WilliamLam.com

64GB memory on the Intel NUCs?

by // 62 Comments

I just got my hands on a pair of 32GB SODIMM memory modules (64GB total) which I had been waiting to evaluate since last Fall. Apparently, it has taken some time for these high capacity memory modules to be readily available in the consumer market. Even after the announcement of the new 2018 Apple Mac Mini last year, which officially supports 32GB SODIMMS, I was not aware of any vendors who were selling these modules direct to consumers.


My primary interests in these memory modules was whether they would work on the latest Intel NUCs, specifically the Hades Canyon (NUC8i7HNK) which are the prosumer versions of the standard Intel NUCs that many folks use for vSphere Home Labs. Both the standard and Skull/Hades Canyon NUCs all officially support a maximum of 32GB of memory (2x16GB SODIMM), however it been hypothesized by the community that they *should* in theory be able to go up to 64GB, especially as some of the newer CPUs technically state support for it.

UPDATE (10/30/20) - Thanks to Ariel Sanchez who shared the Crucial 2x32GB SO-DIMM also work with the Intel NUC. It was a killer deal during Amazon Prime week, at $164 for 2x32 (64GB) but as of right now, they are going for $219 which is still cheaper than the Samsung which are going for $120 per 32GB SO-DIMM.

[Read more...]

GPU Passthrough of Radeon RX Vega M in Intel Hades Canyon

by // 71 Comments

With the latest Intel Hades Canyon now being able to run ESXi, a number of folks have been interested in taking advantage of the integrated GPU that is included in the system. There are two models of the Hades Canyon, NUC8i7HNK which is the lower end system with Radeon RX Vega M and the NUC8i7HVK which is the higher end system with Radeon RX Vega GH. One of the first thing I had attempted after getting ESXi working on the Hades Canyon was to try to enable passthrough of the iGPU into a Windows GuestOS but in all my attempts, it resulted into a PSOD'ing the ESXi host once you start installing the AMD Drivers from Intel.

A few days ago, one of my readers, Chris78 shared an update where he was able to prevent the ESXi host from PSOD'ing by adding a VM Advanced Setting but he he was still having issues where the Windows GuestOS would now BSOD. This sounded promising, I figure it would not hurt to gave it a try and to my surprise, I was able to successfully passthrough the iGPU to a Windows 10, Windows Server 2016 and 2019 system from my limited testing. After reporting the success back to Chris78 who was still having issues even after using the settings I had used, his conclusion was there may be a difference between the HNK and HVK models, with the latter having BSOD issues. For now, it seems like iGPU can only be passthrough if you have the NUC8i7HNK model.

[Read more...]

Update on running ESXi on Intel NUC Hades Canyon (NUC8i7HNK & NUC8i7HVK)

by // 55 Comments

The Intel NUC is one of the most popular and affordable hardware platform for running vSphere and vSAN Home Labs. For customers that want a bit more computing power, Intel also has their Skull Canyon platform which was released back in 2016 and has also gained in popularity amongst VMware Home Labbers. To be clear, the none of the Intel NUC platforms are on VMware HCL and therefore are not officially supported.

Earlier this year, Intel released their second generation of their higher-end Intel NUCs dubbed Hades Canyon which comes in two flavors NUC8i7HNK and NUC8i7HVK, with the latter being the higher-end unit. Based on the previous generation of hardware, most customers assumed ESXi should just work and went out and purchased the lower-end "HNK" version just to find out that was not case. The ESXi Installer would boot up to a certain point and then stop with the following error:

“Shutting down firmware services…..

Using “simple offset” UEFI RTS mapping policy”

To add to the confusion, this issue was not observed with the higher-end NUC8i7HVK model which was also quite interesting. Over on the nucblog.net, they also confirmed ESXi runs fine on "HVK" model and the issue seems to be isolated to the lower-end "HNK" model.

[Read more...]