Supermicro today started to note the market’s very first industrial servers based upon Intel’s Data Center GPU Max ‘Ponte Vecchio’ calculate GPUs The devices utilize Ponte Vecchio in add-in-board and OAM module kind elements, targeting at high-performance computing and massive AI training.
Supermicro presently has 2 servers received Intel’s Ponte Vecchio calculate GPUs: the SYS-421GE-TNRT device that can house as much as 10 Information Center GPU Max 1100 cards with 480GB of HBM2 memory (48GB per board) in addition to the SYS-821PV-TNR that can accommodate as much as 8 Information Center GPU Max 1550 OAM modules with 1TB of HBM2 memory onboard (128GB per card) and combined efficiency of 6.7 BF16/FP16 PetaFLOPS at 4.8 kW.
Both devices are based upon 2 of Intel’s 4 th Generation Xeon Scalable ‘Sapphire Rapids’ processors that can be mated with as much as 8TB of DDR5 memory utilizing 32 256GB modules. When it comes to storage, both devices have 24 2.5-inch hot-swap bays for U. 2/SATA/SAS drives (8x 2.5-inch NVMe hybrid; 8x 2.5-inch NVMe devoted), and the SYS-421GE-TNRT likewise has 2 M. 2 slots for PCIe drives.
In the meantime, Supermicro offers the SYS-421GE-TNRT with Nvidia’s A100 80GB or H100 80GB boards, however it appears like if asked for, it can set up Intel’s Data Center GPU Max 1100 AIBs rather: the device is totally certified to run them, so tossing them in and setting up needed software application needs to not take too long.
On The Other Hand, the SYS-821PV-TNR that can house as much as 8 Information Center GPU Max 1550 OAM modules is noted as ‘coming quickly.’
Supermicro is presently the only provider of industrial devices for AI and HPC work that uses systems geared up with Intel’s Data Center GPU Max cards and modules. Nevertheless, it is affordable to anticipate other leading providers of servers to begin offering comparable items soon.
Although Supermicro uses Intel Ponte Vecchio-based servers, it does not divulge their costs, as it constantly occurs with AI and HPC devices set up separately and offered in amounts.
Source: Supermicro ( @SquashBionic)