CARDINAL
EXACT VARIANT REFERENCE

B200 SXM6 180GB

SXM accelerator reference for qualified Blackwell AI server platforms

SPECIFICATIONS

Memory per accelerator180 GB HBM3e
Memory bandwidth8000 GB/s
Form factorSXM6 module
Maximum accelerator power1000 W

Manufacturer specifications describe the product, not the condition of an offered unit. Checked 2026-09-08

PLATFORM AND SOFTWARE

  • 180GB per GPU is explicitly stated in the reference architecture; this is not a 192GB GB200 record
  • Requires a qualified SXM6 baseboard, cooling and server integration
  • HGX baseboard and DGX complete system are separate purchasing scopes
  • Exact module mechanical dimensions are not established in these sources
  • Verify Blackwell architecture support in the exact CUDA, framework and application build
  • Memory capacity does not establish model fit, precision support or workload throughput
  • Platform firmware, software support and commercial entitlements require separate qualification

Manufacturer explicitly specifies configurable up to 1000W per individual GPU; platform limits may differ

INTERCONNECT

NVLink

Fifth-generation NVLink and NVSwitch at HGX platform level; no per-GPU rate derived from platform totals

MANUFACTURER SOURCES

NVIDIA HGX reference architecture: GPU and node specifications ↗NVIDIA · Checked 2026-09-08NVIDIA HGX B200 PCF summary specifications ↗NVIDIA · Checked 2026-09-08

HOW CARDINAL HELPS

Share your quantity, destination and server model, if known. We help turn those requirements into a clearly specified request.

  • Help confirm the exact configuration and platform requirements
  • Request supplier records for condition, unit identifiers and available testing evidence
  • Clarify included accessories, delivery and proposed warranty terms before you decide

Any additional testing or guarantees must be agreed in the written offer.