ESXi on SimplyNUC extremeEDGE 3000 Series

The folks over at SimplyNUC has recently launched an exciting new Edge-focused product line called the extremeEDGE (EE) Servers and I recently had the opportunity to try out their latest 3000 Series (3rd generation of their EE platform) in my homelab for some testing.

At first glance, the kit reminds me of the SolidRun V3000 kit, which I had reviewed last year. It turns out SimplyNUC has partnered up with SolidRun to design their latest EE Servers and one of the major design change is the inclusion of a Baseboard Management Controller (BMC) interface for remote out-of-band (OOB) management.

Compute

The EE 3000 Series includes three models leveraging various AMD Ryzen-based processors:

• EE-3000 (AMD V3C18I)
• EE-3100 (AMD 7840U)
• EE-3200 (AMD Pro 8840U)

All three AMD kits support 16 CPU Cores and up to 96GB of memory using dual DDR5 SO-DIMM memory modules. For testing, I was provided with the EE-3200 kit (pictured above) which is a Zen 4, Ryzen Pro 8000 CPU series (Hawk Point) and the system was fully loaded with 96GB of memory.

Network

The networking on the EE-3200 comes with 4 x 2.5GbE (Intel i226-IT) which are fully recognized by latest ESXi 8.x as you can see from the screenshot below and should be plenty for all of your networking requirements from a basic vSphere or vSAN  deployment to a full VMware Cloud Foundation (VCF) solution, providing flexible networking options based on your needs.

While the EE-3000 kit includes an additional dual 10GbE SFP+, these are NOT recognized by ESXi as shared in my SolidRun V3000 review and PCIe passthrough of these devices to a VM will also not work. If you plan to run ESXi, you should consider either the EE-3100 or EE-3200 systems.

Storage

All EE 3000 Series systems support up to 3 x M.2 PCIe x4 Gen 4 (2280) and 1 x M.2 PCIe x4 Gen 3 (2242), which provides plenty of storage capacity whether you are considering vSAN Original Storage Architecture (OSA) or Expression Storage Architecture (ESA) along with using one of the M.2 for ESXi installation and local VMFS. For my EE-3200 kit, only the M.2 2280 were populated and again, these are fully recognized by ESXi as you can see from screenshot below.

With the number of M.2 slots, you can also take advantage of the new NVMe Tiering capability that was introduced with vSphere 8.0 Update 3, which increases the available memory for your workloads. See the VMware Cloud Foundation (VCF) section below to see the power of NVMe Tiering!

Baseboard Management Controller (BMC)

As mentioned earlier, all EE Series systems includes a BMC interface for remote management, which is a very welcome feature, especially for folks deploying these systems at the Edge or at locations that is not easily accessible. OpenBMC is used as the underlying implementation, which means there are no additional licensing and it is fully compatible with the Redfish specification.

The initial setup of the BMC will require the use of the serial console port located at the top of the unit, especially if you want to setup a static IP address which you can follow instruction guide HERE.

Once the BMC is on the network, you can open a browser to the configured address and login with the default credentials which is root/0penBmc as shown in the screenshot above.

Since OpenBMC is Redfish compatible, you can also automate various aspects including remotely power on and off the system using the Redfish API. The following cURL snippet will confirm that you can remotely connect and discover the system name which by default is named "system"

BMC=1.1.1.1
USER=root
PASS=0penBmc

curl -s -u $USER:$PASS -k --header 'Content-Type: application/json' --header 'Accept: application/json' -X GET https://$BMC/redfish/v1/Systems

The following command will remotely power on the system:

curl -s -u $USER:$PASS -k --header 'Content-Type: application/json' --header 'Accept: application/json' -X POST -d '{"ResetType": "On"}' https://$BMC/redfish/v1/Systems/system/Actions/ComputerSystem.Reset

The following command will remotely shutdown the system:

curl -s -u $USER:$PASS -k --header 'Content-Type: application/json' --header 'Accept: application/json' -X POST -d '{"ResetType": "GracefulShutdown"}' https://$BMC/redfish/v1/Systems/system/Actions/ComputerSystem.Reset

It is really nice to see this enhancement from SimplyNUC as it is quite rare to see a proper remote management interface for small form factor systems, hopefully other vendors take note 🙂

Security

All EE 3000 series kit includes a Discrete TPM 2.0 device, but the default behavior seems to enable only the fTPM setting, which is not supported by ESXi. On the EE-3200, the option to enable the dTPm is slightly different from the SolidRun V3000 and it took me awhile to find the correct setting. To enable dTPM, you will need to go into system BIOS and go to Advanced->AMD CBS->SOC Miscellaneous Control->Trusted Platform Module and then select the "Enable dTM" option and then reboot and then ESXi will be able to use the TPM device.

Graphics

The EE-3200 includes a Radeon 780M integrated graphics (iGPU) and while attempting to enable PCIe passthrough using the ESXi Embedded Host Client UI, the setting could not be toggle which got into this weird loop/refresh of the screen. The observation I have had with AMD-based kits is that its integrated graphics is typically connected to several other devices including USB, so I am wondering if multiple devices are connected, so passthrough of the device is causing some unexpected UI issue.

Luckily, you can easily toggle PCIe passthrough vSphere API or ESXCLI, which I have demonstrated by running the following command, allowing you to consume the iGPU within a VM.

esxcli hardware pci pcipassthru set -d 000:09:00.0 -e true

Once passthrough has been successfully enabled for the iGPU (no reboot needed) you should now be able to add the device that is labeled Phoenix3. Below is a screenshot of the AMD iGPU being successfully passthrough and fully recognized by an Ubuntu 24.04 VM without any additional driver installation.

I did attempt to passthrough the iGPU to a Windows 11 system, but none of the AMD Graphics drivers was able to successfully detect the iGPU. In fact, it was complaining that it was not detecting a supported AMD platform and after trying several different drivers and Windows versions, I eventually gave up.

Form Factor

I was very impressed at how small and compact the design of the EE-3200 was, definitely fitting for Edge locations. The fanless design can withstand some impressive temperature ranges from -40C to 85C and best of all, no fan noises! While the picture above gives a relative size comparison to other popular kits (Top Left to right: Intel NUC 12 Enthusiast, ASUS NUC 14 Performance, ASUS NUC 14 Pro, EE-3200 and Supermicro E200-8D), the small footprint of the EE-3200 is just slightly larger than your classic 4x4 system and is definitely packed with ton more capabilities, not to mention it looks pretty cool too!

ESXi

The latest release of ESXi 8.0 Update 3 installs fine without issues, no additional drivers are required as the Community Networking Driver for ESXi has been productized as part of the ESXi 8.0 release. If you want to install ESXi 7.x, you will need the Community Networking Driver for ESXi Fling to recognize the onboard 2.5GbE network devices.

VMware Cloud Foundation (VCF)

The EE-3200 can also be a nice kit to deploy either a physical VMware Cloud Foundation (VCF) if you have several of these units or you can deploy it on a single host using Nested ESXi, which is made possible by leveraging the new NVMe Tiering capability as part of vSphere 8.0 Update 3 and as you can see from the screenshot below.

I was able to expand the amount of available memory from the physical 96GB of DRAM to almost ~500GB using a 400% configuration, giving you the ability to deploy VCF and other workloads with no noticeable impact from my testing. The full VCF deployment took a little over ~2hrs and having access to full 16 cores is certainly welcome when comparing to other consumer platforms.