Dear Friends,
Our recent events have been cancelled because of the COVID-19 pandemic. To keep our valued customers on track with TYAN’s latest product information, we are going to conduct an online virtual exhibition in the late June timeframe. Event schedule will be released soon, please contact your Tyan sales representatives for more details.
We’d like to introduce our featured product -
Thunder HX FT83-B7119 on this month's newsletter.
FT83-B7119 is a new 4U GPU server which accommodates most of the
FA77-B7119's boards but transforms the system storage from 14 top-load 2.5" drives to 12 front-access 3.5" drive bays.
Furthermore, the 4U server deploys 12 new 3.5" drive trays with a tool-less design that enormously reduces required service time on drive installation or replacement. Same as the
FA77-B7119, the new
FT83-B7119 accommodates up to 10 double-wide or 20 single-wide GPU cards with an additional single-wide PCIe x16 slot for a high speed networking card, such as 100 Gigabit EDR InfiniBand or 100 Gigabit Ethernet deployment. The mechanical design of the chassis allows passive GPU cards to be properly cooled and operate smoothly. When deploying full-sized active GPU cards, users of the FT83-B7119 can purchase a specialized domed lid for the top mounted power connectors of active GPU cards to fit within the chassis. Furthermore, the Tyan
FT83-B7119 features a new storage mezzanine carrier for deploying a SAS IOC or NVMe mezzanine cards as storage options. Customers can run any combination of 12x 3.5" SATA drives with up to 8 SAS or 4 NVMe drives without consuming any expansion slots. Finally, featuring AMI MegaRAC® firmware, the
FT83-B7119 server barebone offers support for IPMI 2.0 with iKVM and the DMTF Redfish management protocol.
Thunder HX FT83-B7119: 4U High-Density HPC GPU Server
The Thunder HX FT83-B7119 platform has 4 SKUs. The "-N" SKUs are configured with all of the GPU cards evenly connected to two CPUs and provides high CPU-to-GPU bandwidth for various parallel workloads, while the "-NS" SKUs are configured with all 10 GPU cards connected to one CPU, which can lower latency when moving data between cards in applications such as artificial intelligence, machine learning, and deep neural networks.
|
SKU |
Storage |
CPU-GPU Link Topology |
Pre-installed
LAN Mezz. |
Pre-installed
NVMe Card |
| B7119F83V8E4HR-2T-NS | (8) 3.5" SATA 6G Hot-swap Bays + (4) NVMe U.2 | All GPUs connected to CPU0 | (1) M7108-X550-2T | - |
| B7119F83V12HR-2T-NS | (12) 3.5" SATA 6G Hot-swap Bays | All GPUs connected to CPU0 | (1) M7108-X550-2T | - |
| B7119F83V8E4HR-2T-N | (8) 3.5" SATA 6G Hot-swap Bays + (4) NVMe U.2 | 5 GPUs connected to each CPU | (1) M7108-X550-2T | (1) M7106-4E |
| B7119F83V12HR-2T-N | (12) 3.5" SATA 6G Hot-swap Bays | 5 GPUs connected to each CPU | (1) M7108-X550-2T | - |
- Dual Intel P0 socket
- Support 2nd Gen Intel® Xeon® Scalable Processor with TDP up to 205W
- Intel® C621 PCH + (2) Broadcom PEX8796 PCIe switches
- (12+12) DDR4 DIMM slots
- Support DDR4-2933/2666 RDIMM/LRDIMM up to 3,072GB
- PCIe expansion slots:
- (11) FH/10.5"L PCIe Gen.3 x16 + (10) multiplexed PCIe Gen.3 x8 slots support
- (10) double-width GPU cards, or
- (20) single-width GPU cards @ PCIe x8 bandwidth
- (1) PCIe Gen.3 x16 storage mezz. slot
- Storage (B7119F83V12HR-2T-N / -NS):
- (12) hot-swap, tool-less 3.5" drive bays support (12) SATA 6G devices (screws are required for 2.5" device installation)
- Storage (B7119F83V8E4HR-2T-N / -NS):
- (12) hot-swap, tool-less 3.5" drive bays support (8) SATA 6G + (4) NVMe U.2 devices (screws are required for 2.5" device installation)
- Network:
- 10GBase-T LAN + (1) 1000Base-T dedicated IPMI ports
- AST2500 BMC with IPMI & Redfish support
- (3+1) 4,800W RPSU, 80 plus Platinum
|
