Radeon RX 9070 XT

AMDconsumer
Overview
ArchitectureRDNA 4 (2025)
Launch date2025-03-06
Transistor count53.9B
GPU target (gfx)gfx1201
Memory
VRAM16 GB GDDR6
Memory bandwidth640 GB/s
Cache64 MB
Compute — vector
FP3248.7 TFLOPS
FP1648.7 TFLOPS
Compute — matrix / tensor
FP16 (dense / sparse)195 / 389 TFLOPS
FP8 (dense / sparse)389 / 779 TFLOPS
INT8 (dense / sparse)389 / 779 TOPS
INT4779 TOPS
Cores & clocks
Compute Units64
Shader cores4,096
Matrix / Tensor cores128
Peak clock2,970 MHz
Board & system
TDP304 W
PCIePCIe 5.0 x16
InterconnectPCIe only (no Infinity Fabric multi-GPU link)
MSRP$599

Compatible ROCm versions

6.4.06.4.16.4.26.4.37.0.07.0.17.0.27.1.07.1.17.2.07.2.17.2.27.2.37.2.47.14.010.0.0
Re-verified directly against AMD's own product page (the correct URL is amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html, no hyphen before "xt"). fp16TFLOPS (195, dense) and fp8TFLOPS (389, dense) are both directly published ("Peak Half Precision (FP16 Matrix) Performance: 195 TFLOPs", 389 with structured sparsity; "Peak 8-bit Precision (FP8 Matrix) Performance: 389 TFLOPs", 779 with sparsity) — RDNA4 is AMD's first Radeon architecture with native FP8 matrix support. int8TOPS/int8TOPSSparse mirror the FP8 figures exactly (389/779); int4TOPS listed but its sparsity variant (1,557 TOPs) isn't captured in a separate schema field. matrixCoreCount uses the page's "AI Accelerators" count (128). tdpWatts (304W, "Typical Board Power") and memoryBandwidthGBs (640) match this file's prior corroborated figures exactly. No Launch Date field is published on this page; launchDate (2025-03-06) was added 2026-09-20 from AMD's own RDNA 4 launch press release of 2025-02-28, which states availability beginning March 6, 2025 for both this card and the RX 9070 (amd.com/en/newsroom/press-releases/2025-2-28-amd-unveils-next-generation-amd-rdna-4-architectu.html). Source: amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html ("Expand All" spec accordion), pulled 2026-09-01.

← All GPUs · Compare GPUs →