Mp5works BPC plumbing

[TT158]

Hi All

I've tried to copy the Framechasers chap's plumbing for the Mp5works BPC. Looks aside, can you please comment on whether or not this will work? I'm unsure about the whole parallel plumbing thing that this part requires.

Image of proposed plumbing here:
http://imgur.com/gallery/5ZJj317

 

[TT158]

Yeah I agree it looks a bit ****. Maybe a bit better once it's installed and the BPC hoses will be held down by the gpu inlet hose.

All of my loop will be front facing though, so I'm thinking my only alternative is to have the BPC inlet hose splitting off of the pump exit, and the BPC outlet hose join the top of the reservoir, which will also look kinda messy, and require more fittings.

I suppose I could also cut down the hoses and unsleeve the tube, so long as it's not hard to get such hoses again.

Anyway thanks for confirming that this can work

 

[TT158]

not convinced. the main water inlet to the gpu block will be straight thru to the back block and then that goes straight to the outlet of the gpu block. this therefore puts the back block in parallel with the main gpu block. If I were you, I'd be trying to have the back block parallel to a tubing run rather than a component.

Forgive me for being dumb while I try to get my head around this, but I am struggling to see a big difference right now:

On entry: yes, the water is straight through to the BPC inlet, but will the resistance of the small hose not force water into the GPU block? Although I have to say I'm worried (and admittedly uneducated) about the fittings on the BPC which you just push the hose into. Would a T junction before the GPU block be better? i.e. Direct flow of water into GPU block, 90 degree turn before GPU block into BPC inlet.

On exit from the GPU block: the water gets directed vertically through the T-fitting, and the BPC return joins it (10mm or so after the GPU block exit). How is this different from water exiting the GPU block, going through a hose, and then having the BPC return enter the hose (say, 100mm after the GPU block exit)?

Inlet from the pump is fine but the return by the rad depends on your set up. Although i said it will work its definitely not ideal .

OK. So, inlet from pump, return at top of pump reservoir sounds better?

 

[TT158]

The difference is that with it parallel to a tube run, all the water still goes through the gpu block. the way you have it you're splitting the flow between the gpu block and the back block. the gpu block is likely far more flow resistant than the back block and you may find that the flow through the more critical component is too reduced.

You'll have to try it and see I suppose.

Sorry, I get you now, I think! I wasn't considering the high resistance of the GPU block.

Ideally the return before the rad so it cools before going into the pump resoiver but depends on your loop set up.

That makes sense, I think I'll feed the backplate cooler after reservoir, then return to the loop before first or second radiator.

I would avoid the gpu completely and have the inlet from the res tapped on a t piece and then return at a rad port on a t piece again.

Much cleaner look.

Did you do a test for comparison for before and after? Especially on the core temps, i've seen people say they've had a drop of upto 2-4c on the core as well with the backplate.

I've not built the loop yet as one of my orders got delayed and I only recently got all of the stuff. Work permitting I'll do it this weekend though.

My plan is to plumb the back plate cooler in but not attach it to the backplate, run some tests, then attach it. Very interested to see what difference it makes!

Thanks All for your help!

 

[TT158]

Dont forget a drain valve at the lowest point .

Planned it for one of the spare ports at the bottom of the pump housing

 

[TT158]

They have just released a G1/4 version. You can either buy the entire unit, or a replacement top for £30

Thanks for the heads up! Might get that in black. I imagine it's quite restrictive, but could be good

 

[TT158]

Ill do a test on restrictiveness between the parallel version, serial version, X4 Ram Block and X6 Ram block when my parts arrive. Ill let you know.

That would be amazing for the community, thank you!

 

[TT158]

@OC2000 That's awesome, thank you! Minus 2 core and minus 4 memory junction is actually quite decent, I think! I've actually just hooked up the serial version to my first loop (leak testing now before firing up tonight), so I'll stick with that for the time being. I might go serial for cleanliness at some time in the future, and a couple of degrees are always welcome

Will be interesting to hear how the Kryographics goes. That's what I wanted but they're not doing one for the FE yet, I don't think.

 

[TT158]

I’ve installed the Aqua computer water block and it performs worse than the EK even with its active back plate.

was a nightmare to install and produces louder coil whine than my Strix did.

Results are under same conditions

GPU temp - 46c
Memory temp - 78C!!
Flow - 4.89 (an improvement there)

not going to bother trying it with the MP5works. No point. The flow will be the same, but the temps most likely won’t be better than the EK.

im going back to the EK water block and applying the revised thermal pad placement and will test without VRM pads to see how that works as without thermal pads on the back of the VRMs there’s no coil whine.

one thing to note though. I have the first revision parallel MP5 works that has no fins. You most likely have the revised version, so apart from the serial version bleeding faster there may not be a difference in temps.

it does look good though

Wow that's really surprising from Aquacomputer. A touch disappointing but at least you've got the EK!

I just powered up my first custom loop and tested with and without the BPC. I first ran Port Royal stress test for about ten minutes then went straight to OCCT for ten minutes. The results surprised me a bit, with a fairly decent memory junction temp drop from the BPC in Port Royal, a massive drop in OCCT for memory temp, but basically no drop for GPU core temp. Oddly with the BPC in OCCT there seemed to be more heat generated. Maybe just a warmer room by then or something.

I wouldn't rule out user error on my part along the way, but here's what I saw

Port Royal
No BPC
GPU 46-47, Memory Junction 70
Fans 1,400 rpm , water temp 33c

BPC
GPU 46, Memory Junction 64
Fans 1,400rpm, water temp 33c

OCCT
No BPC
GPU 48, Memory Junction 68
Fans 1,400 rpm, water temp 33c

BPC
GPU 48-49, Memory Junction 58
Fans 1,450 rpm, water temp 34c

Block is a Corsair XG7 3090 FE.

Edit: 400W draw, +140 core, +750 memory

Anyway, thanks to all on this thread for your help on this

 

[TT158]

forgot to mention l had +1000 on the mem on after burner for those tests. Can’t believe I forgot that . The gpu was at stock using the 520W EVGA bios though. I’ll edit my post.

Oh good point.

I wish the FE had an option for a bios like that! Mine was at 400W draw, +140 core, +750 memory

 

[TT158]

I’ve gone back to the EK water block.

the difference in temps between the thermal pads on the back plate VRMs and with out is 1C all be it at a short stress test, but the coil whine is gone and I’ll monitor it.
My reasoning behind removing them is based on the quality of the stock back plates. They are not going to be doing much to help the VRM so I’m sure it will be fine. The top of the VRM is still thermal padded and water cooled too

No coil whine by removing thermal pads from the VRMs? Would this apply to other cards, too? My 3090 FE appears better than it was stock, but there is still a bit.

 

[TT158]

With the 3090 KFA SG with an Alphacool block having just added the parallel mp5works block I see about 6-8C less on memory junction temperature on load, bringing it down from ~74C to ~66C, and that brings these temps to a range I'm much happier with. Haven't yet touched or measured the backplate since adding the mp5works block though. It got really hot before.

Sounds like it's in the expected range. One of the YouTube channels (maybe eTeknix?) said that their OC was more stable with the addition of the BPC. I didn't try going past my air cooled OC (+140 core) with the waterblock on but BPC off, but did some testing earlier today with it up to 185 and it seems rock solid (passed Port Royal and Timespy Extreme stress tests, and a couple of hours of Watchdogs legion). Hard for me to say what it would have behaved like without BPC, but would be interested to know if you find a difference.