if it'd help, i'm sure we could promise a lot of work.... ;)

how much is a regular cascade?

I didnt know polycarb was that good!

It's just expensive to make.

Consider the price of a DTek WhiteWater, and compare it to the price of a Cascade: :o

It's the machine time that's the issue. There's about 1 hr of CNC machine time in each block, and a further 30 minutes of manual labor.

Hi Cathar,

I posted a e-mail to you a while back regarding the SS Cascade. Any progress toward having enough people for another batch to be worth your time?

Thanks man, You do great work!

Woohoo!!

Just off the phone to the machinists. They successfully cut the middle plate in polycarbonate on the CNC. They warned that it is a little rough due to the cutter being used being somewhat blunt, but overall the operation was a success. Apparantly a couple of the tubes have split on the edges, but this was more due to a small programming fault that they're now aware of and will fix. Such is the way with first-run prototypes.

Anyway they'll be machining up the copper plate tomorrow or the day after and by this weekend I'll have the first working XXX prototype (in copper) to play with.

That's definitely good news!

Looking forward to pics.

nice, i can't wait to see it.

More looking forward to performance myself. I already predict that at low flow rates that it'll put the regular Cascade to shame, since that was one of my goals when designing it. Let's say I've been influenced too much by people putting heavy weight in foreign tests run at <2LPM. The XXX should continue to offer the same rough rate of performance gain as flow rate is increased as the regular Cascade. Perhaps more, perhaps less. I don't know. Depends on which variable is the dominant factor in reality as opposed to theory. Maybe hoping for a 1C gain over the Cascade.

Really though, and what was the focus of the block's design, was to pay more careful attention to smoothing out hot-spots. This is something that I can't really measure except in terms of overclock stability. An improved overclock will be the true measure of success for the XXX for me, and not so much the actual temperatures seen.

So more focus on jet/cup geometry interaction, inter-cup geometry, and matching that off against the base-plate thickness. Am very excited.

measured temps are NOT the measure, chase the (CPU to CPU variable) overclock, better resolution
yes, FAR more work but more significant results
(glad its Cathar's time and not mine)

? how can you 'SMOOTH' a hot spot whose location is not known ?

Will have 3 separate CPU's to play with. Yes - a lot of time.

Regarding "smoothing", apart from location - also the size?

Really what I'm attempting here is a prediction of the lateral heat-spread though the bp given the assumption that the cooling is not going to be equal across the base of a cup/wall "cell". So assuming an infinitessimal point - it's a case of figuring out what amount of material needs to be between the CPU and the convection area to ensure that effectively an entire cooling "cell" is engaged for a point source of heat, but also not putting so much material in the way that conduction becomes a major cost.

On the original Cascade I could only say that really about 1/3rd of any particular cooling cell was acting on any given point of heat. In fact the size of the cooling cells made it somewhat hard to set the bp thickness at an acceptable level without the cost of copper conduction becoming a major concern.

In essense the 25% shrink (in each dimension) of the XXX was absolutely necessary to provide a better tradeoff between lateral heat spread and conduction. Ideally I'd like the XXX to undergo a further 20% shrink to get it all "perfect", but that's a battle for another day.

Overall in the XXX the net conductive thermal resistance hasn't dramatically changed from the regular Cascade, just better balanced, or "smoother", as I put it. Am hoping for tweaks to the jet geometry/velocity to buy some overall performance gain, but again, not exactly holding out really high hopes.

Of course the savvy of you will have worked out by now that the XXX has a thicker base-plate than the Cascade, and that does directly translate into better low-flow performance. Incidentally the SS has a 40% thicker bp than the regular Cascade. The only real doubt in my mind is how the thicker bp affects higher-flow performance. Will the copper conduction cost become the dominant factor within the limits of the flow rates that I'm capable of enacting on the block? This is the question that I have really yet to wholly figure out.

That's great Cathar! I hope the rest of the machining goes well.

As a physicist, I too am interested in the temperature results regarding the change in baseplate thickness and other factors you have mentioned. Testing should have some interesting results I'm sure.

-OMP

Cathar
you are aware of EK's testing, some of which yielded a higher OC with a higher CPU temp ?

you realize that with a resistor network you can model your bp ?
then you can back-calculate to get a fit for the different bps
but then you will need to address the thermal lag (capacitance) which affects the OC
gets too hairy for me when it is then necessary to characterize the magnitude and duration of the thermal transients

Hmmm....the next step in the Cascades evolution? Where to go from here? This stuff supposedly boasts a rating of up to 1200 W/m*K. If it's all they say it is I imagine lateral heat spread would improve quite a bit. Not easily machinable tho'. Here is a responce from them regarding a few questions I had.

"Knoop Hardness is around 60 GPa, depending on DiaCu-II formula used. Lower diamond volume percentage DiaCu-II versions can be machined with wire-EDM, however, normally DiaCu-II must be laser-cut. Typically we provide machining and metallization services for our customers. DiaCu-II is not permeable to water; as far as creating a water seal, our current samples are shipped with a minimum 0.25mm layer of O-Free Cu on both flat sides of the DiaCu-II disks – so, assume that a seal would be created against the desired finish of the copper or any external plating/sputtering on top of the copper. Factory finish is fine-lathed OFHC copper surface, but can be mirror polished if desired (Cu is fairly easy to lap for most any customer). 90mm slugs will be available by Q2’04, depending on demand."

