GPU Wiki Architecture

Ampere

Architecture notes pending source review.

Launch era
Pending source review.
Source review
Pending

Related GPUs

118 entries
View
Reset filters
GPU Architecture Launch Date CUDA Cores Memory Bus Width TDP/TBP Segment
CMP 90HX Ampere 2021-02-18 6,400 10 GB GDDR6X 320-bit 320 W Desktop
CMP 70HX Ampere 2021-01-01 3,840 8 GB GDDR6X 256-bit Desktop
A800 SXM4 80 GB Ampere 2022-08-11 6,912 80 GB HBM2e 5,120-bit 400 W Data Center
GeForce RTX 3060 3840SP Ampere 2021-01-01 3,840 6 GB GDDR6 192-bit 185 W Desktop
A800 PCIe 40 GB Ampere 2022-11-08 6,912 40 GB HBM2e 5,120-bit 250 W Data Center
A800 PCIe 80 GB Ampere 2022-11-08 6,912 80 GB HBM2e 5,120-bit 250 W Data Center
A10M Ampere 2022-02-01 7,168 20 GB GDDR6 320-bit 150 W Data Center
PG506-207 Ampere 2021-04-12 3,584 24 GB HBM2 3,072-bit 165 W Desktop
PG506-242 Ampere 2021-04-12 3,584 24 GB HBM2 3,072-bit 165 W Desktop
PG506-217 Ampere 2021-04-12 3,584 24 GB HBM2 3,072-bit 165 W Desktop

More architecture facts

Architecture name
Ampere
Predecessor
Not yet documented.
Successor
Not yet documented.
Process nodes
Not yet documented.
Important chips
Not yet documented.
Memory technologies
Not yet documented.
CUDA / RT / Tensor generation
Not yet documented.
Consumer series
DRIVE, GeForce MX Series, GeForce RTX 20 Series, GeForce RTX 30 Series, GeForce RTX 40 Series, GRID, Mining GPUs, Nintendo Switch GPU, NVIDIA A-Series, NVIDIA RTX A Series, Tegra
Workstation / professional series
Not yet documented.
Data center series
Not yet documented.