Spot the difference - Maze 4 GPU

[pdf27]

Having been rather hacked off for a while with how badly the standard Danger-Den barbs seal with 1/2" ID tubing, I've gone and done something about it.
The unmodified one is for comparison - I'll do the same to that next weekend when I get some time.
Holes are drilled & tapped out to 3/8" BSP, barbs are 5/8" OD.

[Marci]

Hmmm.... DD brought out the PerfectSeal fitting to combat this issue, but everyone moaned that the thicker lip hindered smooth flowpath and created turbulence at block entry, so they made em available with HiFlow again... if ordering direct from DD you get a dropdown box with choice of Hiflow or PerfectSeal.

Personally I'd've just ordered a pair of PerfectSeal fittings first to see if they improved the situation... they're available separately from a number of UK Resellers (inc. me)

Only limiting factor is which Maze4 u have... (this is more for everyone else's reference than directed at you btw...) if u have the older 9/16-18 UNF threaded block, PerfectSeal aren't available for it. If it's the newer G1/4 threaded block then they are. If it's the REALLY old 1/4" NPT threaded ali-topped block, then there's no option but to drill and rethread, but most just replace top with newer delrin one as they're also available separately, then use the fitting of choice from the entire G1/4 range.

[pdf27]

Yeah, both the blocks I've got are 9/16" unfortunately so I don't really have any option - looked around for adaptors, but it was either replace the tops or drill and tap them out myself - in which case I could use fittings I'm used to and have confidence in.
PerfectSeal are probably going to be less of an issue with 3/8" than 1/2" though simply because of the thread size they use. 1/4" BSP needs an 11.5mm hole for the tap, 3/8" needs a 15mm hole. Throw in an allowance for thread depth, barb thickness and the like and 3/8" BSP gets very close to 1/2" (12.5mm). G1/4" gets down to 8 or 9mm in comparison - that's a substantially smaller area. For 3/8" tubing that's probably a good match for the tube ID, but for 1/2" you've got an unnecessary section change as well as the restriction caused by the thread.

The performance differences are probably too small to detect in real life, but I reckon they're probably there. In any case, the major benefit is that they're massively easier to seal and clamp than the Hi-Flow fittings. That and the geek factor

[Marci]

Throw in an allowance for thread depth, barb thickness and the like and 3/8" BSP gets very close to 1/2" (12.5mm)

Yup - hence we use 3/8" on ThermoChill rads...

[Cathar]

The alternative would've been to just use 7/16" ID tubing on the hi-flow barbs.

[pdf27]

Originally Posted by Marci

Yup - hence we use 3/8" on ThermoChill rads...

Indeed (and my PA.160 is excellent by the way ). Considering that they have the room, I've no idea why on earth they've gone to 1/4" when 3/8" gives them a heck of a lot more flexibility in terms of barb choice without the pressure drop penalty of a smaller aperture.

Originally Posted by Cathar

The alternative would've been to just use 7/16" ID tubing on the hi-flow barbs.

I did think about that, but I didn't realise how bad the hi-flow barbs were for sealing (particularly when the tube is coming in at an angle, as for graphics cards) until rather too late. Besides, ignoring any effort involved in retapping the holes the trade-off to be made is between weight/volume and performance. As weight is a non-issue (the case also has a hell of a lot of sound damping material in it) and the case (Silverstone TJ-06) isn't volume limited, 1/2" ID gives me a bit of extra performance without a "cost" that bothers me. As in my case I'm exchanging performance for dBA, moving to 1/2" is close to getting something for nothing.
For most people it makes sense, but when you're going to the trouble of running AC into your case just for the fans it isn't worth it...

[Cathar]

Okay, just so long as you're aware that for flow rates <10lpm, and tubing lengths <3m, there are no discernable differences in flow rates between 7/16" and 1/2" ID tubing, so long as the 7/16" ID tubing is being stretched over 1/2" OD barbs.

[SiGfever]

7/16" is sooooooo much easier to fit and route in a case and when you put it on 1/2" barbs you have a great seal. The only place that I have 1/2" tubing in my case now is from the Micro-Res to the pump. The DDC-2TPMP was pulling pretty hard on the 7/16" Tygon so I gave it a little more supply volume. I just wish that I had checked the pump's amp draw before and after to see if there was any difference.

