ASUS has announced they are showcasing its MGX Series AI supercomputing-ready solutions over at GTC 2024.
Although not exactly “The latest and greatest” Grace x Blackwell products, these NVIDIA MGX-powered 2U servers with model names ESC NM1-E1 and ESC NM2-E1, are still fairly competent through the GH200 Grace Hopper Superchip packed with Arm Neoverse V9 CPU cores with Scalable Vector Extensions (SVE2) and NVLink-C2C together BlueField-3 DPUs and ConnectX-7 network adapters that comes together to provide data throughput of up to 400Gbps. The software stack provided NVIDIA AI enterprise with a bunch of tools and platforms to build and deploy enterprise-grade AI apps further making these servers more attractive.
The brand also adopts a Direct-to-Chip (D2C) cooling solution that can be deployed quickly and reduces Power Usage Effectiveness (PUE) ratios quickly over time. Aside from ESC N8-E11 and RS720QN-E11-RS24U supporting manifolds and cooling plates, all ASUS servers can be fitted with a rear-door heat exchanger made to be compliant with standard rack-server designs for industry-standard deployment.
ASUS will also showcase the ESC4000A-E12 engaging in “No-Code AI” through an integrated software stack that can help any business to accelerate AI development such as LLM pre-training, fine-tuning, and inference with tokens count at 7B, 33B, and even 180B, all according to client needs. Everything will be configured to push for maximum efficiency and ROI too.
If you happen to attend the event, you can find them over at booth no. 730.
