Today we are going to take a quick look at the Supermicro Hyper-E that I recently got some hands-on time with at Supermicro headquarters. The 2U Hyper-E platform we are looking at is specifically the SYS-220HE-FTNR. This system can pack two 3rd generation Intel Xeon Scalable CPUs and a system full of storage and accelerators into a 22.6″ / 574mm chassis. Since it is designed for the 5G edge space, it also has front I/O making it very different than some other options in the market.
This is part of our visit to the Supermicro HQ series. Last year we looked at the Supermicro 3rd Gen Xeon Scalable Server which we showcased alongside the Cooper Lake launch. We also looked at the Simply Double platform and the company’s new 60-Bay Top Loading Storage Server and a 1U Half-Petabyte EDSFF Server. This year, we were able to take a look at four new systems, starting with the Hyper-E:
We are going to have more detail in this article, but want to provide the option to listen. As a quick note, Supermicro allowed us to film the video at HQ, provided the systems in their demo room, and helped with travel costs to go do this series. We did a whole series while there and are tagging this as sponsored. I was able to pick the products we would look at and have editorial control of the pieces (nobody is reviewing these pieces outside of STH before they go live either.) In full transparency, this was the only way to get something like this done, including looking at a number of products in one shot, without going to a trade show. Look for more in this series coming to STH over the coming weeks.
Why the Buzz About Edge Servers?
At STH, we have been covering edge servers, often used as appliances, for years. Recently, there has been a major uptick in how often we are covering them, and with industry efforts around developing new models. One of the big drivers of this is the 5G infrastructure build-out and initiatives like open radio access networks. Somewhat related, as AI has moved from the realm of research to application, performing low-latency inference at the edge especially from video and other sensor sources has become a more pressing need. One of the key challenges is that often in places such as racks below 5G towers, store IT closets, and so forth, the spaces and layouts are considerably different versus traditional data centers. As a result, new servers need to be designed to operate in these environments and that is why I wanted to look at the Supermicro Hyper-E.
Supermicro Hyper-E / SYS-220HE-FTNR Hands-on
Perhaps the first major aspect of the system is that it is a 2U front I/O system. Perhaps more succinctly put, all of the cables that would be connected to this system can be connected via the front faceplate. This means that everything, including the dual 2kW 80Plus Titanium power supplies, can be serviced from the same aisle.
Other cables that can be connected are those such as cables to the IPMI management port, USB 3 ports, and VGA ports. This standard I/O block does not include onboard networking.
Instead, Supermicro has its AIOM slots. These are OCP NIC compatible slots and there are two of them in this system. As opposed to older risers and PCIe card NICs, these slots are rear serviceable. That means one can change NICs without having to remove the system.
The three risers are all tool-less. making them very easy to remove. These risers service the eight PCIe expansion slot I/O plates on the front panel and can fit NICs or GPUs.
The system we had access to was pulled (still warm) from a Supermicro lab and did not have the GPUs installed, however, there were GPU power cables available. This system can handle not just high-performance NICs and FPGA accelerated NICs, but also high-performance GPUs and up to four double-width accelerators in the system.
Our system had a Supermicro AOC-SHG3-4M2P PCIe switch card to further extend customization. One can add NVMe SSDs or potentially even lower-power AI inferencing accelerator M.2 cards to the system to further extend expandability.
M.2 storage is not limited to just the card. Instead, there are two onboard M.2 slots if we look below the middle riser.
At the edge of the motherboard PCB (right), there are two M.2 slots that extend into the chassis. Drives are secured via spring-loaded push tabs. Our unit did not have M.2 SSDs, but hopefully, you can see the black push tabs mounted on the metal mounting plate.
Behind all of this I/O expansion are two 3rd generation Intel Xeon Scalable processors, code-named “Ice Lake.” Last year when we did our visit to Supermicro we looked at a 3rd Gen Intel Xeon Scalable “Cooper Lake” system. The new Ice Lake generation brings features such as more cores, PCIe Gen4 with up to 128 lanes in a dual-socket server, 8-channel memory, and PMem 200 support, as well as acceleration for AI inference. There is a lot new with the new generation that you can read about in our Intel Xeon Ice Lake coverage.
This system can support up to 270W TDP CPUs so one could configure it with a Platinum 8380, Intel’s top-bin SKU. Even though it is a shorter depth server, it still has a full set of 16 DDR4 DIMM slots per CPU for 32 total enabling higher memory capacities.
Moving to the rear of the system, we can see fans and SSD trays, but nothing that requires cabling.
One newer feature in this server is that this is using Supermicro’s tool-less 2.5″ drive trays that are thinner to allow for up to six NVMe SSDs across the rear of the server.
There is a small backplane that provides power and data feeds to these drives. The PCIe lanes are delivered via cabled connections.
The six fan units are hot-swappable and come in their own carriers.
It seems as though the key design principles here were that the front should have all cabled, or potentially cabled components while the rear has hot-swappable components that do not use external cables. This is one of those key design principles for the 5G/ edge servers that we discussed earlier in this article.
Overall, the Supermicro Hyper-E 2U system showed us a balance of leveraging technology used in other Supermicro products while applying that technology to a specific edge format.
When Supermicro let us use their demo room, this is a system I wanted to look at since it really shows not just the format differences for these edge servers, but just how powerful they are. Even with a relatively short depth, this still is a full-fledged server.
I just wanted to again say thank you to the Supermicro team for “giving me the keys” to their demo room for the day. We have some other cool systems that we did during this trip that we will show off over the next few weeks.