[pdf27]

Originally Posted by Cathar

Okay, just so long as you're aware that for flow rates <10lpm, and tubing lengths <3m, there are no discernable differences in flow rates between 7/16" and 1/2" ID tubing, so long as the 7/16" ID tubing is being stretched over 1/2" OD barbs.

Out of curiosity, is that 1/2" tubing over 1/2" barbs, or 1/2" tubing over 5/8" barbs you're comparing it to? I suspect the change in section from not stretching it oversized barbs makes quite a big difference.
Fair enough on the diminishing returns though - otherwise if I followed my train of thought to the logical conclusion I'd have 2" tubing throughout my case

[Cathar]

Comparing to 1/2" ID tubing over 1/2" OD barbs.

[Cathar]

Originally Posted by pdf27

Fair enough on the diminishing returns though - otherwise if I followed my train of thought to the logical conclusion I'd have 2" tubing throughout my case

Call me crazy, or perhaps even European, but I'd like to work on creating an effective cooling system around this sort of tubing.

[ricecrispi]

That tubing is 8mm and makes 3/8" look like a gardenhose. 7/16" has clear advantages over 3/8" and 1/2", a great alternative between the two tubings and that is why people use it.
I'm loop a system from 1/2" to 3/8" and it's so easy to loop. Makes me realize there is little reason to create a system that effectively centers on 5/16" tubing. Unless for the pure challenge of it.

[Cathar]

5' (1.5m) of 7/16" (8mm) ID tubing presents around 60cm of pressure drop at 4LPM, compared to 3/8" which has 25cm of PD.

Given that 3/8" barbs really have a ~7mm ID, this creates a boundary transition PD of 9cm per 8mm->7mm ID interface, and 15cm per 3/8" -> 7mm ID interface.

i.e. In a 5' tubing lengthed system, with a pair of GPU blocks, a CPU block, a pump, and a small res, the total tubing and barb interface PD between 8mm and 3/8" ID tubing totally cancels each other out.

What this is saying is that if you're using an unmodified Laing DDC, and your "typical" w/c loop, then you may as well be running 8mm ID tubing rather than 3/8" ID tubing. Much of the 3/8" ID tubing with 1/2" OD has issues with turning corners easily due to the thin wall per orifice ratio, hence the sale of supportive kink resistave items. 8mm ID tubing would be more kink resistant.

i.e. it's not THAT much of a challenge. Use blocks and radiator which are both well into their performance plateau's by 4LPM* flow rates, and is there any need at all for a bigger pump and larger tubing?

(* On the assumption that said blocks and radiator don't total more than 1.5m of PD - barb interfaces excluded since I already factored that in - in order to facilitate 4LPM flow rates with a MCP350 pump).

This is all getting back to designing small, quiet and efficient systems that sacrifice very minimal performance levels to large pump / large tubing solutions, while being easy to route in small cases.

[migueld]

3/8 rocks, it's easy to work with and on up to pro cooling.

I still wonder why the American side considers 1/2 as crucial and bashes 1/4 (or even 3/8) when the European side lives with 1/4 as the best? Who's right? Even if 1/2 is better, does its marginal performance gain outweigh the ease of use loss? I say heck no.

[Cathar]

I think you'll find that even most of the "top-end" European systems are using anything between 8mm to 10mm ID tubing (3/8" is 9.6mm).

The 1/2" ID phenomenon occurred at a time when the water-blocks and radiators of the day (5 years back), including the blocks and radiators coming out of Europe at the time too, almost all showed moderate to strong performance benefits even up to 10LPM, and some even beyond.

Five years ago many in the USA, and even I for that matter, arrived at the conclusion that peak performances were to be attained in the >8LPM range no matter who built the item - whether Europeans or Americans (or Australians). The waterblocks which all had nothing to gain above 4LPM were eclipsed noticably by the high-flow-rate happy items, and this further strengthened the drive towards high flow rate, large tubing solutions.

You have to remember, this is all happening back in the day of stupendous heat dumping overlocked Athlon T'Bird's and first-gen Willamette P4 CPU's, back when an extra 50-150MHz was actually noticable, and even small performance gains at the waterblock through increased flow rates resulted in significant drops in CPU temperature, which then facilitated higher overclocks in a CPU performance hungry world.

Fast forward 5 years, and there's a few blocks and radiators out on the market now which, through correct engineering and a better understanding of the nature of fluid dynamics driving design, which offer close to the best of both worlds. Close to peak performance at moderate flow rates. Sure, higher flow rates will still perform/cool better, but is anyone sitting at the keyboard going to notice it? The answer is no.

