Fujitsu PRIMERGY GX2570 M8s (Liquid-Cooled)
The liquid-cooled member of Fujitsu's PRIMERGY GX2570 M8s pair, and the same NVIDIA HGX B200 8-GPU baseboard as the air-cooled model compressed from 10U into 4U. Eight B200 SXM modules at 180 GB apiece are interconnected by NVSwitch for 1.44 TB of GPU memory, hosted by two Intel Xeon 6900P processors with 24 DIMM slots taking up to 6 TB. Liquid cooling costs two of the ten NVMe bays, leaving eight, and the chassis measures 449 by 895.4 by 174 mm at up to 80 kg with rails, which is 53 kg lighter than the air-cooled build. Power comes from four 6,600 W Titanium supplies in 2+2 redundancy. Fujitsu notes that a CDU, manifolds and a liquid-cooling rack must be purchased separately. No throughput figure and no system power draw are published; baseboard performance is on the linked HGX B200 entry.
We will point you at suppliers who have it. Free, and no signup.
Systems built on this design
9 configurationsBuilt on the NVIDIA HGX B200 8-GPU baseboard, so accelerator performance matches the reference. Chassis, CPU, memory and cooling are this vendor's own.
System Details
GPU Configuration
System Specifications
Precision Performance Breakdown
| Precision | System Performance | Per GPU | Efficiency |
|---|---|---|---|
| 296 TFLOPS | 37 TFLOPS | — | |
| 600 TFLOPS | 75 TFLOPS | — | |
| 9,000 TFLOPS | 1,125 TFLOPS | — | |
| 18,000 TFLOPS | 2,250 TFLOPS | — | |
| 18,000 TFLOPS | 2,250 TFLOPS | — | |
| 36,000 TFLOPS | 4,500 TFLOPS | — | |
| 72,000 TFLOPS | 9,000 TFLOPS | — | |
| 36,000 TOPS | 4,500 TOPS | — |
Where a vendor states a basis, the figure shown is the dense one, and any figure it publishes is listed separately. Vendors commonly headline the sparse number instead: for NVIDIA's 2:4 structured sparsity that is exactly twice the dense figure, though other vendors' sparse modes do not all follow that ratio.
Powered by NVIDIA B200
This system utilizes 8 × NVIDIA B200 SXM 180GB GPUs, each delivering exceptional performance for AI and HPC workloads.
Per GPU TDP
1000W
Per GPU Memory
180 GB
Process Node
4nm
Architecture
Blackwell
Frequently asked questions
How much does the PRIMERGY GX2570 M8s (Liquid-Cooled) cost?
Fujitsu sells the PRIMERGY GX2570 M8s (Liquid-Cooled) on quote through OEMs and resellers, and nobody rents a whole one by the hour. Renting its 8 B200 accelerators separately at the cheapest live rate Flopper tracks, $4.09 per GPU-hour across 15 providers, comes to about $32.72 an hour for the equivalent capacity. B200 rental prices →
How much power does the PRIMERGY GX2570 M8s (Liquid-Cooled) require?
Fujitsu has not published a power figure for the PRIMERGY GX2570 M8s (Liquid-Cooled), and Flopper does not derive one from the accelerator count because the rest of the system draws power too.
How many GPUs are in the PRIMERGY GX2570 M8s (Liquid-Cooled)?
8. The PRIMERGY GX2570 M8s (Liquid-Cooled) carries 8 NVIDIA B200 SXM 180GB accelerators in a rack form factor, and every throughput figure on this page is the total across them. B200 specifications →
Documentation & Resources
Need this system?
Tell us the quantity, timeframe and region. We reply by email with who can supply it.
Flopper Spec Sheet
Fujitsu PRIMERGY GX2570 M8s (Liquid-Cooled) specifications, generated from our database. Printable.
Vendor product page
Fujitsu PRIMERGY GX2570 M8s (Liquid-Cooled) documentation
PRIMERGY GX2570 M8s カタログ (Fsas Technologies)
Fujitsu • Latest version
Typical Use Cases
The Fujitsu PRIMERGY GX2570 M8s (Liquid-Cooled) runs 8× B200 GPUs over One NVIDIA HGX B200 8-GPU baseboard with NVSwitch interconnecting the eight B200 SXM modules; two 10GbE RJ45 ports on an Intel X710, with Ethernet and InfiniBand cards optional, delivering 36,000 TFLOPS FP8 dense.