CPU-360 Midplate Testing

[Waterlogged]

None of them. Single will likely be the "best" of the inserts but will still likely be a little higher. I think the single needs a shorter slot for it to show improvement.

[Vapor]

Double was bad, really bad. Pulled it, put stock in, performance was great again. Put Double back in and performance is crap again. Like a 5C difference between plates that are not very different

It's not flowrate either....stock is 1.61GPM, Double looks to be 1.5ish

EDIT: Jagged is awful too. Going to put my homemade Single in (made from a Koolance brass blank) and see how that fares.

[Vapor]

And my homemade Single looks really, really, really good.....I'm totally confused.

Bei, any chance you can use the Koolance blanks for a small batch? Tape 'em down to some scrap for the waterjet or something....not sure how'd you align the cutting jet though, hmmm. At this point the only thing consistent is that Koolance-sourced plates are performing well and the others aren't. Maybe I should smooth/bevel the edges of the non-Koolance.

[Vapor]

First mount of Homemade Single is done....mount 2 looks even better


(as more mounts come in, the marker will shrink and lose transparency )

Still gotta figure out what's going on with the Bei Fei plates and potentially retest all of them

[Church]

Just like most here predicted/bet on single. Higher restriction, better temps. Just like with p#1@HF. So i wonder why not try to play that game even furthier .. eg. to try narrow that single and see how it affects performance?

[NaeKuh]

Let me know what you need and I will make!

bei vapor said the diamond did not cover all the channels.

only 2/3rds of the channels were actually being used.

could you shrink the diamond so it overlapped the lanes on the plate.

Just like most here predicted/bet on single. Higher restriction, better temps. Just like with p#1@HF. So i wonder why not try to play that game even furthier .. eg. to try narrow that single and see how it affects performance?

yeah but we dont want to go there again.
higher restriction also limits the capacity on the loop.

And a lot of people are wanting to jumbo size there loops with multi blocks.

[Vapor]

3 mounts into my Homemade Single, it's about a degree ahead of stock with only slightly more restriction.

Puts it about 5-7 degrees ahead of Double and Jagged, so something is awry with the Bei Fei plates, and it's not thickness. Not sure what to make of what's going on.

[NaeKuh]

I think we should get TIM in this thread

[Vapor]

Took my last Koolance Brass Blank and made a second Homemade Single (HS2) with a slightly longer and narrower jet and with a bevel on the inlet. I'll see how the second one fares later tonight when HS1 is done and then hopefully get Tim into the discussion to see if these could be brought to market as a cheap add-on for existing CPU-360 owners and as a CPU-361 or CPU-360X or something.

EDIT: 75.25C average after 5 mounts....for reference, that's ~.7C ahead of the next best block I've tested and ~1C ahead of the stock CPU-360r1.1

[Bei Fei]

And my homemade Single looks really, really, really good.....I'm totally confused.

Bei, any chance you can use the Koolance blanks for a small batch? Tape 'em down to some scrap for the waterjet or something....not sure how'd you align the cutting jet though, hmmm. At this point the only thing consistent is that Koolance-sourced plates are performing well and the others aren't. Maybe I should smooth/bevel the edges of the non-Koolance.

Once I get home I can see how LN2 I have avaliable for laser cutting (yes LN2 is good for other purposes) If we have LN2 avaliable I can laser cut the koolance stainless blanks and get some off to you.

[Vapor]

Awesome!

I still have no clue why the waterjet ones wouldn't be working though. Double and Jagged are, frankly, really similar to Stock and yet they're performing like 5C worse

Anyway, if there is a chance to laser the Koolance blanks, I think sticking to simple would be the most fruitful (based on the 1C gain I've already seen with my ugly dremeling). Probably variations on Single using different jet widths. Stock is 1.9mm wide, maybe step down in .3mm steps....1.6mm, 1.3mm, 1mm, .7mm? I think full length (29mm+) is a must, though, gotta use all the microchannels on the base.

[Martinm210]

Took my last Koolance Brass Blank and made a second Homemade Single (HS2) with a slightly longer and narrower jet and with a bevel on the inlet. I'll see how the second one fares later tonight when HS1 is done and then hopefully get Tim into the discussion to see if these could be brought to market as a cheap add-on for existing CPU-360 owners and as a CPU-361 or CPU-360X or something.

EDIT: 75.25C average after 5 mounts....for reference, that's ~.7C ahead of the next best block I've tested and ~1C ahead of the stock CPU-360r1.1

Wow...didn't quite expect that.....nearly 1C is pretty darn good for a simple nozzle change especially considering how close everything is anymore.

Very nice!

[Martinm210]

Is the stock plate brass also??

hmm...

