The cut Strix Halo part: 32 compute units against 40 and an 8-core Zen 5 CPU against 16, at the same 55W default TDP, the same 45W to 120W configurable range and, importantly, the SAME 256-bit LPDDR5X memory controller. Memory capacity and bandwidth are therefore identical to the Ryzen AI Max+ 395 at 128 GB and 256 GB/s, which is the specification that decides what will run. AMD publishes the same 50 NPU TOPS and a lower 106 Overall TOPS, and NEITHER IS A GPU FIGURE, so neither is stored as a precision row. As with the 395, AMD publishes no FP32, FP16, BF16 or INT8 rate and no AI Accelerator count for the Radeon 8050S, so the single FP32 figure here is derived and this row is vendor_claimed rather than official.

AMD Ryzen AI Max 385 Radeon 8050S (Strix Halo)

RDNA 3.5 SoC 2025 4nm Spec confidence: Vendor claimed
FP32
23
TFLOPS
VRAM
128 GB
LPDDR5X
Bandwidth
256 GB/s
memory
TDP
55 W
417.1 TFLOPS/kW
Download datasheet

Overview

The Ryzen AI Max 385 is the affordable way into Strix Halo, and for local model work it gives up less than the specification sheet suggests. It keeps the whole point of the platform, the 256-bit LPDDR5X controller with 128 GB of unified memory at 256 GB/s, and cuts the parts that matter less: 32 RDNA 3.5 compute units instead of 40, and an 8-core Zen 5 CPU instead of 16. Since inference on these machines is bound by memory bandwidth rather than shader throughput, the roughly 23 percent gap in graphics compute is smaller in practice than it looks. The 50 TOPS XDNA 2 neural processor is identical to the flagship's. As with every Strix Halo part, AMD publishes no graphics throughput figure, so the FP32 number here is computed on AMD's own arithmetic rather than quoted from a datasheet.

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
No verified data available
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.

Specifications

Architecture

RDNA 3.5

Form Factor

SoC

Launch Year

2025

Process Node

4nm

Memory

128 GB LPDDR5X

Bandwidth

256 GB/s

TDP

55 W

Max Power

120 W

Stream Processors

2,048

Spec Confidence

Vendor claimed

Full Specifications

Compute Engine
Stream Processors 2,048
Compute Units 32
Boost Clock 2.80 GHz
Memory
VRAM 128 GB
Memory Type LPDDR5X
Bandwidth 256 GB/s
Interface Width 256-bit
Memory Clock 8 GT/s
Power & Thermal
TDP 55 W
Max Board Power 120 W
Enterprise Features
Sparsity No
General
Form Factor SoC
Architecture RDNA 3.5
Process Node 4nm
Launch Year 2025

Datasheet & Resources

Data Provenance

Every figure traced to a source

Primary Source

Publisher
AMD
Published
No verified data available

Data Quality

Spec confidence
Vendor claimed
Clock basis
Boost
Core precisions with figures
1 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 Ryzen AI Max 385 have?

The AMD Ryzen AI Max 385 delivers 23 TFLOPS FP32 at peak. Flopper does not currently have verified FP16 and FP8 throughput figures for it.

What is the power consumption of the AMD Ryzen AI Max 385?

The AMD Ryzen AI Max 385 has a TDP (Thermal Design Power) rating of 55 watts.

How much memory does the AMD Ryzen AI Max 385 have?

The AMD Ryzen AI Max 385 is equipped with 128 GB of VRAM with 256 GB/s of memory bandwidth.

What architecture is the AMD Ryzen AI Max 385 based on?

The AMD Ryzen AI Max 385 is based on the RDNA 3.5 architecture, launched in 2025.

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