Hum yes. The idea was to distribute rads so that the loop gets few degrees cooler before entering the GPU. In practice though, I know it only makes very little difference because a loop tends to equalize temp (but still there is like 4/5° diff). So... Does it worth the complexity... ? Maybe not, you're right!

i dont no. if you had enough rads in between components then i guess it could but like 99% of people say it wont matter.

2 loops could make a difference if you have enough rad per component.  or the difference from cpu and gpu are much more.

but people say that even if the water temp is slightly cooler it wont matter when it come done to ocing witch is dead today. them money spend on cooling can be spent on the nexed tire up.

 

looks and nose are basically the only reason to water cool.

cpus are hot today but you really need to delid them to get any real performance from them. kinda true even for older cpus... 

[Solisius]

On 8/7/2023 at 7:42 PM, thrasher_565 said:

i dont no. if you had enough rads in between components then i guess it could but like 99% of people say it wont matter.

2 loops could make a difference if you have enough rad per component.  or the difference from cpu and gpu are much more.

but people say that even if the water temp is slightly cooler it wont matter when it come done to ocing witch is dead today. them money spend on cooling can be spent on the nexed tire up.

 

looks and nose are basically the only reason to water cool.

cpus are hot today but you really need to delid them to get any real performance from them. kinda true even for older cpus... 

You know what? You're right, I should do direct die cooling with the new EK Waterblock dedicated to it and even use Conductonaut for the GPU!  The EK waterblocks are nickel plated copper, so it should be perfect for it (Conductonaut on pure copper have extreme melting problem in the long run, this is limited with nickel even if there is different quality of nickel coating but EK uses electrolytic coating as far as I know which is very good for this purpose)!

[thrasher_565]

2 hours ago, Solisius said:

You know what? You're right, I should do direct die cooling with the new EK Waterblock dedicated to it and even use Conductonaut for the GPU!  The EK waterblocks are nickel plated copper, so it should be perfect for it (Conductonaut on pure copper have extreme melting problem in the long run, this is limited with nickel even if there is different quality of nickel coating but EK uses electrolytic coating as far as I know which is very good for this purpose)!

now your talking. thow a tec on there too... delid, tec, direct diy and like 3x480 rads... thousand of dollars to get one boost bin!

[Solisius]

12 hours ago, thrasher_565 said:

now your talking. thow a tec on there too... delid, tec, direct diy and like 3x480 rads... thousand of dollars to get one boost bin!

You mean the boost won't worth it anyway? :P. Maybe... May the silicon lottery ever be in our favor!

[thrasher_565]

26 minutes ago, Solisius said:

You mean the boost won't worth it anyway? :P. Maybe... May the silicon lottery ever be in our favor!

well new stuff you only get like 5% more from ocing maybe 10% on water and like 14% on ln2... so you just buy a faster part if you want more speed...

gone are the days from going from 2.66 to 4.0...

 

a bost bin is were you up the speed of the part. it gose in set amounts. but is largely do to temp. just like Mario bros is in sets of 22 frames. even if you get there faster you cant leave till the nexed 22 frames.

[Solisius]

49 minutes ago, thrasher_565 said:

well new stuff you only get like 5% more from ocing maybe 10% on water and like 14% on ln2... so you just buy a faster part if you want more speed...

gone are the days from going from 2.66 to 4.0...

 

a bost bin is were you up the speed of the part. it gose in set amounts. but is largely do to temp. just like Mario bros is in sets of 22 frames. even if you get there faster you cant leave till the nexed 22 frames.

Hummm, only 14% on ln2... Wahou, that's not much :O. Well, at least I'll be able to exploit the 13900K at it full potential, so something like 10% more than with low perf air cooling. I hope the great mobo will help to handle the power needed for like 10% more.

For the 4090 I wanna use the 666W Galax BIOS with the two side EK waterblock and see what I can do.

 

Let's just hope I can provide enough power with a 1300W power supply.

 

Also, I wish idle usage to be quiet and low power consumption. I don't know how much the power saving features can impact overclock.

[thrasher_565]

10 hours ago, Solisius said:

Hummm, only 14% on ln2... Wahou, that's not much :O. Well, at least I'll be able to exploit the 13900K at it full potential, so something like 10% more than with low perf air cooling. I hope the great mobo will help to handle the power needed for like 10% more.

For the 4090 I wanna use the 666W Galax BIOS with the two side EK waterblock and see what I can do.

 

Let's just hope I can provide enough power with a 1300W power supply.

 

Also, I wish idle usage to be quiet and low power consumption. I don't know how much the power saving features can impact overclock.

well i dont no what the true oc number is but that what i got for say j2c and like gn. the stuff comes oced out the box and the xmp basically makes it as good as it gets.  things people do is under volt it to drop the temps some. but you need power to oc so.... i heard there might still be ocing ram but i think you need good ram to start... but i dont really no how to oc so take that as that.

 

but it is fun to water cool thow