martinsliquidlab.org

EXPANDABILITY/MODIFICATION

One of the big selling points of the H220 over your usual AIO kit’s is that it is intended for expansion. While ultimately there have been many demonstrations and user reports now to show the kit is indeed expandable and I’m the type that likes to measure things and understand how they work at a finer detail than just degrees.   That’s where the flow meter and digital manometer come in to test restriciton and pumping power and hopefully shed some light on the hydraulics.   DIY water cooling parts are generally designed for slightly higher flow rates than AIO units and typically seem to have lower restriction. That is good in terms of retaining higher flow rates for easy bleeding and filling, but it also means their thermal performance starts to drop off at lower flow rates. While we don’t have a lot of hard data to understand how really low flow rates impact performance for GPU blocks, we can look at that for CPU blocks.

Using Stren’s recent CPU block flow rate performance data here.

You can see that as you would expect, block performance does increase as flow rate increases. By 1.5 GPM flow performance increases become very small, usually less than a degree, but flow rates from 1GPM down to .5GPM do start adding up to 3 degrees or so. Using that I would say anything much lower than about .5GPM is going to really start suffering at performing well even with the relatively micro designs in CPU blocks. GPU blocks are generally not quite as micro in their fin structure, so I would assume that drop off in performance is even more pronounced. With that I will stick with the .5GPM minimum flow rate, 1GPM+ preferred.

Bleeding Performance

In addition to thermal performance dropping off as flow rates dip down low, even more important is how well a system self purges or bleeds itself. The 1GPM rule used in past DIY planning efforts was predominantly there to help ensure the system will self purge air bubbles from the tubes and parts, etc. This can be mitigated quite a bit by removing the system from the computer when filling and bleeding as it gives you an opportunity to shake parts and help encourage the trapped air around the loop, but it can take a considerable amount of work.  1GPM is however fairly conservative I think.  I did a few very simple tests with 1/2″ ID tubing coiled vertically to see at what point the flow rate was high enough to push the air pocket down and through the tubing.  At ~0.3GPM the pocket remained, and at approximately 0.4GPM the pocket finally pushed along and cleared, .5GPM had more than enough to move the air about.  Inverted double thickness radiators with large boxy plenum chambers mounted inverted at the top of the case probably have the greatest challenge as that’s a very large x-sectional area to move the air down, however in the context of value systems double thickness at twice the price for maybe 10% more performance certainly isn’t as likely as someone expanding with a slim radiator.  To give some room for bleeding radiators top mounted, I’m going to call 0.5GPM a good bare minimum although it is possible to have issues with premium thickness rads not wanting to push the air pockets clear in a top mounted setup.  You really should consider building the loop in the case, then removing is to bleed on the bench or at a minimum plan on turning the case around at different angles to help bleed the air out.

Power Consumption

Power consumption is actually a fairly decent measure of pumping power.  Not all are equal in terms of efficiency, but it can give you a general idea.  A D5 at full speed draws upwards of 21W and when dialed down to setting 1 as little as 3W.  Many of the smaller pumps range from about 6-8W.

The H220 draws approximately 6.4W in my test loop with the valve wide open and reduces as restriction is increased.  It’s a good amount of power to meet the needs for less complex loops.

Pump/Block PQ Curve Testing

While not a typical pump test since the block restriction is also included in the result (giving an a lower result), a PQ curve test is the most complete picture at understanding hydraulic performance and necessary to predict flow rate performance.  I also did a pressure drop test below that shows the block restriction and radiator restriction so you have everything needed to evaluate additional components if you have their pressure drop.

I’m working on the chart for the results but I got this:

Max Head Pressure (0 GPM) = 2.97PSI

Head @ .50GPM = 1.20PSI

Head @ .77GPM = .29PSI

Speed remained constant at 101Hz or 3,030RPM

Voltage was held at 12.02-12.04V

Amperage ranges from .53 at low restriction to .36 at max restriction (0 GPM).

Max Flow Rate ~0.80GPM

So, with just the H220 radiator tested below, you should see around .62 GPM, adding a triple radiator does very little and reduces flow to about .59, adding an MCW82 on top of that and you should get around .53GPM.  Full cover GPU blocks as tested per Stren’s testing here measure very similarly to the lower restriction MCW82 so you should retain over .5GPM with most CPU + GPU + Radiator type loops.

