1st and 2nd Gen Intel Xeon to 5th Gen Intel Xeon Consolidation
From 2017 to 2021, LGA3647-based Xeons were sold in huge volumes. Those same servers are now 3-6 years old, meaning they are squarely in the era where replacement cycles kick in.
From 2017 to 2021, LGA3647-based Xeons were sold in huge volumes. Those same servers are now 3-6 years old, meaning they are squarely in the era where replacement cycles kick in.
Emerald Rapids is an improvement over Sapphire Rapids which just launched 11 months prior. The fact that Intel is able to move this far in 11 months tells you about the server world competition. This is a CloudDC SuperServer SYS-621C-TN12R which is a perfect chassis for mixing different kinds of storage including PCIe Gen 5 storage. It is a 2U chassis designed to support up to 350W TDP CPUs. This server can be configured with two double width or six single width GPUs. Watch the video from LevelOneTechs for more insightful information.
Supermicro has long offered GPU servers in more shapes and sizes than we have time to discuss in this review. Today, we’re looking at their relatively new 4U air-cooled GPU server that supports two AMD EPYC 9004 Series CPUs, PCIe Gen5, and a choice of eight double-width or 12 single-width add-in GPU cards.
This is Siena-based server AS-1115SV-WTNRT. Siena has 96 PCI Gen 5 lanes and it is amazing how much connectivity Supermicro managed to fit into this 1U server. It has 40 PCIe Gen 5 NVMe lanes for hot swap storage with two of the MCIO headers being hybrid to support to support SATA or SAS disks as well. This means no matter what disks you fill in front of the server, you will have more than enough bandwidth for line rate connectivity. Most impressive bit of I/O is actually around the rear where there are three full x16 PCIe Gen 5 slots.
Today we wanted to take a look at the liquid cooled Supermicro SYS-821GE-TNHR server. This is Supermicro’s 8x NVIDIA H100 system with a twist: it is liquid cooled for lower cooling costs and power consumption. Since we had the photos, we figured we would put this into a piece.
[YouTube] — We take a look at a state-of-the-art liquid cooled NVIDIA H100 8x GPU server from Supermicro. We show why modern AI servers need liquid cooling. We also show some of the other cool bits from the company's liquid cooling lab including 2U 4-node servers with Intel Xeon and AMD EPYC liquid cooling blocks.
We take a look at Supermicro's custom liquid cooling rack, and specifically what it is doing around liquid cooling distribution for AI and HPC
If you take nothing else from this review, this should be the image that sticks in your mind when thinking about the Supermicro SYS-551A-T. There are class-leading 16 DDR5 DIMM slots, liquid cooling for the 56-core Xeon W CPU, six PCIe Gen5 x16 slots, and four M.2 slots for an onboard storage array.
AS-2125HS-TNR is an AMD based storage-oriented platform that focuses on the compute. Going from Milan to Genoa, we moved onto PCIe Gen5 and DDR5 memory, so we doubled the amount of PCIe bandwidth that we have. We also have 128 lanes between the two CPU sockets. The fourth high speed GMI link between the sockets is much appreciated. The system has a simple but very effective mechanism for routing cooling the CPUs. DDR5 memory; however, is using a bit more power than the last generation of DDR4.
With the Intel Xeon Max testing at Phoronix that's been ongoing so far for the past month on Phoronix has all been done within the Supermicro Hyper SuperServer SYS-221H-TNR rackmount server.