November 18th 2019, Taipei Taiwan – For the past few years GIGABYTE has established an enviable reputation as an industry leader in high performance accelerated computing platforms, by continuously building new systems based that feature the highest GPU density available in the market, combined with excellent cooling performance, power efficiency and expansion flexibility. These systems are ideal for a range of different accelerated computing workloads in artificial intelligence and deep learning, scientific analysis and simulation, VDI or video streaming, and the excellent raw performance delivered by these products have resulted in them becoming the system of choice for some of the biggest HPC system integrators worldwide.
GIGABYTE is now pleased to welcome another seven new products to join our G-Series GPU server family, and continue setting the pace in the industry as a leader in GPU computing platforms:
New 2nd Generation AMD EPYC™ Based GPU Servers
GIGABYTE is proud add five more systems to our 2nd Generation AMD EPYC™ powered server lineup, bringing to a total of 28 systems that GIGABYTE has released so far this year since the release of 2nd Gen EPYC in August - including rack servers, edge servers, GPU servers and high density 2U 4 node servers - and firmly establishing GIGABYTE’s position as the system vendor with the most complete and diverse 2nd Gen EPYC system lineups on the market.
G292-Z22 / G292-Z42: Maximum GPU Density in the Smallest Footprint
The G291 Series has been one of GIGABYTE’s most popular GPU server platforms, with an ingenious design that allows for a maximum density of 8 GPGPU cards in a 2U chassis. The chassis design features four huge GPU bays, with two on each side of the server. Four large riser cards, which are easily inserted into or removed from each bay via lie flat handles, accommodate two dual-width GPGPU cards each, making the system able to accommodate eight dual-width GPGPU cards in total. This chassis design is not only perfectly optimized for GPU density but also cooling, as the bays on each side of the server become airflow tunnels with air pulled / pushed through from both front and rear located fans.
Our G292 series continues with this excellent design for the 2nd Gen EPYC processors. The new systems are available both in single socket (G292-Z22) and dual socket (G292-Z42) models for different workload requirements, and supports up to eight PCIe® Gen 3.0 GPU cards such as NVIDIA Tesla V100. Two additional models, G292-Z20 and G292-Z40 which offer native support for PCIe® Gen 4.0 GPU cards will be released early in 2020.
▲ G292-Z22 & G292-Z42: Maximum GPU Density in the Smallest Footprint
G482-Z50 / G482-Z51: Excellent GPU Density and Storage Capacity
GIGABYTE’s G482 Series is a 4U dual socket server platform that not only features dense GPU capacity but also impressive onboard storage space, allowing the user to build an integrated storage cluster within the same unit to store data sets for deep learning workloads etc.
The G482-Z50 can support ten PCIe Gen 3.0 GPGPU cards in balanced mode, with each five GPU cards connected to one CPU downstream via a PCIe switch to minimize GPU to GPU communication latency. The G482-Z50 also features 12 x 3.5” SATA / SAS drive bays and 10 x 2.5” NVMe / SATA / SAS SSD drive bays for ample storage capacity, and can be used to build an “AI machine in a box” that offers powerful compute, GPU compute, storage and networking capabilities.
Meanwhile, the G482-Z51 is one of the world’s first PCIe Gen 4.0 GPU server systems to be released to the market, featuring support for up to eight PCIe Gen 4.0 GPU cards (with each GPU connected directly to the processor via PCIe root). It is ideal to integrate an “AMD on AMD” solution using 2nd Gen EPYC processors together with the Radeon Instinct™ MI50, the world’s first 7nm, PCIe® Gen 4.0 data center GPU with a peak performance of almost 27 TFLOPS for FP16 deep learning workloads, or with other Gen 3.0 or Gen 4.0 GPU cards for applications such as cloud gaming or video streaming.
▲ G482-Z50 & G482-Z51: Excellent GPU Density and Storage Capacity
T181-Z70: “RACKLUTION-OP” GPU Server Node
RACKLUTION-OP is GIGABYTE’s server product line based on the Open Compute Project (OCP) Standards, designed for a 21” OCP rack and featuring a separate PSU system, with power supplied to each server node by a bus-bar system running along the rear of the rack. The Open Rack standards have designed to create more efficient, flexible, and scalable datacenter hardware, such as improving rack / node density, better power efficiency and easier maintenance by centralizing power supplies, and so on. RACKLUTION-OP products are also ideal for converged server applications or to build software defined data centers.
The T181-Z70 is a 1OU RACKLUTION-OP GPU computing node, featuring dual 1st or 2nd Gen AMD EPYC 7001 / 7002 Series processors, four PCIe® Gen 3.0 dual-width GPGPU bays in the rear, four frontal 2.5” hot swap SSD / HDD drive bays and two additional Gen 3.0 x16 low profile expansion slots for add-on networking or storage cards. It can be easily integrated together with GIGABYTE’s other RACKLUTION-OP products – compute nodes and trays, storage nodes and racks – to build your complete Open Rack data center infrastructure.
▲ T181-Z70: “RACKLUTION-OP” GPU Server Node
G241-G40 / G591-HS0: New Intel® Xeon® Scalable GPU Server Platforms
GIGABYTE is also proud to announce two new dual socket 2nd Generation Intel® Xeon® Scalable Processor family GPU server platforms:
The G241-G40 is a 2U system that features four PCIe® Gen 3.0 x16 expansion slot bays for GPUs, well as optimal storage capacity with four 3.5” hot swap SATA / SAS drive bays and two onboard M.2 slots. Since each GPU bay has an ample height of 1.5U and exceptional airflow characteristics, this system is extremely flexible to be used with a diverse range of active or passive cooled GPU cards.
Meanwhile, the G591-HS0 is a 5U system that can accommodate up to 40 low profile PCIe® Gen 3.0 x16 add-on cards via two expansion slot trays (20 per each tray) that slide out from the rear of the unit. Add on card options could include up to 32 single-width PCIe® GPU or FPGA cards to make the G591-HS0 a powerful inferencing engine for an intelligent video analytics platform or other applications.
▲ G241-G40 & G591-HS0: New Intel Xeon Scalable GPU Server Platforms
Easy & Powerful Management Tools
All GIGABYTE servers include GIGABYTE Management Console as standard, which is based on an AMI MegaRAC SP-X web-browser based platform and is compliant with the latest standards of Redfish API. GIGABYTE Management Console is also notable in providing HTML5-based iKVM remote management client as a standard feature, with no additional add-on license required to purchase. Also available as a free download is GIGABYTE Server Management (GSM), GIGABYTE’s multiple server remote management software platform, and includes both a desktop and mobile APP.
For those wishing to have a closer look at these new systems, they will be showcased this week by GIGABYTE at SC19, the world’s most important international conference for high performance computing. Please visit GIGABYTE at Booth 671, Denver Convention Center during November 18 to 21 in Denver, Colorado, or contact us directly at server.grp(at)gigabyte.com for more information.