Inference chipmaker d-Matrix moves into NVIDIA's rack architecture
Inference chipmaker d-Matrix will use NVIDIA NVLink Fusion and the MGX rack architecture to deploy its next-generation Raptor XPUs at scale.
AI inference chipmaker d-Matrix announced it will use NVIDIA NVLink Fusion to connect its next-generation Raptor XPUs to NVIDIA's AI infrastructure platform, gaining access to NVLink scale-up networking, Spectrum-X scale-out networking and the NVIDIA MGX rack architecture.
Speaking at a press briefing, d-Matrix cofounder and CEO Sid Sheth said: "Demand for inference is soaring, but capital, time and energy remain finite. With NVLink Fusion and MGX, we can integrate our Raptor XPUs into a broadly deployed, liquid-cooled architecture, giving customers a faster, lower-risk path to deploy and scale ultralow-latency inference."
The numbers
According to NVIDIA, sixth-generation NVLink delivers 3x lower XPU-to-XPU latency than off-the-shelf Ethernet, 10x higher packet rates and 3 TB/s of all-to-all bandwidth per XPU.
d-Matrix plans to link its XPUs in a single high-bandwidth, low-latency scale-up domain. Its racks can also run alongside NVIDIA GPU systems such as Vera Rubin NVL72 for disaggregated inference. The company intends to integrate NVIDIA Vera CPUs, ConnectX-9 SuperNICs, BlueField-4 DPUs and Spectrum-X Ethernet networking as well.
Standardizing on a common rack means a data center can be built once and support GPUs, CPUs and XPUs without a separate rack design for each processor type.
A growing partner list
Alongside d-Matrix, the NVLink Fusion ecosystem includes AWS, Arm, Intel, Fujitsu, SiFive, Alchip, Astera Labs, GUC, Marvell, MediaTek, Samsung, Cadence, Synopsys, Ayar Labs and Lightmatter. The platform supports Arm, x86 and RISC-V architectures alongside NVIDIA GPUs.
Full details are in the NVIDIA blog post.
