Gigabyte Finally Makes A Blade Server

MWC 2025: From small to large Gigabyte does most everything

Gigabyte had two new items at MWC, one bigger than average and one smaller. Of the two, SemiAccurate thought the smaller blade debut was the most interesting.

Blades are nothing new to the industry, they have been around since the days of single core CPUs. At Gigabyte they are brand new, in fact they debuted at MWC this year. The all new B343-C40 line, as shown here by the -AAJ1 variant, is likely the first of many even if Gigabyte would only talk about this specific version. While you probably can figure out the specs from the name, for the unwashed it is a 3U AMD single socket blade, 10 blades in a chassis.

Gigabyte blades

Yes Gigabyte FINALLY makes blades

Socket AM5 can take Epyc 4004 or Ryzen 7000/9000 CPUs and supports 2 DIMMs per channel. This is basically a pimped out desktop with ECC and hot swap features rather than a full blown ‘server’ blade. To some this may seem underwhelming but there is one upside, you can put an APU in it for graphics or video workloads, plus you can add a GPU on top of it for more graphics grunt should you desire. If your workload is more than pure compute/HPC, these blades offer some nice options. All this said, more conventional blades with other CPUs, both AMD and other, are sure to follow in the near future.

Gigabyte Gigapod compute node

One fifth of a Gigapod on display

Next up is the Gigapod which looks to the untrained eye like a rack of servers. Each pod has two main types of rack, a compute rack and a central node. A full Gigapod is made up of four compute pods and one central node for a five rack ‘device’. The racks are water cooled with the controller at the bottom. The variant pictured above is a compute node.

If you can’t tell from the color of the drive bay releases, this is an Nvidia AI node, a combination which will make up the majority of the unit sales for Gigabyte. The individual servers are interesting too with a split design that puts the CPU in one chamber and the GPUs in a larger chamber above it.

The beauty of this design is you can add a sled to either chamber with any CPU you want, AMD, Intel, or Ampere are the main ones at the moment. On the top you can add any GPU you have a particular liking for, Nvidia, AMD, or Intel, or anything else you want. Tseng Labs ET2000s are pretty cheap nowadays and with their TDP you can pack a lot of them in. Think about it. AMD + Nvidia may be the node of choice at the moment but the Gigapod is designed to mix and match at scale with scale defined as a 5-rack minimum unit.

To say that Gigabyte pleasantly surprised us at MWC this year would be an accurate summary of their offerings. The all new blades have a lot of promise, the ability to put consumer CPUs/APUs in the same managed framework as full on server CPUs offers a lot of flexibility. SemiAccurate is run on a Gigabyte server at the moment and the manageability is quite nice. On the other end of the scale having a flexible Gigapod managed as a five rack unit is also quite interesting. I can’t wait to see what they add at Computex.S|A

The following two tabs change content below.

Charlie Demerjian

Roving engine of chaos and snide remarks at SemiAccurate
Charlie Demerjian is the founder of Stone Arch Networking Services and SemiAccurate.com. SemiAccurate.com is a technology news site; addressing hardware design, software selection, customization, securing and maintenance, with over one million views per month. He is a technologist and analyst specializing in semiconductors, system and network architecture. As head writer of SemiAccurate.com, he regularly advises writers, analysts, and industry executives on technical matters and long lead industry trends. Charlie is also available through Guidepoint and Mosaic. FullyAccurate