European and American designs have been quietly converging over the last 2-3 years, but from opposite ends of the spectrum. Many factors have resulted in this, not the least of which is greater understanding of block design at both sides, and a CPU market that is mellowing. Yes, there have been a few American manufacturers, notably Swiftech, who have always stood by the 3/8" approach while still catering for the 1/2" approach, just as there have been some European systems using 8mm tubing while most were still using 4/5/6mm.

Seems to me that somewhere between 8-11mm ID (5/16" - 7/16" ID) lies the "ideal" modern tubing ID. In that range, pick what you want as best suits your needs.

There's no need for pointless trans-Atlantic bashing.

[Seven]

Cathar hit it right on the head.

At this point, watercooling is so advanced and has developed so much (remember the days before DC pumps?) that the debate between european and american systems is really a moot point. I will tell you right now, cosmetically, the european systems look much more spectacular than any american system, in addition to being smaller and seemingly easier to implement. While yes, high flow rates are better for temps, as Cathar says, it's not even noticible at this point.

I will say this though, as dual and quad cores become more common, you will see more of this issue arising.

[pdf27]

Originally Posted by Cathar

This is all getting back to designing small, quiet and efficient systems that sacrifice very minimal performance levels to large pump / large tubing solutions, while being easy to route in small cases.

Looks like one of the big areas for improvement in most systems will be the transition from hose to barb. When (potentially) a mere two barbs will have more pressure drop than your tubing, you're looking in the wrong place to save pressure drop.
Thinking about it, a system with badly selected barbs might be losing as much as 2m worth of pressure drop due to the barbs alone. For well designed barbs, you can presumably eliminate the pressure drop from the hose-barb interface almost entirely. The barb-block interface is more troublesome, particularly as the inside of blocks is non-standard.

At a guess, the ideal barb would look something like a standard riffled connector with the hollow thin walled cone from the Danger-Den Hi-Flow barbs on the end. The objective would be to ensure that when passing through the water experiences no change in section, while there is a large length of tubing stretched tightly over the barb (i.e. the barb OD is substantially greater than the tubing ID) and the barb is long enough to clamp easily.
I suspect this may be one of those cases where subtle system design choices that most people don't even think about come out to bite you without you ever realising it. It may turn out that with the amount of optimisation and investment that has gone into some areas (tubing size, block design, etc.) there is far more to gain by looking at other areas, at far less cost.

[Cathar]

Originally Posted by Seven

I will say this though, as dual and quad cores become more common, you will see more of this issue arising.

Not while CPU makers continue to sell CPU's with non-removable non-flat IHS's.

IHS's have also been responsible for levelling the playing field dramatically. When every single water-block is effectively forced into a "thick base-plate scenario" where the parimary heat paths can occur anywhere from the middle of the IHS, roughly even, or more at the edges, depending on IHS flatness which is being primarily affected by what's going on at the edge of the IHS, then water-block design nowadays is a bit like building a ship in a bottle, it provides a very one-eyed perspective on how the cooling can be best achieved.

I intensely dislike IHS's for their non-uniformity, despite the best intentions of the CPU makers, and I intensely dislike them because they severely limit water-block design freedom. It really is tantamount to having sex with a 1/16" thick rubber condom. Sure, the job is getting done, but all sensitivity, finesse, and control is effectively dead. Further, with reported CPU temperatures being sensed at locations other than where it's important, it's bloody hard to tell whether or not your indications of success are truthful, or merely faked in order make you feel good.

[SiGfever]

Is it just my imagination or do people seem to be having more trouble with the Conroe's IHS being flat then they did with other makes and models?

[Seven]

Originally Posted by Cathar

It really is tantamount to having sex with a 1/16" thick rubber condom. Sure, the job is getting done, but all sensitivity, finesse, and control is effectively dead.

Simply brilliant.

Honestly it's easier for CPU manufacturers and resellers to have IHSes on chips nowadays. It's not your old 1700+ Athalon with a small aluminum heatsink. Most noobs (manufacturers too) would crush their core, and boom, there goes a $400 chip and none of the blame falls on the manufacturer or reseller.

