AMD Instinct MI325X vs AMD Radeon RX 9070 Full Specs - HMC

19,456 Shaders 2.1GHz	3,584 Shaders 2.52GHz
288GB HBM3e6.01TB/s	16GB GDDR6640GB/s
·· 163.4 TFLOPS	·· 36.13 TFLOPS
Form Factor OAM Module	Form Factor PCIe Card
TDP 560W	TDP 220W
Power Connectors -	Power Connectors 2x 8-Pin

Instinct MI325X

Instinct MI325X5.23 POPSINT4 Tensor Sparse

x4.52

Radeon RX 9070

Radeon RX 90701.16 POPSINT4 Tensor Sparse

x1

Clock Speed ···	Clock Speed ···
Peak OPS 5.23 POPSINT4 Tensor Sparse	Peak OPS 1.16 POPSINT4 Tensor Sparse
Tensor FP8-16 2.62 PFLOPS FP8-16 Tensor Sparse 5.23 PFLOPS Tensor FP8-32 2.62 PFLOPS FP8-32 Tensor Sparse 5.23 PFLOPS	Tensor FP8-16 144.5 TFLOPS FP8-16 Tensor Sparse 289 TFLOPS Tensor FP8-32 144.5 TFLOPS FP8-32 Tensor Sparse 289 TFLOPS
Tensor FP16-16 1.31 PFLOPS FP16-16 Tensor Sparse 2.62 PFLOPS Tensor FP16-32 1.31 PFLOPS FP16-32 Tensor Sparse 2.62 PFLOPS	Tensor FP16-16 72.25 TFLOPS FP16-16 Tensor Sparse 144.5 TFLOPS Tensor FP16-32 72.25 TFLOPS FP16-32 Tensor Sparse 144.5 TFLOPS
BF16 - Tensor BF16 1.31 PFLOPS BF16 Tensor Sparse 2.62 PFLOPS	BF16 72.25 TFLOPS Tensor BF16 72.25 TFLOPS BF16 Tensor Sparse 144.5 TFLOPS
Tensor TF32 653.7 TFLOPS	Tensor TF32 -
FP32 163.4 TFLOPS Tensor FP32 163.4 TFLOPS	FP32 36.13 TFLOPS Tensor FP32 -
FP64 81.72 TFLOPS Tensor FP64 163.4 TFLOPS	FP64 1.13 TFLOPS Tensor FP64 -
Tensor INT4 2.62 POPS	Tensor INT4 578 TOPS
Tensor INT8 2.62 POPS	Tensor INT8 289 TOPS
Pixel Rate -	Pixel Rate 322.6 GPixel/s
Texture Rate 2.55 TTexel/s	Texture Rate 564.5 GTexel/s

Shaders 19,456 Shaders	Shaders 3,584 Shaders
TMUs 1,216 TMUs	TMUs 224 TMUs
ROPs -	ROPs 128 ROPs
Tensor Cores 1,216 T-Cores	Tensor Cores 112 T-Cores
RT Cores -	RT Cores 56 RT-Cores
CUs 304 CUs	CUs 56 CUs

Base Clock 1GHz	Base Clock 2.21GHz
Boost Clock 2.1GHz	Boost Clock 2.52GHz

L0 Cache -	L0 Cache 64KB/WGP
L1 16KB/CU	L1 256KB/Array
L2 Cache 16MB shared	L2 Cache 8MB shared
L3 Cache 256MB shared	L3 Cache 64MB shared
L3 Bandwidth -	L3 Bandwidth 2.25TB/s

288GB HBM3e	16GB GDDR6
Memory Bus 8192-bit	Memory Bus 256-bit
Memory Speed 5.9GT/s	Memory Speed 20GT/s
Memory Bandwidth 6.01TB/s	Memory Bandwidth 640GB/s
ECC Yes	ECC No

TDP

560W

TDP

220W

Max Resolution -	Max Resolution 7680x4320
Refresh Rate Calculator -	Refresh Rate Calculator 7680x4320 Resolution8K UHD Refresh Rate 165Hz
Multi-Monitor -	Multi-Monitor 3
Adaptive Sync -	Adaptive Sync FreeSync
DSC Not Supported	DSC Supported
HDCP -	HDCP HDCP 2.3

-	3x DisplayPort 2.1 1x HDMI 2.1

Shader Model -	Shader Model 6.8
-	DirectX DirectX 12 Direct3D 12_2
OpenGL - OpenCL 3 Vulkan -	OpenGL 4.6 OpenCL 2.2 Vulkan 1.3
GFX 9.4	GFX 12

Encoder 2x VCN 2.6	Encoder VCN 4.0
Codec - AVC (H.264) HEVC (H.265) -	Codec VP9 AVC (H.264) HEVC (H.265) AV1

Decoder 2x VCN 2.6	Decoder VCN 4.0
Codec -	Codec MPEG-1 MPEG-2 MPEG-4 JPEG VC-1 VP9 AVC (H.264) HEVC (H.265) AV1

Form Factor OAM Module	Form Factor PCIe Card
PCIe PCIe 5.0 x16 -	PCIe PCIe 5.0 x16 2.5-Slots
-	Height 111mm (4.37") Width 267mm (10.51") Depth 50mm (1.97")
Cooling Passive -	Cooling Open-Air 3x Fans
Power Connectors -	Power Connectors 2x 8-Pin
Multi-GPU Multi-GPU Support Supported Multi-GPU Type Infinity Fabric	Multi-GPU Multi-GPU Support Supported Multi-GPU Type CrossFire XDMA

Manufacturer AMD	Manufacturer AMD
Chip Designer AMD	Chip Designer AMD
Architecture CDNA 3	Architecture RDNA 4
Family Instinct	Family Radeon RX 9000
Branding	Branding
Codename Aqua Vanjaram	Codename -
Chip Variant Aqua Vanjaram XTX	Chip Variant Navi 48 XL
Market Segment Server	Market Segment Desktop
Release Date Jun 2, 2024	Release Date Jan 6, 2025

Foundry TSMC	Foundry TSMC
Other Foundries TSMCActive Interposer	Other Foundries -
Fabrication Node N5 N6Active Interposer	Fabrication Node N4P -
Die Size 1480mm²	Die Size 357mm²
Transistor Count 146B	Transistor Count 53.9B
Transistor Density 98.65 MTr/mm²	Transistor Density 151.2 MTr/mm²