[Sub-Preview] Heatkiller 3.0 LC and LT Compared + Vapor's Testbed 2.0 First Look
So this is a post comparing the LC and LT Heatkillers, as well as examining horizontal vs. vertical orientation of both as well as what happens when you remove the divider from the LT (not the impingement plate, that always stays in). None of this will be posted over at Skinneelabs as, well, this is really a sub-preview....the testing I did over the past few days was more of a test run than anything else, but everything went smoothly enough that I feel I can show the results while I wait for the final parts to come in :yepp:
As such, this is also my first data-showcase of my new testbed and what it can do for testing waterblocks :D There are a few minor improvements to make over the next week (which I will detail), and then it'll be finalized. At which point I'll retest everything I've tested here and then continue on with my regular testing/reviewing. I'll also take pictures of Testbed 2.0 at that time and share those :D
So let's get to it, shall we?
First, let's go over what Testbed 1.1 did wrong and what Testbed 2.0 does to correct those shortcomings.
Testbed 1.1 Shortcomings:
1) I was stubborn with Testbed 1.1 and really wanted to stick to the H20-220 Compact roots it was born from.
2) Going the "sandwich" route was ineffective for anything more than structural reasons....yes, having a really wide base was effective for making it very stable even with a decent amount of mass hanging off of it. But after some tests, I realized I only had ~10-15% more radiator power than using just the MCR320 with the same fannage. :eh:
3) I should have measured water temperatures.
4) Using a Koolance FM17 flowmeter, while it's surprisingly accurate when used with logging, I deemed as 'unprofessional' and also wanted higher confidence in my high-flow readings.
5) When I varied flow through the waterblock, it varied through the radiator as well. (i.e., there wasn't enough variable isolation for the waterblock)
6) Flowrates weren't high enough at the high-end of the spectrum.
7) Flowrates weren't low enough at the low-end of the spectrum.
8) A test of a single waterblock configuration on a single CPU took 5 days.
9) Due to the slight break-in of MX-2, flow vs. temp curves ended up being shallower than they should have been (fortunately I always did the tests in the exact same order! [max flowrates to min]).
10) Not every pump setting was a "hard setting." (i.e., in order to achieve 10v supply for one of the pumps, I needed to use a DMM and dial-in the pump controller; in my opinion, hard settings are always better--on/off and turning a dial all the way up or down are good examples).
Testbed 2.0's Improvements:
1) I actually remained stubborn to the concept of having a 2-piece cooler: one piece at the cpu area, the other being a remote portion containing "everything else." It remains portable, though has increased in weight immensely.
2) I now use two MCR320s mounted together at the hip....they're not sandwich style any longer (unless you count open-face sandwiches I suppose).
3) I measure a lot of water temperatures, I have 4 probes measuring water into the CPU block, 2 probes measuring water out of the CPU block, 2 probes measuring water into the radiator, and 4 probes measuring water coming out of the radiator. I also continue to use 2 probes per intake fan, so 12 for air intake measurement. Overall I have 24 Dallas One-Wire probes in use for the waterblock testbed (note, for the tests below, only 22 were used--I only used 2 coming out of the radiators).
4) I'm still using the Koolance FM17 on the radiator subloop, just to verify that flowrates do not change noticeably (they don't). But I've gone to Dwyer RMC series flowmeters for the CPU subloop. Since I need measurement range from .25gpm to over 4gpm, no single rotameter I could find fit the bill, so I got two Dwyer RMC series flowmeters and put them in series...in all, I can (very) accurately measure flowrates between .2GPM and 7GPM :D
5) I now have a setup with a shared reservoir with two 'subloops' stemming from it. This allows me to vary flowrates through the CPU block immensely (from .25GPM to 4+GPM) and have a constant flowrate through the radiators the entire time. Theoretically I didn't need to do this when I'm measuring water temperatures, but it allows me to use CPU vs. air temperatures again since radiator flow (and performance) is non-variable and disambiguated from the CPU block performance.
6) I've made numerous improvements to the loop: I've added a D5, I've gone from a XSPC Res Top + EK Dual Turbo Top to an XSPC V3 Top and two EK V2 Tops (has better performance at high flowrates, although it's not as good as the XSPC V3 at low-to-moderate flowrates [at below 2GPM, the XSPC is the better top]). I've also reduced restriction for the CPU subloop....gone are the radiators (duh) and the 3/8" tubing and in is a combination of 1/2" and 7/16" tubing.
7) Despite lowering restriction greatly, I'm also capable of measuring significantly lower flowrates due to the addition of the D5 and its incredibly weak Setting 1.
8) A test of a CPU block in a single configuration on a single CPU now takes only slightly more than a day. I can get 3 blocks/configurations done in 4 days now. And my data is more consistent and accurate due to the use of water temperature probes.
9) I get around the MX-2 break-in in two ways. First, I no longer test the entire flowrate gamut each mount (meaning intra-mount comparisons are no longer needed); and second, I do let MX-2 fully cure for 12 hours before running the flowrate gamut tests.
