Test stand and conditions

Test stand

• CPU: AMD Ryzen 9 3900XT;
• Motherboard: ASUS ROG CROSSHAIR HERO Wi-Fi (UEFI v 2103);
• DRAM: Corsair Vengeance LPX 4133C19;
• GPU: ASUS ROG Strix AMD Radeon RX 5700XT;
• Drive: Gigabyte Aorus NVMe Gen4 SSD 2TB;
• Driver: Samsung 960 PRO 512 MB;
• PSU: ASUS ROG Thor 1200;
• CPU cooling: EKWB Quantum Velocity;
• CPU cooling: EKWB Quantum Magnitude;
• CPU cooling: HEATKILLER IV PRO;
• CPU cooling: OPTIMUS FOUNDATION CPU BLOCK;
• CPU cooling: TechN CPU Waterblock;
• Pump: EKWB EK-XRES 140 Revo D5 RGB PWM;
• Radiator: MORA 420;
• Fans: Arctic P14 PWM PST;
• Flow sensor: Aqua Computer's Flow Meter;
• Temperature sensors: Aquaero;
• Controller: Aquaero 6 LT;
• Thermal paste: ARCTIC MX-4;
• Case : Streacom BC1.1 Benchtable;

Test Conditions

Condition No. 1. Thanks to the Aquaero 6 LT and aquasuite the hysteresis of the Inlet port (24 degrees) was set. The essence of this idea is a constant water temperature, which will come into the waterblock. Thanks to this, we get constant test conditions, which are not affected by room temperature. In the end, this will give us an extremely small inaccuracy.

Condition number 2. MORA 420 was used to keep the water temperature constant. The largest radiator with huge heat dissipation capacity (and the largest hydrodynamic resistance). Complemented this tandem 18 fans Arctic in Push-Pull mode.

Condition number 3. Constant laminar flow. Pump speed was selected individually for each waterblock to get 1 GPM (gallon per minute). Note that each hose bend or angle fitting affects the flow, so each user system will have completely different values.

Condition number 4. To minimize the inaccuracy, each test will only start when the water temperature in the system stabilizes on a given hysteresis.

Condition 5. AIDA 64 CPU stress test in FPU mode. This is the only stress test that has minimal load variation (the main reason why I gave up Prime95 with a fixed FFT). The time of one test is 30 minutes. 3 tests for each waterblock.

Condition 6: The waterblock tests were done in two thermal modes. The first is a standard boost with a maximum PPT of 142 W. The second mode - 170 W, this overclocking on the CCX 4500 4500 4400 4400 at 1.325 V.

Condition No. 7. AMD Ryzen 9 3900XT. Processor with high static leakage currents. These processors warm up more but allow to reach higher frequencies at lower voltage. These are the crystals AMD uses in Zen 3 too.

Nuance number 1. OPTIMUS FOUNDATION CPU BLOCK received only 187 l/h due to HDR at 100% pump speed. This is slightly less than the required 240 l/h. But, fortunately, this flow rate is enough to consider the data valid.

Nuance number 2. The EKWB Quantum Magnitude waterblock used a "flat cold plate", because the default "cold plate" was too "humpty". Let me remind you that heat sources in Zen 2 and Zen 3 are shifted relative to the center, the correct contact of the waterblock plays a big role.