Lapping waterblocks?

[Top Nurse]

Hey thanks for replying as I figured you might know about this as I read somewhere you had a similar problem in your prototypes, but I figured that would be proprietary info so I didn't ask. Well the dimensions given above were for the Cuplex Pro that also had a concave surface though it was made out of copper. Here is the link to the Cuplex Pro. However, I just disassembled my Silver XT and with my .001" dial caliper it appears to be within .001 of being square and in thickness around the edges. The thickness of the base plate appears to be about .040 (minus the .002 concavity so say about .038) at the thinnest part where they plunged an end mill in. There is an almost identical gap on my XT as the link shown above and it is .002 right above the CPU.

Interestingly enough all my copper Twinplex's are very flat though and they are about .050 thick at the bottom of the plate.

Do you think I will be able to salvage this block by hand lapping it on a glass plate? Or do you think I should send it back to Germany? I could never get out of their manufacturers forum what specs of flatness were within tolerance.

BTW, how long does it take to get a G5? And will it work well with a 6mm or 8mm ID tube size and my Aquastream or DDC?

 

[Top Nurse]

I think it's prudent, in this forum at least, to mention that glass is a superCOOLED liquid
A sheet of glass that has been sitting in a window pane for the past 20 years is sure to have slouched a bit and will not be very flat. The mirror glass is a good idea - mirrors are likely to suffer the same fate after time, but at least you will be able to tell somewhat more easily.

So what about say a brand new piece of glass that is about 3/8 - 1/2" thick cut to order and a size of say 6" x 18"?

 

[Cathar]

Do you think I will be able to salvage this block by hand lapping it on a glass plate? Or do you think I should send it back to Germany? I could never get out of their manufacturers forum what specs of flatness were within tolerance.

If it's only out by 0.002", or ~50 microns, then I would lap with the block assembled.

A glass plate is fine so long as it, too, is on a fairly flat surface. I would start with 400-grit given your present level of non-flatness, then 600, 800, 1200. Use low-smell kerosene as a cutting lubricant for the paper. Before you start with each grit, scribble all over the base with a black texta/permanent marker so that way you can see when you're finished with a grit - which will be when ALL of the marks are gone.

Do a final lap with 2000-grit, but don't do the permanent marker trick for that one as it can clog the paper. You'll know when it's done.

Do it at night-time, or basically when it is around 15C ambient, if possible. This helps to prevent warpage due to heat flexing.

Most importantly, go slow. Real slow. Take about 4-5 full seconds to do a lap. Listen to it. Hear the grinding/cutting noise it makes. I find that going slow and listening for the particular "note" of the sound is where I get my most effective lapping results. Going fast will heat the block up, causing it to warp again when it cools down.

Don't push down heavily on the block. Don't push down with more effort than it takes to move the block along the paper.

Can take a good hour or two to really lap a base-plate properly, but if you take the time, it'll be flatter than anything that most machines can manage.

BTW, how long does it take to get a G5? And will it work well with a 6mm or 8mm ID tube size and my Aquastream or DDC?

Next round of G5 ordering is about 2 months away, and it takes about 2 months after that before the blocks would head your way. May be a little quicker than that this time around as my CNC machinists are getting more machines this month.

Regarding its performance, Procooling's results show that at 1LPM and above that nothing matches the Storm/G5. Assuming that the Cuplex XT has a similar pressure drop to the NexXxoS XP, you will achieve slightly higher flow rates with the G5 as it is less restrictive.

So a good rule for you to do, I guess, is to look at your present flow rates. If they're 1LPM or more, then nothing comes close.

 

[Top Nurse]

Regarding its performance, Procooling's results show that at 1LPM and above that nothing matches the Storm/G5. Assuming that the Cuplex XT has a similar pressure drop to the NexXxoS XP, you will achieve slightly higher flow rates with the G5 as it is less restrictive.

Thanks for the lapping info

I don't understand how it could be less restrictive if the nozzle holes are only the size of about twice a human hair in thickness.

 

[Top Nurse]

The ultimate question though is just how much effect does this non-flatness have on performance. Thermal pastes are designed to work happily with 50 micron gaps (0.002"), and up to 100microns in a pinch (0.004"). Also remember that it's the non-flatness that's directly above the CPU die that's important. While edge-to-edge non-flatness might be 50 microns, the part that's touching the CPU may only be bowed by 10-20 microns edge to edge, which is something that a thermal paste can easily handle.

In my experience, slight non-flatness doesn't affect temperatures as much as it affects overclocks. While the overall CPU temperature may not vary greatly (0.5-1.0C), it's the CPU's overclock that tends to suffer the most, especially moreso with bare-die CPU's.

Well from what I saw when I dry mounted the block it was only touching the corners of the Winchester 3500+ CPU heat spreader. So I figure it was .002 directly over the CPU and tapering off to zero to .001 depending on what side you measured it on. I guess I was just expecting something a little better for twice the price of a standard XT. So do you think that .002 concavity is a standard tolerance for most manufacturer's? Or is this something that slithered out the back door unbeknownst to AC?

 

[Cathar]

So do you think that .002 concavity is a standard tolerance for most manufacturer's? Or is this something that slithered out the back door unbeknownst to AC?

I would say that is about standard for manufacturers that make a minimal attempt to achieve flatness (basic machine-finish level flatness).

Ensuring 0.0003-0.0005" flatness requires quite strict measures which many manufacturers do not take.

I don't understand how it could be less restrictive if the nozzle holes are only the size of about twice a human hair in thickness.

Oh, the jets themselves are quite a deal larger than twice human hair. I said the finest machined structure in the block was about that size, not the jets themselves.

A lot of the XP/XT's restriction comes from the water trying to force its way out of the pin maze, as well as the restriction of the jets. With the Storm design the only real restriction is the jets themselves. Once the water leaves the jets it is largely free to leave the block relatively unhindred after it does its impingement deal.

 

[Top Nurse]

I would say that is about standard for manufacturers that make a minimal attempt to achieve flatness (basic machine-finish level flatness). Ensuring 0.0003-0.0005" flatness requires quite strict measures which many manufacturers do not take.

I hope you patented your design.

Well I suppose then that this is an abberation of Aqua Computer then. But any block I buy from now on gets the flat measuring tools out

 

[DFI Daishi]

is there a special reason that you are reccomending kerosene as opposed to straight water?

i image that it ensures cooler temps, but i would think that generously poured water would do better at carrying the sanding debris away from the area of contact, and so help more with avoiding scuffing and uneven wear.

unless the kerosene is not evaporating as fast as i think it would.....oh well, i guess that's why i'm asking.

 

[Cathar]

is there a special reason that you are reccomending kerosene as opposed to straight water?

I tend to find that it lubricates better as you're sanding. Some people use water and say it's okay. Having used a variety of wetting/cutting agents, I swear by (low smell) kerosene and that's my choice.