Published method
AI accelerator memory specification methodology
1-datacenter-accelerators-v1-2026-07-26
Version
Jul 26, 2026
Published
1
Board
ai-accelerator-memory-specs
Method ID
Where the facts come from
- Source Policy
Only linked NVIDIA or AMD first-party product documentation and announcements; no reseller or inferred specifications.
How the order is produced
Memory capacity descending, then memory bandwidth descending, then stable entity slug.
When the method reruns
Each board shows the date of the snapshot it is built from.
Boards
Boards using this version
| Board | Primary measure | Rows | Confidence | Snapshot | Board version | Refresh | Board |
|---|---|---|---|---|---|---|---|
| AI accelerator memory specifications | accelerator-memory-capacity-gb | 5 | 100% | Jul 26, 2026 | 1-datacenter-accelerators-v1-2026-07-26 | Not stated | Open the board: AI accelerator memory specifications |
Versioned contract
Complete measurement ledger
Cohort
- Name
NVIDIA H100 SXM
- Vendor
NVIDIA
- Entity Slug
nvidia-h100-sxm
- Form Factor
SXM
- Memory Type
HBM3
- Source Locator
H100 SXM GPU specifications
- Memory Capacity Gb
80
- Memory Bandwidth Tb S
3.35
- Name
NVIDIA H200 SXM
- Vendor
NVIDIA
- Entity Slug
nvidia-h200-sxm
- Form Factor
SXM
- Memory Type
HBM3e
- Source Locator
H200 GPU specifications
- Memory Capacity Gb
141
- Memory Bandwidth Tb S
4.8
- Name
NVIDIA B200 SXM
- Vendor
NVIDIA
- Source Url
- https://docs.nvidia.com/enterprise-reference-architectures/hgx-ai-factory/latest/components.html ↗
- Entity Slug
nvidia-b200-sxm
- Form Factor
SXM
- Memory Type
HBM3e
- Source Locator
NVIDIA HGX B200 GPU specifications
- Memory Capacity Gb
180
- Memory Bandwidth Tb S
8
- Name
AMD Instinct MI300X
- Vendor
AMD
- Entity Slug
amd-instinct-mi300x
- Form Factor
OAM
- Memory Type
HBM3
- Source Locator
MI300X accelerator specifications
- Memory Capacity Gb
192
- Memory Bandwidth Tb S
5.3
- Name
AMD Instinct MI325X
- Vendor
AMD
- Entity Slug
amd-instinct-mi325x
- Form Factor
OAM
- Memory Type
HBM3e
- Source Locator
MI325X memory capacity and bandwidth announcement
- Memory Capacity Gb
256
- Memory Bandwidth Tb S
6
Selection
A fixed, versioned set of generally available datacenter AI accelerators with first-party capacity and bandwidth values.
Reviewed At
2026-07-26T00:00:00+00:00
Missing Data
A missing source or value rejects the complete reviewed cohort.
Not Measured
- benchmark performance
- workload throughput
- price-performance
- power efficiency
- availability
- utilization
- GPU hours consumed
Rank Formula
Memory capacity descending, then memory bandwidth descending, then stable entity slug.
Review Due At
2026-10-24T00:00:00+00:00
Source Policy
Only linked NVIDIA or AMD first-party product documentation and announcements; no reseller or inferred specifications.
Universe Version
datacenter-accelerators-v1-2026-07-26
See how often AI names you.
Enter your website. In about two minutes, Rank.ai asks ChatGPT, Claude and Gemini 12 questions your customers ask and grades how often they name you.
- Your grade out of 100How often AI names you, cites your site, and how high it ranks you.
- Who gets namedEvery competitor in the answers, most named first.
- The pages AI readsThe sources behind each answer.
- Three fixesWhat to fix first, with a brief for the first article.