Saving money and the environment with redundant power solutions

Started by socialreger, Aug 09, 2022, 12:27 PM

Previous topic - Next topic

socialregerTopic starter

Don't rush to get rid of unused server cases, as there's an opportunity to save money on a new platform.

 Standard Common Slot power supplies that appeared around 2010 are inexpensive and in good condition on eBay and AliExpress. Wholesale lots of power supplies can be purchased at a time. Common Slot power supplies are quiet, with an efficiency of 93-95%, and up to eight can work in parallel. To turn them on, connect pins 36 and 37 on the bus through a resistor of about 22 kOhm and pin 33 to a protective conductor. A ready-made distribution board can be bought or one can be made using a Wingtat Model 2.54 EDGE SLOT DIP 180° SINGLE LEAF TYPE WITHOUT EAR High Power S-64M-2.54-5 slots power connector.

Installing a universal 750 W HP Common Slot power supply into a server case is possible by cutting off excess and fixing the block into the case with double-sided mounting tape. This saves money compared to buying native PSUs, which can be expensive and worn out.

By using Standard Common Slot power supplies, we can save up to €600-€700 on new platforms and keep the cost of renting dedicated servers at market level. Additionally, we can reuse old cases and reduce waste in the environment. Power supply failure is rare, and a dedicated test stand is available to ensure functionality.

Using two power supplies can provide redundancy and fault tolerance, which is necessary in a data center where a single power beam cannot guarantee uninterrupted power. A test bench video will show that the solution described in the topic is effective, allowing for quick and low-cost assembly with guaranteed or redundant power.


It's essential to have the ability to maintain servers without interrupting services, whether redistributing loads or migrating services to other capacities. In a cluster, automatic redistribution occurs when a server dies, but dedicated servers require hot service work.
Based on personal experience, replacements are usually for dead screws, faulty bp, or i/o issues in storage systems. Motherboards, processors, and memory rarely fail during operations, and broken memory is typically detected during pre-operational tests.


In the past, a burnt-out power supply unit on an old Intel server posed a problem as the manufacturer no longer supported it and obtaining used parts was not possible. The solution was to replace the power supply unit with a compatible PC unit, though the case could not be closed. The server functioned well for several years without issues, though web server reboots occurred occasionally.

Upon inspection, the power supply unit was identified as the problem and a replacement resolved the issue. It was surprising that the super micro server did not have a hotswap bp, leading to the conclusion that there may be optimized PSUs that work using double-sided tape.