• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Good Idea / Strange Results

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Cathar said:
Copper XS ~= Cascade SS (silver). Storm/G4 pipped the XS across the board, except at ultra low flow (<2LPM).
oh well... just be sure to let me know when the downpayments for the G5's begin :attn:

-Justin :thup:
 
I gave up on motherboard monitor and switched to speedfan a while ago...I seem to remember that MBM isn't made anymore either-dunno that last part.
Give speedfan a try. Nice and stable for me, and doesn't need any corrections to which sensor is which.

I have an unusual design for a block too, but I don't know if it's ever gonna see metal or not. I can't decide. My Swifty 6000-P is doing just fine, but I think my basement truck radiator would make any block look good.

Swell block Hoot, and it's great to see you're still getting your "feet" wet!!! (you betcha that pun's intended :D ).
 
Lookin nice Hoot, its been a while:)
I love their Pin Fin design.
I thought the same exact thing when i saw it:) about identical

A possible problem I see with what you got is the concentration of the jets and the area for the water to escape is quite tiny. I think your loosing the impingment effect with the restriction.

How many hours do you have into that baby?

If you ever wanna drop by and program up a block like that and cnc it, give me a hollar;)

Jon
 
Hi Hoot,
you should love our pin-fin design, it was the forging you used that inspired it

you're well into the mounting, but add flatness
if you have a ever so slightly convex bp, you will play hell getting consistant mountings because it will rock to one side and then with another mount to another
(I'm thinking of your last wb's bp)
look CLOSELY at the grease, it reveals all
non-uniform grease = non-uniform flatness = inconsistant results

if you want to send it to me I'll lap it on the machine and diamond polish
(Hoot only guys, sorry)
 
JFettig said:
Lookin nice Hoot, its been a while:)

I thought the same exact thing when i saw it:) about identical

A possible problem I see with what you got is the concentration of the jets and the area for the water to escape is quite tiny. I think your loosing the impingment effect with the restriction.

How many hours do you have into that baby?

If you ever wanna drop by and program up a block like that and cnc it, give me a hollar;)

Jon

The diameter of the cups was considered in chooing one which would provide escape area greater than or equal to the area of the 1mm ID of the pipe going down into the cup.

Design considerations:
5/64 Inch ID cup area = .0048 Sq.Inch
1/16 Inch OD tubing area = .0031 Sq.Inch
Return Gap area = .0017 Sq.Inch
1mm ID tubing area = .0011 Sq.Inch
.0011 time 167 pipes = .183 Sq.Inch
1/2 Inch ID barb area = .196 Sq.Inch
Bottom line, not a lot or restriction.

I've been learning that part of the problem attaining meaningful metrics is the fact that this MSI board manipulates the raw data from the thermal diode before passing it along to whatever software you use for displaying it. It's been discussed/alleged by many owners in different forums. My gut feeling is that MSI is compressing the results to keep the ignorant blissful. I'm also not fond of the Winbond W83627THF Super I/O chip as a measuring device for thermal diodes. It does okay with thermistors, but that's about all the credit I'll give it. Too much hassle isolating the pads on the 939 socket for the thermal diode to allow porting it to a more accurate measuring IC, not to mention that MSI does not have an SMBUS port header to use. I bore out the Swiftech barbs to .4375 ID, but that yielded no increase in performance. It still delivers an MSI alleged core temperature 20C warmer than the water going through it. In my case, after 24 hours priming it yielded 44C core for 24C water. Idle drops back to 38C. All the blocks I've tried fall within 2C at load of one another on my second performance run and that makes me suspicious of this MSI board. Heck, I can turn off my pump and the MSI only reports a core temperature of 49C while priming. Somethings fishy there... I currently have this new block remounted for the upteenth time where I feel good about it and it'll be another day before I feel good about the AS5 settling in, but it looks like it's going to fall real close to the MCW6002-64 and my Whitewater Clone in performance. I hope to get my hands on a real TDX shortly to play with.

