Hmm ok.  I dont see a vid.  I see a MSpaint chart and an article that compares rad cooling performance per fan RPM.  But nothing comparing a system's component order with a before/after based on a constant rad/fan setup .  Agree to disagree, i am new to this after all so I suppose this was dumb.
 
Cheers!

redteamgo
Hmm ok.  I dont see a vid.  I see a MSpaint chart and an article that compares rad cooling performance per fan RPM.  But nothing comparing a system's component order with a before/after based on a constant rad/fan setup .  Agree to disagree, i am new to this after all so I suppose this was dumb.
 
Cheers!

Your post is #121 of the thread, go to post 110

nosomo
Only the initial.  The hybrid fix is close to the POC as seen here https://imgur.com/a/BQvcTzG.

The part for air units is a little more unique and will be shown / shipped at the same time I put a provisional on it. 

I have a feeling that a lot of the poor memory cooling off the hybrid comes from the plate being thin. I noticed with the stock pads, there is a lot of deflection in the plate, which will produce a less than perfect interface between the plate and AIO.

You can just about double the plate thickness. I think the stock plate is 0.030" and the AIO had an offset of 0.059". Using normal materials, you could go up to 0.050" material and not worry about it being too thick. Unfortunately, the next size up in standard copper is 0.062", I think it would be too thick. Might be worth a try though.

From there, you can shim it like you have to get the thermal pad thinner.

FWIW, I've hear the Strix LC card runs really cool on the memory. It uses a full size flat plate and then ~2mm thick pads. This goes along with my thinking the issue is more the interface between the plate and the AIO.

https://www.techpowerup.c...led/images/cooler6.jpg

03whitegsr

nosomo
Only the initial.  The hybrid fix is close to the POC as seen here .

The part for air units is a little more unique and will be shown / shipped at the same time I put a provisional on it. 

I have a feeling that a lot of the poor memory cooling off the hybrid comes from the plate being thin. I noticed with the stock pads, there is a lot of deflection in the plate, which will produce a less than perfect interface between the plate and AIO.

You can just about double the plate thickness. I think the stock plate is 0.030" and the AIO had an offset of 0.059". Using normal materials, you could go up to 0.050" material and not worry about it being too thick. Unfortunately, the next size up in standard copper is 0.062", I think it would be too thick. Might be worth a try though.

From there, you can shim it like you have to get the thermal pad thinner.

FWIW, I've hear the Strix LC card runs really cool on the memory. It uses a full size flat plate and then ~2mm thick pads. This goes along with my thinking the issue is more the interface between the plate and the AIO.

I recently converted a 3090 FTW3 to a hybrid. Previously with air, GPU temps were in the 70's and the VRAM was in the 80's celcius. With the Hybrid conversion I'm gaming with the GPU in the 40's and the VRAM/Hotspot is in the 60's. During Benchmarks I may hit low 50's for the GPU depending upon ambient and high 60's with the VRAM/Hotspot. I used the supplied thermal pads, and re-used the old Thermal pads from the air cooler on the stock backplate for the hybrid. Only thing I did to lower the VRAM/CGPU temps further after the conversion was placing a couple of ML120's on the backplate. I believe that lowered the temps an additional 6 degrees or so.


 
 
 
 
 
 

nosomo

redteamgo
OK.  Lets lay out some basic math.  Hopefully I got it right.  If your loop throughput is even just 25L/m.  These rads have a volume of what?  200-350ml tops?  and you're moving 25L/min on a "normal" bios speed?  That means you've got 400+ml per second of fluid moving through the rad.  How much is a typical loop in total volume?  1-2L ?  Maybe 3?  And you're saying rearranging the order of the rads is going to drop 4C?  How can that be when the same droplet of coolant will have visited every component in 4-6 seconds?  If all of the rads have a fan, whether push or pull, the claim is that your rad order will shift loop temp by -4C favorable?  Lets also not forget that water's viscosity increases as temperature increases.
 
I am dubious but would love to be proven wrong by a systematically designed test with a temperature sensor before and after each rad and one in the res.
 
Not trying to argue, just understand the claim.  I can't get past the (seemingly?) logic faults in such a large improvement relative to component order where throughput, volume, surface area and fans are constant on a system on even a conservative curve.

There's literally a video above with 7 rads and temp sensors at each one -- where do you think the data is pulled from?

thanks for pointing out the exact post.  stacking radiators in that test does not extrapolate to what the OP has in an apples to apples manner.  OP will not get 4C from your suggestion.  Maybe if he or she is lucky, it will squeeze 1-2C tops, probably much less.  Is that worth the work?  Only OP can decide.

rzelek506

nosomo
For optimal efficiency / cooling, you want the hottest air coming in contact with the hottest water first.  With lower flow pumps like a D5 with that many items in the loop, the loop temp could end up a few degrees out of balance vs optimal.  In a setup like yours, that's where the rads like nexxxos st30 come into play where they inlet on one end and outlet the other.  That lets you move from the front to the back of the case w/ out a dedicated line run (they're also less restrictive to flow too).