I feel as if watercooling is getting to a point where we cannot improve upon it much more. It's like trying to set the world record for a 50m sprint among hundreds of other world class runners. However, we always repeat the general cycle in CPU technology. Underdeveloped hot-running chip>More efficient chip>Prime chip for overclocking and enthusiast use>Pushing the same technology to it's limits>Revising the technology and repeating the cycle. Remember the transition from the late P3 days to the Early P4s? The first P4s couldn't even hold a candle to the later chips that came down the line. I see the same thing happening as more and more cores are stuffed onto a single CPU.

However, I feel as if there are other areas to be improved on, such as GPU cooling and pumping technology. It's really a search for a pump with good pressure, high flow rates and something that doesn't dump too much heat into the water.

[Cathar]

Originally Posted by Seven

It's really a search for a pump with good pressure, high flow rates and something that doesn't dump too much heat into the water.

Laing DDC/DDC+, with a modified top.

If Laing provided a straight-in top, rather than the silly right-angled bend, in one stroke they'd near kill off the after-market top scene and people would have a pretty hard time justifying using anything other than 3/8" ID tubing.

That right angle at the inlet is about the only thing standing in the way of me calling the DDC/DDC+ the "ideal PC water-cooling pump". Instead, an aftermarket top is required today to achieve that status (for me).

[Azazel]

Originally Posted by Cathar

Laing DDC/DDC+, with a modified top.

If Laing provided a straight-in top, rather than the silly right-angled bend, in one stroke they'd near kill off the after-market top scene and people would have a pretty hard time justifying using anything other than 3/8" ID tubing.

That right angle at the inlet is about the only thing standing in the way of me calling the DDC/DDC+ the "ideal PC water-cooling pump". Instead, an aftermarket top is required today to achieve that status (for me).

From a pressure/flow POV, I agree completely. From a noise perspective, no.

There are still no pumps out there (DDC with modded top included) that are quiet enough to compete with a well-designed aircooling rig. Even with graphics cards, I can get good performance out of a modded vf900 with an 80mm nexus fan running at 7v. At these speeds, it keeps my 7900GT (volt-modded to 1.4v) running happily at 700/780 core/clock. With one of any number of great heatsinks (the Scythe Ninja comes to mind, as always) on the CPU, you can have an aircooled computer running at 90% of the maximum clockspeeds you can get from watercooling and do so with less noise (in almost all cases. exceptions always exist). The death of the megahertz race was the beginning of the end for enthusiast watercooling. Until somebody comes out with a pump noticably quieter than a DDC or an eheim 1048, I've hung up my raincoat.

ps. Yes, I saw the '(for me)' caveat. No, it wasn't going to stop my SPCR evangelical rant.

[ricecrispi]

Yeah a "well designed" aircooling setup will be more silent than WCing but how well design is the WCing setup.

I tried really hard to isolate my DDC. Diffrent foams, multiple layers, zip tied so it doesn't rest on the floor, and put as much effort as I would silencing an air-cooled
setup.

I guess since Wcing cost more money and extra work people don't feel invest as much time in their aircooled rig when it is so much easier.

[Cathar]

Azazel, you'll get no argument from me on the "90% over-clock is good enough" attitude. I completely agree.

Still, the faint noise of a DDC or an Eheim 1048 doesn't bother me, since disk drives either make about the same noise as either, so I'm happy to live with the faint noise of a pump given that I have to live with the noise of the disk drives anyway.

[Azazel]

Originally Posted by Cathar

Azazel, you'll get no argument from me on the "90% over-clock is good enough" attitude. I completely agree.

Still, the faint noise of a DDC or an Eheim 1048 doesn't bother me, since disk drives either make about the same noise as either, so I'm happy to live with the faint noise of a pump given that I have to live with the noise of the disk drives anyway.

Fair enough, but I'm running an enclosed laptop drive, for all intents and purposes, silent.

I tried really hard to isolate the pump noise (in one of these, with foam and a gelpack) and even tried undervolting. I really wanted to get wet again after a year of aircooling. It was a sad day realising that I'd just spent NZ$400 to prove that I could never go back to the constant light humming of a waterpump.

As always, YMMV, but my point is (was) that silence is no longer the realm of watercooling, even when overclocked. An aircooled rig unfortunately (or fortunately, depending on how you look at things) is now cheaper, easier and quieter than a water system for the vast majority of cases, while only running slightly hotter. Of course all of this may change with the next gen of graphics cards, but I really can't see anyone who values quiet computing buying a 300w (if the reports are correct) graphics card.