The Supermicro SuperServer 4029GP-TRT2 is a GPU-enhanced 4U system with 24x HD bays up front and support for two Intel Xeon Scalable processors, plus up to 6TB of memory. The real impressive aspect of this system is support for up to 10x high-performance computational accelerators or graphics processing units. There are three other chassis in this family.
Supermicro’s 4U SuperServer line are all about machine learning, artificial intelligence, big data analytics and other high-performance computing needs. The Supermicro SuperServer 4029GP-TRT2 server supports up to 10x double-width Nvidia GPUs or computational accelerators. The difference between this chassis and the other three, the 4029GP-TRT, 4029GP-TRT3, and 4029GP-TVRT all come down to the daughter board and the number of GPUs supported. Three of the systems, including this one, use a standard GPU to CPU interface through the PCIe lanes. The last in that set the 4029GP-TVRT, uses Nvidia’s NVLINK circuit board for 7 to 12x more bandwidth and a transfer rate of 300GB/s plus, compared to 32GB/s using the standard GPU connection, like on this system. This single root system offers excellent GPU peer to peer communication with sufficient CPU to GPU throughput.
Two Intel Xeon Scalable processors can be installed on this system with up to 28 cores and a TDP of 70 – 205W. Both Gen 1 and Gen 2 Intel Xeon Scalable processors are supported on this system but you will get support for more memory and faster memory speeds using the Gen 2 processors. Gen 2 processors will support up to 4.5TB of memory each using processors with an “L” suffix and offer speeds of up to 2933MT/s. Gen 1 processors max out at 1.5TB each using processors with an “M” designation and support memory speeds of up to 2633MT/s. Both generations support six memory channels. Four PSUs for 2+2 redundancy are supported.
As stated in the Performance section above memory speeds and capacity are dependent on the CPU and memory speeds supported by the modules themselves. There are 12 memory module slots per CPU with six slots to either side of the processor, and 24x active in a 2-processor configuration. Registered (RDIMM), Load-Reduced (LRDIMM), and Intel Optane Date Centric Persistent Memory Modules (DCCPM) are also supported on this system and will provide up to 6TB using the appropriate processors and memory modules.
24x 2.5-inch storage bays on the front of the Supermicro 4029GP-TRT2 support up to 24x SAS and SATA drives. Another configuration supports up to 20x SAS or SATA drives plus up to 4x NVMe drives. The system will support up to eight SATA drives and 2x NVMe drives using the universal backplane with direct connections to the motherboard. For SAS drives and to support two more NVMe drives, you will need an optional HD/RAID controller. An M.2 slot located below the daughter board, can be used to support the OS, and a micro SD card slot is located just underneath where the M.2 drive would be positioned and can be used to support a hypervisor.
Two PLX modules on the Daughter Board control the data flow to enable up to 11x PCIe 3.0 x16 slots and one x16 mechanical slot with a x8 electrical lane. 10x of those slots are used for double-wide Nvidia graphics cards including Tesla and Quadro GPUs. The other two slots can be used for enhanced network communications or HD/RAID controllers to support more than 8x SATA drives or for SAS implementations. The system features 2x 10GbE ports on the back of the system plus a dedicated management port to access the IPMI.
Management of the system is through the Intelligent Platform Management Interface (IPMI) leverages the ASPEED AST 2500 controller for at chassis and remote management of the server. Super Doctor5 provides an intuitive and easy to read graphical interface showing health status plus hardware assets. Tell-tale LED lights on the control panel ear on the front of the system also provide at a glance status on HDDs, PSUs, LAN ports and others.
The Supermicro SuperServer 4029GP-TRT2 with support for up to 10x GPUs or computational accelerators is definitely designed for specific applications in mind. Those applications include astrophysics, business intelligence, artificial intelligence, deep learning, machine learning and other high-performance computing workloads. With up to 24x storage devices up front and support for up to 6TB of memory, there’s plenty of room to scale as your business needs change.
If you know what you want but can't find the exact configuration you're looking for, have one of our knowledgeable sales staff contact you. Give us a list of the components you would like to incorporate into the system, and the quantities, if more than one. We will get back to you immediately with an official quote.