[Vapor]

Stock is stainless. Bei's are stainless. Retail Koolance blanks are brass (easier to manage with at-home tools). Bei's Koolance blanks are stainless (much better for a laser).

[Martinm210]

Brass has nearly 7X the thermal conductivity as stainless, perhaps touching the fin tops is transferring heat from the plate to the water more efficiently...or maybe just distributing the load over the fins more evenly...may not have as much to do with the nozzle shape...?

[Martinm210]

Was this already tested? I'm behind..

[Vapor]

Doesn't explain why the stainless ones Bei made are so far behind Stock (specifically Double/Jagged, which are pretty similar to stock on jet dimensions). They're not on the graph since I stopped the runs early, but they were 4-5C behind stock, easily.

[Martinm210]

Let me know if you need some brass plates machined. I could order a few and machine them and send them your way for the good of the cause..

[NaeKuh]

Brass has nearly 7X the thermal conductivity as stainless, perhaps touching the fin tops is transferring heat from the plate to the water more efficiently...or maybe just distributing the load over the fins more evenly...may not have as much to do with the nozzle shape...?

then i guess a good way to test this would be to make many many many many many many many lines on a brass plate?

[Martinm210]

Doesn't explain why the stainless ones Bei made are so far behind Stock (specifically Double/Jagged, which are pretty similar to stock on jet dimensions). They're not on the graph since I stopped the runs early, but they were 4-5C behind stock, easily.

Yeah I don't know, have you checked thickness with calipers or a micrometer?

Or maybe the brass is ever so slightly thicker.? Something like .010" may be significant but hard to measure without a mic.

[Vapor]

Stock = .95mm thick
Bei's = .87-.88mm thick
Brass = 1.00mm thick

I don't think it's enough to explain a 4-5C difference, but there is a difference, good call.

I'll take my modified Millipede plate and put it in now....it is .99-1.00mm thick.

Mount 6 of HS1 is completing in a minute

[Martinm210]

Stock = .95mm thick
Bei's = .87-.88mm thick
Brass = 1.00mm thick

I don't think it's enough to explain a 4-5C difference, but there is a difference, good call.

I'll take my modified Millipede plate and put it in now....it is .99-1.00mm thick.

Mount 6 of HS1 is completing in a minute

Yeah not sure on why there would be that much difference, but maybe it's just not getting as much contact pressure between the fin tops and plate. If the nozzle plate is acting like a heat spreader, it would probably depend pretty highly both on the pressure as well as the thermal conductivity of the plate itself. Just like mounting pressure between the block and IHS...pressure may be even more important without TIM to help the transfer...?

Makes me wonder what an optimized copper plate nozzle would do??

Some 19 gauge copper sheet like this (.043"/1.09mm) might be ideal:
http://www.onlinemetals.com/merchant...118&top_cat=87

Maybe try lapping the nozzle plate bottom to a nice polish for better contact...curious if that would help too.

[Vapor]

Updated with final HS1 data:

Final number is 75.25C at 1.51GPM

Good idea with the copper if there is a heatspeader effect, but wouldn't be able to judge that unless we isolated that variable.

Modified Millipede is looking okay, somewhere around Stock I'd guess. Maybe it is a thickness thing

Going to pull it now though and put in HS2.

Stock: jet is 1.9mm wide, 31mm long. Slight bevel at the inlet.
HS1: jet is 1.34mm wide, 26mm long. No bevel at the inlet.
HS2: jet is .67mm wide, 29mm long. Moderate bevel at the inlet.

Microchannel array is 31mm tall.

[Accerian]

I've made and tested custom jet plates on ek supreme (not HF). All of my jate plates was worster from 2C to 5C than orginal one. But mine was litle thicer than orginal one. Thicknes may couse water to go under the seal and disturb water flow.

[Vapor]

I've made and tested custom jet plates on ek supreme (not HF). All of my jate plates was worster from 2C to 5C than orginal one. But mine was litle thicer than orginal one. Thicknes may couse water to go under the seal and disturb water flow.

Yeah, the increase of thickness is whole ideal behind the silicone mods I've posted, but those were 1mm increases, not .1mm. As for the regular Supreme, considering the incredibly shallow microchannels, the plate does need to be completely flush. Supreme HF and CPU-360 have microchannels that are ~3-5x deeper and don't have quite that requirement, but it is still important.

Anyway, four mounts into HS2 and it's about .1c ahead of HS1, but significantly more restrictive (1.51GPM vs. 1.32GPM, stock is 1.61GPM). When it's done I'm going to modify HS2 to have a 1mm by 30mm jet and that will probably be my final test for the foreseeable future (testbed scheduling reasons).

I don't need anyone else to make any midplates either, but thanks for the all the offers