Larkooler iSkyWater 300 DIY 240mm Kit

Posted: February 24, 2013 in Kits
Tags: , , , , ,

THERMAL TESTING (Phase 1 Kit Only)

My intent in Phase 1 testing is to test as closely to real world as possible including a case condition. Through trials testing I’ve established that I will be using the NZXT Switch 810 bottom mount location. The case fans are all left at full speed throughout the testing. The Larkooler fans and pump do not have PWM capability so the only means to control speed is through a manual fan controller. So with that I will focus on reducing the fan speeds using my Crystalfontz CFA-633 to run the fans at full speed and log RPM for max performance testing.

Larkooler-iSkyWater-300-25

PHASE 1 “KIT ONLY” TESTING SPECIFICATIONS

  • NZXT Switch 810 Full Tower Case, Bottom Mount, Dust Filters Removed (Bottom and Front).
  • Case top grill open, Case doors on and close, Case front panel on (filters removed) – Closed Case Testing
  • NZXT 140mm 1300RPM case fans, 2 intake front, one recirc, one exhaust back, one exhaust top.
  • Corsair TX650 Power Supply
  • EVGA 570GTX Air Cooled
  • ASrock X79 Extreme 7
  • i7-3930K processor, overclocked to 4.5Ghz at 1.344-1.352Vcore, Per CoreTemp this is putting out roughtly 170-182W so it’s very aggressive and about double that of many processors.
  • Prime 95, Custom Small 8K FTT is started to load 12 worker threads at 100%.
  • Thermal Sensor logging: Crystalfontz CFA-633, logs up to 32 temperature probes and up to 4 fan channels:http://www.crystalfontz.com/products/633/index.html
  • 2 Air “Ambient” thermal sensors, 1 Air “Out” sensor. 6 core sensors
  • Minimum 20min warm up time, likely a 1 hour log minimum. Review and trim stable data.
  • Phase 1 testing will be kit only parts, Kit Fans, Kit TIM, Kit everything, no mods.
  • Kit TIM using the large “+” method.
  • For Kits without fan speed control, I am using my Crystalfontz CFA-633 to adjust fan speed and log fan RPM. This allows precise 1% speed control and the desirable logging capability.

100% Pump, 100% Larkooler Fans In Push UP FULL SPEED

First I ran a full speed and produced the following result:

Larkooler-iSkyWater300-TH01

Unfortunately, it did not do very well. Even at full speed I was pushing one of my cores to an 89C max with a fairly large spread between minimum and max core temps (about 14C). The average core temp hovered right under 80C and resulted in an core-air delta of 56.5C. Perhaps this was due to a bad mount, but that’s what I got. Here is the thermal spread after removing the block:

Larkooler-iSkyWater-300-26

The block was slightly shifted to the right and pressure was also more rotate which happens when blocks have no fixed pressure system. The bowing on this block also looks to be fairly aggressive leaving only a very small contact patch. In my past testing this sort of bowing is more desirable for smaller processors such as the 2600K, where the 3930K generally prefers much more gradual bowing since much of the core extends out past the optimal contact patch area.

Retesting Full Speed, Mount #2 “Spread Method” & Careful Centering.

So, I removed the block and decided to try the instructions recommended spread thin method instead of the plus.  I did notice that during my first mount that the block was slightly shifted to the right.  It seems the corners of the IHS on the block actually don’t extend far enough to cover the entire processor either.  The mount was not terrible, but definitely a bit off centered.

Larkooler-iSkyWater-300-27

So..I retested and spent some extra time to visually center the block.  The hold down plates are a bit loose, so it seems you can move the block around a bit more than usual.  The second mount did improve differential between cores rather significantly, but the average was still within two degrees of the last but better for sure.

Larkooler-iSkyWater300-TH04

That’s about a 1.8C improvement over mount #1 which is good, but more importantly the core variations is much better.

Unfortunately, I believe the LEDs are making speed control a challenge for my Crystalfontz to lower speed. As soon as I lower RPM just a little below 100%, it falls down to nothing and I am unable to control speeds to 1300RPM as desired.  The Crystalfontz does use PWM to control fan speeds and I’ve seen this before, I think it has to do with the LEDs causing an error in the RPM readout or something. I will have to find a different way to reduce fan voltage to test at the slow 1300RPM speed.

Comparison So Far

Since I only have two high-speed runs on the Larkooler and no noise data, I’ll leave you with this until I get more testing complete.  I averaged the two Larkooler mounts and compared to max setting on the H220 and XSPC kits:

Larkooler-iSkyWater300-TH05

Pump 100%, Fans 1300RPM

So in order to test out slower speeds, I had to add in my sunbeam rheosmart to voltage control the fans manually.  The LED fans just don’t like PWM control from either the crystalfontz or the Sunbeam while in PWM mode.  Not sure why LED fans have trouble with PWM control, but they do.  Anyhow, good old analog voltage control does work, so I dialed the fan speeds down and got this result.  It did not quite finish the logging of the test before crashing, but I think I got enough.  Core 0 was my hottest core and made it up to 85C, but I’m not quite sure why it crashed.  I’ve seen temps up to tjmax before without crashing, but it didn’t like something.

Larkooler-iSkyWater300-TH06

The result is actually fairly good for slow 1300RPM speeds, so I think the radiator is actually more low speed tuned than high speed.   I expected the temps to rise more than they did which is good.  The end result is still a bit behind the H220 and XSPC kits, but I thought it was still interesting how the temp rise was less than the other two.

Larkooler-iSkyWater300-TH07

The XSPC and Swiftech kits still have a good advantage at like RPMs, but the gap does shrink a bit.  This is just an RPM look though.

Phase 1 Bottom Line Results Core Temp vs dbA

I have noise runs done, just need to process the video, but this is what I got when converting the RPM data over to noise level and then comparing noise to core temperature.

Larkooler-iSkyWater300-TH08

This is kind of the meat and potatoes result of phase 1.   From a noise level of min to max, all three kits are relatively similar although the Swiftech H220 is in the lead on both noise and max performance fronts by a bit.

[TESTING IN PROGRESS, MORE TO COME…]

Logo-FrozenCPU

Pages: 1 2 3 4 5 6 7 8

Comments
  1. M3TAl says:

    Did you end up getting Phelan’s H100i shipped to you?

    This message thing keeps changing my username, I think I’ve fixed it now.