10) Every pump setting is a hard setting now...D5@1, I just turn the D5's dial down all the way. DDC3.2 at 7.6V, I just turn the pump controller's dial down all the way. D5@5, I just turn the D5's dial up all the way. DDC3.2 @ 12V, I bypass the controller and just plug in the XSPC DDC3.2 (note, this test was not included in this comparison, but it will be there when I do the real testing). 2xDDC3.2 @ 12V, I just plug in both EK DDC3.2s. All pumps on = all pumps on :)
11) There are a lot of little improvements here and there as well...I use BP Fatboy barbs, I use rotary extenders at the CPU block so that I don't get twisted tubing (and so that even the REALLY tightly spaced CPU blocks can be mounted with no modification to my setup), the entire front fascia of the cooler is unobstructed, the CrystalFonts and pumps and fans are on their own Zippy PSU (meaning I no longer get droop when all the pumps are going at full bore and it allows me to not have to reconfigure the CF software each time I unhibernate).
Testbed 2.0 Specs:
2xMCR320s + 12x YL D12SH-12s (push+pull)
3xDDC3.2 (2x with EK V2 top, 1x with XSPC V3 top...all in the block subloop)
1xD5 (stock top....in the block subloop)
1xD-Tek DB-1 (pump for the radiator subloop)
1xBitspower 5-port Res with a Swiftech MicroRes attached to the top to help with bleeding (still sucks with bleeding)--getting switched out to an 8-port PrimoChill Typhoon in the next few days
24xDallas One-Wire DS18B20 temp probes (12 in the air, 12 in the water)
1xDwyer RMC-142 (.2GPM to 2.2GPM)
1xDwyer RMC-144 (.8GPM to 7GPM)
1xWhatever waterblock I'm testing
Changes being made to Testbed 2.0 between the tests below and next week:
1) Switching to an 8-port Typhoon reservoir, should help me with bleeding (which is very slow right now due to a design flaw in the BP res I currently use), as well as a few minor improvements.
2) Going from serial to parallel flow through the radiators via the use of a BP Y-split coming out of the pump and using the extra ports on the Typhoon res. I really have no rationale for doing it other than I want to try it :)
3) Adding in some Tygon Silver tubing.
4) Adding two more temp probes at the water-out portion of the radiator (increases precision and maintains balance in restriction between the two subsubloops of the radiator subloop).
5) Removing a Killcoil that got jammed against some temp probes (damn you high flowrates!)
6) Switching to the Zippy PSU for the WC components (already done and tested--works great and boosts flow noticeably at the high-end due to a lack of droop--but was not used for the tests below).
Alright, some tests!
This isn't a formal preview or review, just a testlog as much as anything, but my procedure is my normal one (which I don't think any of you read anyway, lol)....it can be found in any of my other waterblock test threads. Any differences in procedure are due to new equipment, which is detailed above....fill in the blanks :stick:
CPU vs. Water Temps:
http://img527.imageshack.us/img527/8576/capturebp.png
CPU vs. Air Temps:
http://img156.imageshack.us/img156/764/capturebg.png
CPU vs. Air Temps (with 1/3rd the radiator power of my setup):
http://img142.imageshack.us/img142/715/captureda.png
(this data was calculated by taking water to air deltas and adding it three times to the water temps...it shows what someone with a single 3x120 radiator with ~1600RPM fans can expect in terms of block performance vs. flow)
Each of the 5 datapoints for each curve was at the same pumping power setting....from left to right:
Very Low: D5@1
Low: DDC3.2 + XSPC V3 top @ 7.65V
Medium: D5@5
Medium-High: DDC3.2 + XSPC V3 top @ 12V (***not used in these tests***)
High: 2xDDC3.2 + EK V2 tops
Very High: All 4 pumps on at max--D5@5 + all three DDC3.2s at 12V :devil:
Recap
This really is a quick and dirty preview of what's to come from my upcoming testing....I expect with a super-low restriction block like the Sapphire Rev. A, I should hit nearly 4GPM (I'm at 3.8-3.85ish with the Heatkiller LC with the Zippy PSU).
I think overall this testbed is a significant step forward for my testing and should be able to really hammer out some useful results quickly and accurately.
In addition to the Heatkiller LC and LT, I have a Swiftech GTZ, Koolance CPU-350, EK Supreme, EK Surprme LT, and a D-Tek Fuzion V2 (though I'm thinking of retiring the FV2, it's seen better days). I'm always on the look out to test more blocks, so if you know of an opportunity, drop me a note :)
Of course this sub-preview has none of the fit and finish of my regular reviews, but it's still fun/usable data, IMO. Seems to me the Heatkillers enjoy vertical and for an LT+i7 system, pulling out the little plastic divider at the inlet is a good idea :) Their flow dependence is really similar, with the LC not doing as well with low flow (likely due to the lack of impingement plate sealing off secondary flowpaths, IMO)....and overall they seem to have very minimal flow dependence, not having much performance scaling past a single D5.
Anyway, thanks for reading :)
(sorry for the wall of text....that's what happens when I leave a post open for a couple days and just keep adding to it :()