Restriction Testing

While restriction is not related to thermal performance it does relate to how much pumping power remains for adding additional part. In very simplified terms, pressure differential is what makes the fluid move about, the pump adds pressure and the parts consume that pressure. When pump and loop planning for these smaller kits, I am most interested in the pressure added and dropped in the .5-1.5GPM range. First I started with the radiator pressure drop using my usual tools to measure flow rate (King Instruments 7520 250mm Acrylic Flow Meter) and pressure differential (Dwyer 477-5 Digital Manometer). It’s a pretty simple test, I simply vary the flow rate and record the flow vs pressure drop incrementally until the full curve is developed.

That’s a bit surprising to me as I expected the H220 radiator to test much lower than that since the radiator core and tube sizes look very similar to the MCR320. The H100i in contrast does visually look like it uses much thinner tubes which is also somewhat evident in the “Less Rounded” curvature. I’ve noticed this in testing the HWlabs GTX which uses more restrictive tubes that also produced a flatter curve. My only explanation for this is the flatter tubes expand as pressure increases. In the end, the H220 and H100i are both generally high restriction radiators compared to their DIY counterparts by about 5-6X. I’m not quite sure why the H220 measures higher in restriction, but I suspect it is the much lower profile end tanks and swivel fittings since the tubes look very similar to the MCR320-QP. While radiators are generally low in restriction compared to CPU blocks, the H220 radiator is not and measured on the high side compared to usual DIY rads. The H100i rad measured higher in restriction at .5GPM, where the H220 was higher at 1GPM and they were roughly equal at .7GPM.

I did the same test on the pump/block combo to get a sense of block restriction.

While the H220 is considerably lower in restriction than the H100i by about 4X, it is still a bit on the higher side relative to other DIY blocks. It shares a very similar pin matrix to the the older Apogee XT2 except for the pump housing, impeller restriction, and swivel fittings. What this means is that even if the H220 pump could spin up to MCP35X speeds, the restriction of the block alone would limit flow rate to 1.1GPM. Now add in the radiator restriction and you are likely looking at flow rates between .5GPM and 1GPM as a guess. Generally the restriction is a bit higher than I would have liked, but thermal performance and pressure drop is always one of those balancing things. I’m pretty sure it will mean some planning is needed in regarding to what you add and you can expect flow rates below 1GPM.

Is Expansion possible, Yes. – You should have no trouble with most CPU + GPU + Radiator type loops.

Is it going to be super easy to fill and bleed, Maybe.  It is below the 1 GPM rule of thumb most use for DIY loop design, but that’s generally a high enough flow rate to ensure super easy fill it and turn it on bleeding.  If performance only drops a couple to a few degrees from 1GPM down to .5GPM, then running at .5GPM is a relatively minor performance loss.  The more notable difference will be how easy it is to get the air bled out.  If you are below 1GPM, I would recommend filling and bleeding with the loop outside of the case just to make sure you can rotate parts around and get the air out properly.

Will it perform as well as a full custom loop?  Close. – I would estimate 2-3C difference due to slightly lower flow rates, but that’s still very much a good performance.  You can also always add another pump to push in series if you wanted to step up pumping power and maximize flow rate, but your greater gains could probably be had with more radiator if you have the space.  Using something like a preimum quad radiator plus 35X2 pump could probably net you 7-8C as a guess but you are talking about a $200 pump plus a $100 for just the rad, not to mention fans, reservoir, block, etc.  The $140 H220 price point is very much a great value, you have to spend a LOT more on DIY custom cooling to get much more.

Overall, I would consider the H220 very expansion capable.  It certainly does have enough pumping power to push through the kit plus a couple of extra parts as long as you don’t mind giving a little more patience to fill and bleed outside the case and has been demonstrated to be capable of running multiple GPUs and radiators just fine.  I wouldn’t suggest stringing together an extremely complex loop, but it should be plenty for most users simply looking for a little expansion.

Regarding radiator capacity, that really depends on the heat level.  I would generally recommend at least adding an MCR120 or other 120 sized radiator if you add a GPU, but CPU + GPU could be very possible with just the H220 rad.  It just depend on heat level.  My extremely hot 180W 3930K as you’ll see in later testing does push the limits when dialing fan speeds way down, so a little extra rad would be encouraged depending on heat level.  About 100W per 120mm rad section is about max I would recommend for average fan speeds.

Pump Comparison H100i

Here is a quick comparison of pump+block vs pump+block that I have so far.  Will add more as I get the others done:

The Corsair H100i is specifically designed as a sealed unit and operates at very low flow rates, where the H220 produces much higher flow rates.  If we hold .5GPM as the “DIY parts Minimum”, then the H220 is compatible with DIY expansion and the H100i is not.  The H220 produces roughly 3X the maximum pressure head and about 7X higher maximum flow rate.

 

Pages: 1 2 3 4 5 6 7 8 9 10 11 12