If you feel like playing with the loop and can control the room temp (maybe put it in a larger room w/ temp control), try the  pump - CPU - GPU - 120 rad - top 360 rad - side intake rad -res setup after collecting current config data.

a vid on radiator placement / stacking (test 1 and test 2 are what you should pay attention to)
https://www.youtube.com/watch?v=vauAJl29xlw

redteamgo
 
thanks for pointing out the exact post.  stacking radiators in that test does not extrapolate to what the OP has in an apples to apples manner.  OP will not get 4C from your suggestion.  Maybe if he or she is lucky, it will squeeze 1-2C tops, probably much less.  Is that worth the work?  Only OP can decide.

My current loop order is: pump - side intake 360 rad - GPU - 120 exhaust rad - CPU - top 360 exhaust rad - res
I don't think it will really be worth moving the pump or getting new rads for this build, because it would involve completely deconstructing the build and I doubt the st30 rad would even fit in this chassis, given how I have everything setup. But I'll probably do that for my next build instead.
But from the video and from what I believe nosomo is saying, the actual improvement (vs balancing the loop temp sensor data) would mainly come from moving the 120 rad so that it's after the CPU as opposed to before the CPU, when referring to my current loop order.
That way the hot air would be moving through hotter water in the 120 rad, while the coldest air (intake rad) would still be moving through the coldest water, which already occurs after all the exhaust rads.
I don't really know how much of an improvement moving that 120 rad would be, since my loop is already pretty similar to what nosomo recommended (in terms of which air is hitting which rad).
Even if the loop temp is guaranteed to be better by X amount, I would probably try and do that for the next time I plan on draining the loop, as long as everything fits ok, in terms of spacing/routing.
But I certainly wouldn't drain the loop only just to see this small improvement.

I did not mean to be argumentative.  I misread the original post and I get what nosomo meant now.  my fault.
 
rig looks amazing 

Kasque
how did you manage to put your hand on this graphic card ?

Waited in queue like the rest of us.
See https://forums.evga.com/N...-Process-m3108092.aspx

rzelek506
Update 10/2/2021: Edited EVGA HC vs Optimus Absolute GPU block comparison section (my second post in this thread):
- Reran some tests
- Removed some graphs from playing games because that workload has too many variables IMO (i.e. actual scenes/stages in the game are not always consistent between runs). So I just kept 3dmark results because that should be a perfectly consistent workload between runs.
- Added an average temp graph for more stable results besides just the peak temps.
- Removed TimeSpyEST graph because it looked like that test doesn't keep the system temps in equilibrium for a long enough time, when trying to get an average temp graph with stable results. 
 
I think these updated graphs and results in that section are a little more accurate than some of the previous graphs that I posted earlier.

very cool, the optimus block is a beast.
 
just curious and confirming - when testing the EVGA HC block, was it truly factory?
 
or was it repasted, backplate pads, front pads etc?

I've just converted my 3080ti ftw3 ultra to hybrid, using the evga hybrid kit which I bought new second hand for cheap.
I've replaced all pads which came with it with high quality "Nab Cooling thermal pad 15/wmk" (can find on amazon)
used 3mm and 1mm pads, I did squeeze alil-bit the 3mm before closing up and made sure contact was even and well made (also added 3mm pads to the backplate)
I didn't replace the default radiator fans it comes with (although thought of putting Arctic P12 instead)
 
as for my result, radiator placed as intake at push/pull config in P500A case:
 
Used aggressive fan curve profile for tests / Room temp around 26C
Heaven benchmark@ 150/1000 OC peaked 80C temp on air ->>> on hybrid peaking now 55 
TimeSpy@ 170/1200 OC AVG temp 53C :  3dmark dot com/spy/23248347 on hybrid ->>> was averaging at 74C on lower OC before 
 
Ain't fully remember my hotspot/mem on air before but think they were alil high before, now on hybrid hotspot peaking avg 60s and mem avg 70s
Quite satisfied with the conversion overall, lower noise and lower temps overall with higher avg boost clocks.
 
(gonna sort gpu cables&radiator soon ^^)

I realize this thread long ago went dormant, but you guys were rather helpful in a world where most forums have a bunch of arrogant asses who make it less desirable to ask questions.
So, I've long finished my first couple loop builds just mostly learning as I went.  I decided to swing for the fences and picked up a Lian Li DK-05F desk case. (I don't really care that a lot of people find them outdated at this point, it looks badass to me and that is all that matters)  I think I'm going to go hard tube this time for aesthetics, but I'm not sure which material to use.  I'm leaning Acrylic at this point.  I have an absolutely massive amount of space to work with, and I think I'm going to install two systems in it since I have two systems running my 3 rtx3090's at this point.  I have one ftw3 model that has front and back active blocks, one ftw3 that has a VRAM plate and the factory water cooler unit that I am going to swap out to the front and back active blocks.  My 3rd 3090 is a factory Dell card that has active blocks front and back.  Running as I have them now they run at a pretty constant 72-74 celsius on the VRAM. 
I recall that hard tube can deform, so I'm not sure what kind of room I have to work with there.  I read a bunch of what seems to be negative posts on petg tubing and it's reaction with certain other products.  I'm not doing glass, that's just outside my wheelhouse. Anything you guys want to throw at me, please feel free.  I'd rather ask 1000 questions before I get into this than stumble and buy a bunch or worthless crap as I go.