Aguasto / gpus

The GPU Library.

Accelerators and rack systems, with the boundaries that make comparisons useful. Manufacturer specifications, configuration differences and deployment considerations—not a list of available Aguasto inventory.

Original Aguasto conceptual illustration. Not manufacturer geometry, an operating Aguasto asset, or a measured benchmark.

12 of 12 references

Individual accelerator

NVIDIA H100 SXM

Hopper. 80 GB published memory. A documented commercial SXM accelerator; this record does not establish supplier stock or an Aguasto allocation.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

NVIDIA H200 SXM

Hopper. 141 GB published memory. The manufacturer product page explicitly says available. No supplier stock or Aguasto deployment is verified.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

NVIDIA B200 SXM

Blackwell. 180 GB published memory. NVIDIA's HGX page states HGX B200 is shipping. This record describes one SXM GPU, not the baseboard.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

NVIDIA B300

Blackwell Ultra. Configuration-dependent memory. NVIDIA states HGX B300 is shipping. A single generic capacity is withheld because current manufacturer documents describe different configurations.

Manufacturer reference · reviewed Oct 12, 2026
Rack system

NVIDIA GB200 NVL72

Grace Blackwell. 13,400 GB published memory. NVIDIA reports GB200 NVL72 shipments began in 2024. Availability of a specific OEM build remains supplier-specific.

Manufacturer reference · reviewed Oct 12, 2026
Rack system

NVIDIA GB300 NVL72

Grace Blackwell Ultra. 20,000 GB published memory. NVIDIA reports GB300 NVL72 shipped in 2025; a specific OEM configuration and delivery date still require confirmation.

Manufacturer reference · reviewed Oct 12, 2026
Rack system

NVIDIA Vera Rubin NVL72

Vera CPU / Rubin GPU. 20,700 GB published memory. Manufacturer disclosures describe systems shipping and production racks delivered by September 2026. This establishes a shipping ramp, not universal availability or an Aguasto allocation.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

AMD Instinct MI300X

CDNA 3. 192 GB published memory. Launched accelerator with manufacturer-reported deployment/benchmark evidence; supplier stock and Aguasto availability are not verified.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

AMD Instinct MI325X

CDNA 3. 256 GB published memory. Launched accelerator with manufacturer-reported deployment/benchmark evidence; supplier stock and Aguasto availability are not verified.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

AMD Instinct MI355X

CDNA 4. 288 GB published memory. Launched accelerator with manufacturer-reported deployment/benchmark evidence; supplier stock and Aguasto availability are not verified.

Manufacturer reference · reviewed Oct 12, 2026
Individual accelerator

AMD Instinct MI455X

CDNA 5. 432 GB published memory. AMD lists a July 2026 launch and published specifications. Reviewed Helios sources still describe partner volume deployments as expected in the second half of 2026; broad shipment availability is not independently established here.

Manufacturer reference · reviewed Oct 12, 2026
Rack system

AMD Helios

MI455X / EPYC Venice / Pensando. 31,000 GB published memory. AMD explicitly identifies Helios as a reference design rather than an AMD product for sale. The reviewed page expects OEM/ODM volume deployments in the second half of 2026.

Manufacturer reference · reviewed Oct 12, 2026