Test Methodology

Robela's H20 performance was measured on the Opteron 148/DFI LAN Party UT running the system under full LOAD using S&M CPU/Memory stress tester, specifically S&M v.1.7.6 (160) . The CPU tests were run at default (11x200FSB) and overclocked (10x250FSB) both at default Vcore. I repeated these tests with VGA water-block removed from the loop, this is the last temp included on that chart. The Ambient temp outside the case fluctuated between 18-21°C during tests.

To test Robela's VGA cooler in particular I ran ATI Tool 0.25 Beta 14, "Scan Artifact" utility. As I did with the CPU test I removed the CPU Cooler from the loop, replacing it with a stock cooler to test Robela's cooling potential running the VGA cooler alone.

AMD Test System
CPU	Opteron 148 (CABE2) Socket-939 Retail
Mainboard	DFI LAN party UT nF4 Ultra-D
Memory	Corsair XMS-Pro TwinX2048-4400PRO
Graphics	Sapphire Radeon X800XT
Power Supply	In-Win IP480Q3-2 480W
Cooling	AMD Stock Retail cooler for Opteron Thermaltake RX-K8 Silent Boost Alphacool CPU-waterblockNexXxos XP nickel plated. Alphacool pump AP1510 Centrifugal w/Voltage regulator. Alphacool CAPE Cora 642 passive radiators Titan Robela
Operating System	Windows XP

As indicated above using the stress test utility S&M, Robela's included water-blocks were mounted on the Opteron 148 and X800XT GPU. Tests were performed at 2.2GHz and 2.5GHz. For comparison both AMD stock air-cooler and Thermaltake's RX-K8 comprised two more tests while concurrently running the Sapphire X800XT with its stock air-cooler. For an H20 comparison I installed the Alphacool NexXxos XP CPU-waterblock mated with the Alphacool CORA passive radiator.



Testing Robela's VGA block simply involved running Artifact Scan on the ATI Tool panel. Where both blocks were installed and VGA core temps were increased using ATI Tool I did notice the CPU temp rose along with the GPU temp, unfortunately this is indicative of over-taxing Robela's H20-cooling system.



Sound Testing:

This is one area where Robela out-performed every other system I've ever owned or tested to-date. Similar to what has occurred while operating Alphacool's passive system, I actually had to check to ensure the pump and radiator fans were working. Titan allows for a 60m front intake fan, although that fan was not included with the kit. This leaves us with 2 x 120MM intake radiator fans and a 90m rear exhaust fan.



Conclusion:

The Robela water-cooling case is most definitely an enigma in design execution. Many of the attributes which piqued my interest in this case became the source of problems during testing.

The primary means of measuring/monitoring CPU-temps is through a supplied thermistor. Instructions are specific; the thermistor is to be placed on the CPU-water block inlay. A thermistor may have been acceptable a few years ago when CPU temps were measured via Socket thermistor; however, today CPU's with internal thermal diodes are much more accurate then any thermistor. At the very least they’re more consistent in their inaccuracy compared to thousands of end-user's placing a thermistor even with the most precise instructions.

In fact due to my concern for this I decided to try installing the thermistor between the CPU IHS and water block base, albeit at the edge so as not to disturb contact between the two. In some respects I didn't disturb the contact between the two, because there was barely any contact at all, seen below.



The photo above is one reason I reject spreading as opposed to placing a dollop of thermal paste on a CPU IHS and allow mounting pressure to compress the paste, forcing air out. It would be much more difficult to determine how little contact had occurred between these two surfaces had the paste been spread. Very few base plate/CPU surfaces are truly "flat" and are usually replete with micro-striations and micro-pores. Once a dollop of paste is paced in the center of the IHS, mounting compresses the paste so that it seeks its own natural “footprint”.

TIM install methods aside the only perceivable difference between mounting the thermistor incorrectly and re-mounting the block were lower temps. It’s difficult to determine what effect this had on the cooling system because moving the thermistor may have dropped the temp measured, but the CPU temp also changed as temps decreased from a more effective mount.

With a few more improvements such as a more powerful pump and two more 120mm fans I believe Robela would give the competition a spanking. If it were my choice I'd eschew any self-automating system and simply place a rheostat so the end-user can adjust fan speeds, possibly adjusting pump voltage as well, but why even mess with it. Automation can be your best friend or your worst enemy, unfortunately with the Robela it is the latter.

PRO
All in one solution
Able to accommodate just about any motherboard/CPU video card combo.
Perfect solution for first time H20-enthusiasts
Quiet!

CON
Robela's large and potentially high-performance radiator which could accommodate four 120mm fans with room to spare, only uses two.
USB/Firewire front connectors cumbersome to hook up
The front intake fan at 60mm would be inadequate for any case, and offer a poor performance/noise ratio.
Robela's water-pump is woefully underpowered.
While Titan made a valiant effort to automate their case, the complex series of temp and water-level sensors may actually place restrictions on the end-user as well as potentially limiting performance. If any aspect of this system were to fail the solution would involve shipping the entire case since the sensors and monitoring hardware are located in three different areas.

At approximately 300USD the case isn't prohibitively costly nor is it a bargain, not in its current configuration. There's room for improvement and with that improvement Robela would be an extraordinary case, at this juncture I'd still say it's one of the best solutions out there for first time H20-users.

Questions/Comments: forum thread