Hoot
 
BillA said:
Hi Hoot,
you should love our pin-fin design, it was the forging you used that inspired it

you're well into the mounting, but add flatness
if you have a ever so slightly convex bp, you will play hell getting consistant mountings because it will rock to one side and then with another mount to another
(I'm thinking of your last wb's bp)
look CLOSELY at the grease, it reveals all
non-uniform grease = non-uniform flatness = inconsistant results

if you want to send it to me I'll lap it on the machine and diamond polish
(Hoot only guys, sorry)

Hi Bill;
Yeah, I know flatness is always an issue, especially with the IHS removed. That little core isn't the best aligning platform for an inverted pyramid which a big honking block is like. With only two mounting holes, you get good control in one axis, but poor control in the other, which just so happens to be the vertical axis in a tower case, which is where block weight works against flat alignment. Since the MCW6002-64 retainer is a known variable, I mounted and unmounted it many times with grease on the little 90nm core and the grease displacement is never the same upon removal, often being thinnest on the lower half of the core, implying a very slight downward tilt. There may be no satisfactory mounting method for A64s with the IHS removed, save mounting the motherboard in a desktop case. When I move on to some other block, I'll take you up on the lapping offer. I still have to get your Innovatek block back to you. Sorry for taking so long.

Hoot
 
Last edited:
Ven0m said:
And what do you think about my 4-point mounting idea?
This way you could use screws at corners or at edges.

I gave it some thought, but have you ever tried tweaking a table with one short leg? ;) Using a digital depth guage is tough in a crowded area around the CPU, but that's what I used to adjust the block as closely as possible. I am still thinking through the whole balancing act.

Hoot
 
Just thinking outloud:

Take some of that brass you used to make the block with, and make two bars about 3/4" wide and as long as the block's width.
Countersink 3 holes, two facing up to match holes in the block, and one facing downward to mount to the board, sandwiching some neoprene to prevent shorts.
Solder in some brass countersink screws and make the posts permanent.

You could even extend the bars, and bend each end upward to mimick a P4's mounting system.

Again, just thinking outloud...I'm sure you've got a dozen or so cures thus far :attn:

BTW, no hilarious Hootstory this time?...I feel cheated somehow :D
 
Oh, almost forgot!

Since you're feeding water into that cap the way you are, the inlet doesn't need to be centered so you can make a new one offset and build a bar/mount bridging over the center with a central torquing screw. That would definately center up the mounting pressure and help level things a bit.
 
Diggrr said:
Oh, almost forgot!

Since you're feeding water into that cap the way you are, the inlet doesn't need to be centered so you can make a new one offset and build a bar/mount bridging over the center with a central torquing screw. That would definately center up the mounting pressure and help level things a bit.

That's exactly the direction I am working on.

FWIW

From the MSI Forums, posted by an informed person:

Everybody should know that temp monitoring / readout of AMD64 is nothing more than a guess of what the ambient tcase (air temp under the lid/HS) is. A64 has a predefined register and the winbond temp chips use this as a base for calculating data to increase / decrease temp based on this register. It's not based on a real temperature sensor , if it were all our problems reading temps had been solved ages ago .

So sum it all up. Temp reporting on a A64 is bogus as hell , and based on a guess of what temp it's inside in the air under the lid/heatsperader .

The temp register is set at the same initial temp value nomatter what kind of cpu cooler ambient temp , case vent you have . So it's not exactly difficult to see that temp recordings of different setups (coolers/cases/ambient temps) is inclusive .

Now that's truly troubling :eh?:

Hoot
 
Hoot
IF your grease is not uniform upon dismounting you're flogging a dead horse
'better' measurement will only reveal greater variations
and yes, the on-board 'measurements' are not even consistent
a can of worms

once you get it flat all this variation will disappear, the 2 bolts will work fine
 
2nd Time Around

My first effort was as much a proof of design, though not absolutely necessary, given Cathars hard work already along those lines. I was as much a proof of technique verifying I could fabricate what I needed to make it work. Now for my second attempt. I got PMs asking me to give more details in the process, so I've included more pictures and steps. This can be done with a Drill press, hack saw, files, dremel tool sand paper and patience. If you're steady-handed, it could even be accomplished with just a good hand drill, using careful layout and measuring along the way, though it would take more time.

