Folding@Home Power Usage

Introduction

As I discussed recently, I started a SQLFamily Folding@Home team that anyone is welcome to join. Since I have far more desktop machines than any sane person should, I have a few available that I can have running Folding@Home. I was curious about how much power usage each machine was generating, so I have taken some measurements to answer that. I used a Kill-A-Watt meter to measure the system power usage from the wall at 120 volts.

BTW, the Folding@Home team has a news update that discusses their efforts to get some work units that are directly related to COVID-19 available on their public servers.

Four Similar Lab Machines

I have four recently rebuilt desktop machines that have very similar parts configurations. I bought new motherboards, CPUs, DDR4 RAM, and M.2 NVMe drives. I reused the cases, power supplies and video cards.

I took advantage of some of the sales and price reductions at Micro Center over the past several months to affordably buy the parts to make this possible. As you might guess, these are all AMD Ryzen-based systems.

Two of the systems (Systems 1 and 2) are in identical Fractal Design cases, while the other two are in identical Corsair cases. All four systems have Corsair CX500M 80 Plus Bronze power supplies and all four systems have one M.2 NVMe drive. They are all running Windows 10 Pro 1909 with the High Performance Windows Power Plan. All four systems are using the stock AMD CPU coolers that came with the processor.

Systems 1 through 4 are shown left to right in Figure 1.

System 1

This has a 6C/12T 7nm AMD Ryzen 5 3600X in an MSI B450 Tomahawk Max motherboard with 32GB of DDR4-3200 RAM. It has an MSI Radeon R7 240 video card with 2GB of RAM. This system uses 51 watts at idle and 134 watts when running Folding@Home.

System 2

This has a 8C/16T 12nm AMD Ryzen 7 2700X in an MSI B450 Tomahawk Max motherboard with 32GB of DDR4-3200 RAM. It has an MSI Radeon R7 240 video card with 2GB of RAM. This system uses 63 watts at idle and 245 watts when running Folding@Home.

System 3

This has a 6C/12T 12nm AMD Ryzen 5 2600X in an MSI B450 Tomahawk motherboard with 32GB of DDR4-3000 RAM. It has an MSI Radeon R7 240 video card with 2GB of RAM. This system uses 51 watts at idle and 187 watts when running Folding@Home.

System 4

This has a 6C/12T 12nm AMD Ryzen 5 2600 in an MSI B450 Tomahawk motherboard with 32GB of DDR4-3000 RAM. It has an MSI Radeon HD6450 video card with 1GB of RAM. This system uses 55 watts at idle and 142 watts when running Folding@Home.

Table 1 shows some performance vs. power usage metrics for these four systems. The CPU-Z Multicore score is a very rough measure of the CPU capacity of the system. I have no idea how it correlates Folding@Home CPU performance, but it is an easy way to compare the four systems.

System	CPU-Z Multicore	Idle Power Usage	Load Power Usage
Ryzen 5 3600X	4219.7	51 watts	134 watts
Ryzen 7 2700X	5028.5	63 watts	245 watts
Ryzen 5 2600X	3762.2	51 watts	187 watts
Ryzen 5 2600	3465.9	55 watts	142 watts

If you divide the CPU-Z Multicore score by the Load Power Usage (when running Folding@Home), you get a CPU-Z/watt measure as shown in the list below. The Ryzen 5 3600X system is giving much better performance per watt. This is due to its more efficient 7nm process node and architectural improvements in Zen 2.

• Ryzen 5 3600X is 31.49 CPU-Z/watt
• Ryzen 7 2700X is 20.52 CPU-Z/watt
• Ryzen 5 2600X is 20.12 CPU-Z/watt
• Ryzen 5 2600 is 24.41 CPU-Z/watt

Final Words

If I was really concerned about power usage, I could have used more efficient power supplies that would have been considerably more expensive. If I wanted better Folding@Home performance from the GPUs, I could use better (but more expensive) video cards (that would also use more power).

The 3rd Generation Zen 2 AMD Ryzen 5 3600X is significantly more efficient than the 2nd Generation Zen+ Ryzen 2xxx systems. The temperature of the air coming out of the rear exhaust fan on the Zen 2 system is MUCH cooler than the other three systems. As you might expect, the temperature of the air coming out of the rear exhaust fan on the Ryzen 7 2700X system is MUCH hotter than the other three systems.

Balancing electrical power usage and cooling needs vs. performance is very important for large scale data centers. It is also important when running an ad-doc data center in your basement! These four systems use 708 watts while running Folding@Home. If I run them 24×7, they will use about 510 KwH per month. That is nearly as much as my whole house uses.