I did not get a quote so pricing is up in the air. It'd be fun to play with though. :)

jlrii
keep digging, better things out there
and the pricing will cool your ardor

Late to the game, but wanted to toss out some comments about those Cathar made regarding 304/316 stainless. We use these materials frequently for items with variable water contact, ranging from stagnant to turbulent and varying between submerged or not. Generally speaking, in my experience rust is only a problem if the steel has been contaminated, generally by manufacturing processes that handle both stainless and plain carbon steels. A good passivation process normally takes care of such contamination.

That said some of the water we deal with does have bad actors that will attack the 18-8 family of stainless. In many of these cases this is addressed through either going to a duplex stainless ala 2205 or a specific paint procedure.

I would not expect the fluid contents and temperatures of a typical water cooling system to cause any trouble for 304/316.

I'm just wondering what experience or data has led to your conclusion regarding the rusting of 304 and 316.

Yeah from what I've seen CVD=$$$. Their PDF brochure says,"Costs: Competitive with existing high-end thermal
materials; less than 1/10 the cost of traditional CVD
diamond materials"

The stuff I've seen b4 is around $1800 for a single slug so if their advert is true that puts it around $2-300 per 50mm slug (Minimum "sample" order of 10 pcs BTW). Nice to think about but too much money for DIY. I imagine you could use a fairly small piece tho', using impingement there isn't really any need to cover much more area than the die takes up.

EK? Am aware of my own testing though where I found the same thing, and promptly drew the ire of a certain forum admin leading to my banning from LiquidNinja's for daring to suggest such.

Basic summary of my prior testing which led me to this point:

Test 1) Cascade => Overclock stability quite dependent on flow-rate despite relatively small changes (<2C) in overall CPU temperature. By altering the water temperature I was able to ascertain that the overclock stability issue had a far wider effect than the overall CPU temperature. i.e. The high-flow scenarios would remain stable at 5C higher indicated CPU temperatures than the low-flow scenarios. Hypothesis: Uneven cooling effect across each cell resulting in hotter locations for the lower flow rates affecting overclock

Test 2) Thicker bp Cascade (essentially a copper Cascade SS without jet mods) => Overall CPU temperatures slightly higher and overclock stability was about the same at higher flow rates. Better overclock stability and better temperatures at lower flow rates than regular Cascade. Hypothesis: thicker bp provides a better "smoothing" effect

Test 3) Cascade SS => thicker bp than regular Cascade, jet modifications. Markedly improved overclocks in all scenarios (low/high flow rates), including holding higher overclocks at higher CPU temperatures. Improved overall temperatures. Undoubtedly some of the benefit seen was due to the silver used facilitating greater lateral thermal spread. Some gain also due to jet mods.

Basically all that is telling me that despite the original Cascade giving pretty good temperatures across the board, the actual temperatures weren't really telling the full story. Mind you, the original Cascade was giving me the best overclocks of all the blocks I tested it against, so I guess that's why I rested on it for a while.

Bill, I'm sure that a test setup can be made to measure all this. Electronics aren't my strong point though as I'm a relative noob in that area. Sure, I can find my way around a multimeter, but proper part selection and setup and control is something I'd need assistance with.

Hi Dave,

My experience with 316 is rather limited, but based upon the Laing D4 pumps I ordered in. Even after sitting overnight and damp after a test, I opened one pump the following morning to discover rust spots had formed.

Now I guess I should point out that where this happened was where the metal had been "marked" by various inspection chops (red stamped writing), and it may very well be that this was enough of a "contamination" to result in the rusting I saw.

In any event that immediately scared me off 316. My other concern arose from the galvanic differences between stainless and copper/silver. When in an passive setup, which I understand to mean running water where there is no fixed ionisation loop occurring, galvanic corrosion isn't really an issue, but for an active setup where the two metals are in constant ionised contact with each other then the PD appears to be enough to sustain a slow reaction. Not as dramatic as aluminium/copper, but enough that over time the 316 will get eaten away.

Now I'm no expert on this, and this is what I've managed to deduce from Googling around, so I will happily defer to someone with more experience who can definitively answer yes/no to either of those two points.

banned ? you mean like JoeK did to me here ? lol, their loss
EK are the initials of OPPainter of xtremesystems.org, bit of an OCer

did not mean to suggest a hardware 'solution' using resistors, I was talking about software simulation as thermal and electrical resistances are calculated the same way
examples here, all nodes are connected via resistors

Actually I think I got banned right after I told a certain admin what I thought of their practise of reading and using the content of Private Messages between forum members in public posts. Some comment about me being antagonistic towards the site for what the site says is common practise as far as they are concerned. I don't know what came over me - getting upset about something that is a federal offense here in Australia.

The issue is not so much with breaking down the simulation like you describe, indeed that's pretty much what I've done. The issue is confidence that I've done it right. At some stage practical results are going to yield better data than theory, and I'm pretty much at that point.

Does your simulation predict better cooling directly under a "cup"?
My crude models using heatspreading and convection coefficients give little enlightenment.
Think that a detailed analysis(including convection profile) by programs like TAS may be necessary to answer even that simple question.

I have TAS and it won't do this as it has no CFD solver
one has to enter a value for h, the convection coefficient
this guts the calc as h is the great unknown (for me at least)

h is also the great unkown to me.
I would probably try a Flomerics calculated h for an average value within a cup. Rather unsatisfactory and not a profile but ...................