Fungible has announced Fungible GPU-Connect (FGC) to solve the challenges caused by the growing demand for AI/ML that results in stranded and underutilized GPUs, especially in Edge Data Centers. The FGC product addresses how data processing power is accessed in enterprise and service provider infrastructures.
Fungible has announced Fungible GPU-Connect (FGC) to solve the challenges caused by the growing demand for AI/ML that results in stranded and underutilized GPUs, especially in Edge Data Centers. The FGC product addresses how data processing power is accessed in enterprise and service provider infrastructures.
Fungible GPU-Connect
The Fungible GPU-Connect solution includes the FX-108 GPU-Expansion Chassis, the FC200 Accelerator Card, and Fungible Composer Software. The FX-108 GPU-Expansion chassis is a 4-U, rackmount chassis, supporting up to eight double-width/single-width PCIe 3.0/4.0 x16 cards. Optional GPU-GPU Interconnect support via NVIDIA NVLink Bridge. Current validated GPUs include NVIDIA A2, A10, A16, A30, A40, A100-40G, A100-80G.
Leveraging the Fungible DPU, FGC dynamically composes GPU and CPU resources across an Ethernet network to provide economic benefits and agility to organizations that provision, allocate, and manage those expensive GPU resources. With FGC, organizations can create a GPU-powered infrastructure free from the physical limitations of PCIe.
Fungible DPU
The Fungible DPU delivers 800Gbps processing, implementing the entire storage, networking, security, and virtualization stack. The F1 DPU also facilitates highly efficient data interchange among server nodes through its TrueFabric technology. This enables disaggregation and pooling of all data center resources at scale.
TrueFabric is a large-scale IP-over-Ethernet fabric protocol that provides full cross-sectional bandwidth with low average and tail latency, end-to-end QoS, and congestion-free connectivity. The TrueFabric protocol is fully standards-compliant and interoperable with TCP/IP over Ethernet. This ensures that the data center spine-leaf network can be built with standard off-the-shelf Ethernet switches.
Fungible FC200 DPU
Organizations that deploy FGC in their data centers can centralize existing GPU assets in a single resource pool to be attached to servers on-demand instead of dedicated GPUs sitting idle. The disaggregated solution provided by the FGC product removes the constraints of having GPUs and CPUs physically co-located. These resources can be located anywhere within the same data center.
The Fungible DPU creates a secure, virtual PCIe connection between the GPU and the server that is transparent to the server and applications requiring no additional software or drivers. This connection is managed in hardware by the DPU, ensuring a high-performance and low latency connection. The transparency means FGC can be easily retrofitted into existing environments and can scale with growing demand, ensuring GPUs are always available when they are needed.
According to explains Toby Owen, VP of Product at Fungible;
“Fungible GPU-Connect empowers service providers to combine all their GPUs into one central resource pool serving all their clients. Service providers can now onboard new data-analysis intensive clients quickly without adding expensive servers and GPUs. By leveraging FGC, data centers can benefit from the collective computing power of all their GPUs and substantially lower their TCO with the reduction of GPU resources, cooling and physical footprint needed.”
FGC Benefits for Service Providers
Service providers deploying FGC differentiate themselves by creating net-new revenue opportunities offering customers more flexible plans and the specialized performance of GPU servers that are in high demand.
Eric Burgener, Vice President of Research, Infrastructure Systems, Platforms and Technologies Group, IDC, said:
“GPUs are needed for service providers who are increasingly getting into AI/ML processing. Composable infrastructure supports the agility they need to successfully monetize these expensive resources. Products like Fungible GPU-Connect enable service providers to cost-effectively make GPUs available to their customers, rapidly provisioning requested resources to meet customer requirements on-demand while easily sharing these expensive resources across multiple customers to drive a higher return on investment.”
To learn more about the new Fungible GPU-Connect, check out the product brief, read the blog, or contact [email protected].
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | Facebook | TikTok | RSS Feed