First off what did I change from the first version. I decided to use less pipes of greater diameter, for ease of fabrication (completed in one weekend) and to reduce the resistance to flow. Here's a diagram comparing the two concepts:

url]


Here's a stereo microscope image showing a rea-time difference in the area of the two pipe sizes:

diffsize.jpg


To expedite the process and standardize the length of the individual pipes, I determined the length needed to reach near the bottom of each cup, drilled an old piece of stock to that length and used it as a length jig. Just insert the raw brass tubing stock into the hole, cut it off, remove the pipe and repeat:

jig1.jpg


jig2.jpg


jig3.jpg


rawpipes.jpg


When you're done, you need to deburr the pipes. I used a dremel tapered bit in my pin-vise for that:

debur.jpg


I Drilled the top plate holes with the bottom plate attached so that the drill bit (3/32) penetrates the top plate and then continues down to drill the first diameter holes for the cups. That way, the pipes are registered with their respective cups. I then insert all the 3/32 pipes until they bottom out in their respective cups. Then, using a tapered remnant of a broken drill bit, I swage the end of the pipes into the top plate. You can find swaging techniques all around the web:

swaging.jpg


When completed, the top plate looks like this:

pipetop.jpg


And this:

pipebot.jpg


Once all the holes are drilled, I then enlarge the cups to their final 1/8 diameter and depth just slightly deeper than the pipes insert into them. When the three plates are completed, they look like this:

Edit Whoops, the top plate doesn't show the hole for the return flow barb. Sorry...

3parts.jpg


Before sandwiching the three plates together, I lap them as flat as can be accomplished in a reasonable time, thoroughly degrease them, spread a thin, emphasis thin layer of Silicone rubber on them to form a sealing gasket, then bolt them together. The final product looks like this:

final.jpg


As with all my blocks, I then lap the baseplate, again as flat as I have patience to pursue. I use 220, 400 and 600 Wet-or-Dry with soapy water to get this result:

baseface.jpg


The block is in my PC at home, priming and waiting for the AS5 to settle in. From my initial temperatures, it looks like a winner. Once I get the problems with how my motherboard reads the thermal diode resolved, I'll be able to scale my reading and give you some actual values. Relative to a commercial block as a comparison standard, it exceeds the Swiftech MCW6002-64 by 1C idle and 2C load.

The brass tubing for the pipes can be either ordered from smallparts.com or obtained locally at a decent hobby store. In my case, "Hobbytown". If anyone is considering a shot at this, I have an extra 3-pack or 1 foot long 3/32 tubes left over which you are welcome to for the price of postage.

Hoot

Edit, not sure why some of the images didn't show. I rechecked the links...
 
Last edited:
BIG thanks for step-by-step guide, as I think this post can be called a guide.
Great job - with both block and explaination :)
 
Here is an idea that I have seen before, instead of connecting the block to the holes on the board, put a suspended frame above the block, and put a sping between the fram and the block which is the same diameter as the barb and hose coming out of the top. If the barb is centered over the chip, the pressure will be directly in the center.

just my 2cents worth
 
robjen said:
Here is an idea that I have seen before, instead of connecting the block to the holes on the board, put a suspended frame above the block, and put a sping between the fram and the block which is the same diameter as the barb and hose coming out of the top. If the barb is centered over the chip, the pressure will be directly in the center.

just my 2cents worth

That's also a good idea. As it turned out, with the second block, using the springs and screws that came with the MCW6002-64, I was able to secure a good, tight, flat bond. Rocking the block moderately does not make the temperature jump around. Even though the screws are intended for the appropriate force for bearing down on the IHS, the exposed core does not seem to mind it. Since my case does not travel, it is an acceptable risk.

When I was using the MCW6002-64, I actually put a nitrile o-ring under the plate, between it and the block, centered on the intake barb and the centric force stabilized that block very well.

Hoot
 
Back