RDNA 2 has no matrix cores. AMD's 46.08 TFLOPS is packed FP16 vector math at exactly twice the FP32 vector rate, not a matrix or tensor rate. AMD publishes no INT8, INT4, BF16, FP8 or FP64 number for this card, and its current product page publishes no clock frequency of any kind.

AMD Radeon RX 6900 XT 16GB

RDNA 2 PCIe 2020 7nm Spec confidence: Official
FP32
23
TFLOPS
VRAM
16 GB
GDDR6
Bandwidth
512 GB/s
memory
TDP
300 W
76.8 TFLOPS/kW
Download datasheet

Overview

The Radeon RX 6900 XT was AMD's RDNA 2 flagship at launch in late 2020, a full 80 compute unit Navi 21 with 16 GB of GDDR6 and 128 MB of Infinity Cache holding a 256-bit bus at 512 GB/s. It is the same die and the same core count as the later RX 6950 XT, which differs only in memory speed and power budget. RDNA 2 has no matrix cores, so its 23.04 TFLOPS FP32 and 46.08 TFLOPS packed FP16 are the whole compute story: there is no BF16, FP8, INT8 or INT4 rate published, and nothing on the die that would make one interesting. The card matters to this catalog mainly as a 16 GB local inference option and as the reference point for how far a vector-only architecture sits below a tensor-core one.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
32-bit floating point
23TFLOPS
TF32
No verified data available
BF16
No verified data available
FP16
16-bit floating point
46TFLOPS
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
No verified data available

Every figure here is dense. The vendor documents no structured (2:4) sparsity mode for this part, so there is no second number to quote.

The Radeon RX 6900 XT 16GB in the GPU landscape

Peak FP16 TFLOPS (dense) against TDP, single-GPU parts tracked by Flopper

06001,2001,8002,4003,0000 W250 W500 W750 W1000 W1250 W1500 WInstinct MI355XRadeon RX 6900 XT 16GB

Higher and further left is better: more half-precision throughput for less power.

Specifications

Architecture

RDNA 2

Form Factor

PCIe

Launch Year

2020

Process Node

7nm

Memory

16 GB GDDR6

Bandwidth

512 GB/s

TDP

300 W

Max Power

~345 W

Transistors

26.8 billion

Stream Processors

5,120

Spec Confidence

Official

Full Specifications

Compute Engine
Stream Processors 5,120
Compute Units 80
Chip Design
Transistors 26.8 billion
Process Node 7nm
Chiplets 1 (Monolithic)
Memory
VRAM 16 GB
Memory Type GDDR6
Bandwidth 512 GB/s
Interface Width 256-bit
Memory Clock 16 GT/s
Cache
Infinity Cache 128 MB
Power & Thermal
TDP 300 W
Max Board Power The vendor publishes no maximum board power; the Max Power tile above estimates it from TDP
Power Connector 2x8-Pin
Enterprise Features
Sparsity No
Compute APIs DirectX 12, OpenGL, OpenCL, Vulkan
Physical & Media
Card Length 267 mm
Width 2.5-slot
General
Form Factor PCIe
Architecture RDNA 2
Process Node 7nm
Launch Year 2020

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
AMD
Published
No verified data available

Data Quality

Spec confidence
Official
Clock basis
Boost
Core precisions with figures
2 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

Frequently Asked Questions

How many TFLOPS does the AMD Radeon RX 6900 XT have?

The AMD Radeon RX 6900 XT delivers 23 TFLOPS FP32 and 46 TFLOPS FP16 at peak. Flopper does not currently have a verified FP8 throughput figure for it.

What is the power consumption of the AMD Radeon RX 6900 XT?

The AMD Radeon RX 6900 XT has a TDP (Thermal Design Power) rating of 300 watts.

How much memory does the AMD Radeon RX 6900 XT have?

The AMD Radeon RX 6900 XT is equipped with 16 GB of VRAM with 512 GB/s of memory bandwidth.

What architecture is the AMD Radeon RX 6900 XT based on?

The AMD Radeon RX 6900 XT is based on the RDNA 2 architecture, launched in 2020.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs