GeekRadar / The comparison desk
COMPARE.
GPURadeon Instinct MI355X
Graphics processor
- Video memory
- 288 GB*
- Memory bus
- 8,192 bit*
- Rated memory bandwidth
- 8,190.00 GB/s*
Benchmark results
- LLM Inference: MLPerf: llama2-70b-99 (mxfp4) · version unreported · score10,128.88* tokens/sPlatform unreported Compare results
- LLM Inference: MLPerf: llama2-70b-99.9 (mxfp4) · version unreported · score10,128.88* tokens/sPlatform unreported Compare results
- LLM Inference: MLPerf: llama3.1-405b (mxfp4) · version unreported · score252.38* tokens/sPlatform unreported Compare results
Processing & platform
- Architecture
- CDNA 3.0*
- Process node
- Reported value 1: 5 nm*Reported value 2: 3 nm*Reported value 3: 3 nm*
- GPU cores
- 16,384*
- GPU base clock
- 1,000 MHz*
- Graphics interface
- PCIe 5.0 x16*
- Reported chip die area
- 2,380 mm²*
- Transistors
- 185,000 million*
- GPU boost clock
- 2,400 MHz*
- GPU L2 cache
- 32 MB*
- Graphics processor
- MI350*
- Texture mapping units
- 1,024*
- Raster output units
- 0*
- Chip foundry
- TSMC*
- Card or module format
- OAM Module*
- Declared GPU TDP
- 1,400 W*
- Suggested system PSU
- 1,800 W*
- External power connectors
- None*
- OpenCL
- 3.0*
Memory & storage
- Memory options
- 288*
- Video memory
- 288 GB*
- Video memory type
- HBM3e*
- Memory bus
- 8,192 bit*
- Rated memory bandwidth
- 8,190.00 GB/s*
Power specifications
- Thermal design power
- 1,400 W*
System configuration
Detailed specifications
Specifications 1 · 22 fields
- GPU architecture — AMD Instinct MI355X — 288 GB HBM3E
- CDNA 4*
- Theoretical peak: BF16 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 2,516.6* TFLOPS
- Theoretical peak: FP16 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 2,516.6* TFLOPS
- Theoretical peak: FP32 non-tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 157.3* TFLOPS
- Theoretical peak: FP4 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 10,066.3* TFLOPS
- Theoretical peak: FP64 non-tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 78.6* TFLOPS
- Theoretical peak: FP64 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 78.6* TFLOPS
- Theoretical peak: FP6 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 10,066.3* TFLOPS
- Theoretical peak: FP8 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 5,033.2* TFLOPS
- Theoretical peak: INT8 tensor dense per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 5,033.2* TOPS
- Native hardware support: TF32 tensor — AMD Instinct MI355X — 288 GB HBM3E
- No*
- Physical configuration — AMD Instinct MI355X — 288 GB HBM3E
- OAM*
- GPU interconnect bandwidth (Infinity Fabric) bidirectional — AMD Instinct MI355X — 288 GB HBM3E
- 1,075.2* GB/s
- GPU interconnect bandwidth per direction — AMD Instinct MI355X — 288 GB HBM3E
- 537.6* GB/s
- Infinity Fabric link count — AMD Instinct MI355X — 288 GB HBM3E
- 7* links
- PCIe interface declaration — AMD Instinct MI355X — 288 GB HBM3E
- PCIe Gen 5 x16*
- Maximum board power — AMD Instinct MI355X — 288 GB HBM3E
- 1,400* W
- Peak memory bandwidth per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 8,000* GB/s
- Memory interface width per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 8,192* bit
- Memory capacity per GPU — AMD Instinct MI355X — 288 GB HBM3E
- 288* GB
- Memory technology — AMD Instinct MI355X — 288 GB HBM3E
- HBM3E*
- GPU model — AMD Instinct MI355X — 288 GB HBM3E
- AMD Instinct MI355X*
Specifications 2 · 33 fields
- proc transistors
- 185,000 million*
- cores
- 16,384*
- base clock
- 1,000*
- boost clock
- 2,400*
- memory clock
- 2,000*
- ram
- 288*
- name
- AMD Radeon Instinct MI355X*
- test name Graphics benchmark result
- LLM Inference: MLPerf: llama2-70b-99 (mxfp4)*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- test name Graphics benchmark result
- LLM Inference: MLPerf: llama3.1-405b (mxfp4)*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- test name Graphics benchmark result
- LLM Inference: MLPerf: llama2-70b-99.9 (mxfp4)*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- ram type
- HBM3e*
- bus width
- 8,192*
- bus interface
- PCIe 5.0 x16*
- ram bandwidth
- 8.19 TB/s*
- fp16 performance
- 629.1*
- fp32 performance
- 78.64*
- fp64 performance
- 78.64*
- score Graphics benchmark result
- 10,128.875*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- score Graphics benchmark result
- 10,128.875*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- score Graphics benchmark result
- 252.38*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- tdp
- 1,400* W
- suggested psu
- 1,800* W
- slot width
- OAM Module*
- is retail board
- No*
- gpi value
- 93.52*
- unit Graphics benchmark result
- tokens/s*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- unit Graphics benchmark result
- tokens/s*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- unit Graphics benchmark result
- tokens/s*
Scope and conditions
- benchmark_context
- suite
- Per-row third-party test name*
- test_configuration_verified
- No*
- comparable
- No*
- vendor
- amd*
- manufacturer
- AMD*
- card release date
- 2025-06-12*
- proc foundry
- TSMC*
- proc process size
- 3* nm
Specifications 3 · 31 fields
- OpenCL
- 3.0*
- Graphics interface
- PCIe 5.0 x16*
- Memory bus
- 8,192* bit
- Reported shader processors
- 16,384*
- Reported theoretical FP16 peak
- 78.64* TFLOPS
Scope and conditions
- Arithmetic density
- Not stated*
- Reported theoretical FP32 peak
- 78.64* TFLOPS
Scope and conditions
- Arithmetic density
- Not stated*
- Reported theoretical FP64 peak
- 39.32* TFLOPS
Scope and conditions
- Arithmetic density
- Not stated*
- Declared L1 cache configuration
- 32 KB (per CU)*
- Declared GPU L2 cache
- 32 MB*
- Reported display outputs
- No outputs*
- Theoretical pixel fill rate
- 0 MPixel/s*
- External power connectors
- None*
- Reported chip die area
- 2380 mm²*
- Chip foundry
- TSMC*
- Process node declaration
- 3 nm*
- Transistors
- 185,000 million*
- Graphics processor
- MI350*
- Reported video memory
- 288* GB
- Rated memory bandwidth
- 8.19 TB/s*
- Video memory type
- HBM3e*
- Raster output units
- 0*
- Reported shading units
- 16,384*
- Card or module format
- OAM Module*
- Suggested system PSU
- 1800 W*
- 2:4 tensor sparsity
- No*
- Declared GPU TDP
- 1400 W*
- Theoretical texture fill rate
- 2,457.6 GTexel/s*
- Texture mapping units
- 1,024*
- GPU cores
- 16,384*
- Video memory
- 288* GB
- Video memory type
- HBM3e*
Specifications 4 · 49 fields
- Brand
- AMD*
- Top BUS WIDTH
- 8,192* bit
- Render Config MCM
- 2*
- Graphics Processor GPU Name
- Aqua Vanjaram*
- Render Config Matrix Cores
- 1,024*
- Graphics Processor Architecture
- CDNA 3.0*
- Graphics Processor Process Size
- 5* nm
- Graphics Processor Transistors
- 153,000 million*
- Graphics Processor Die Size
- 1,017* mm²
- Graphics Card Release Date
- 2025*
- Graphics Card Generation
- Radeon Instinct (MIx)*
- Graphics Card Production
- Unreleased*
- Name
- Radeon Instinct MI355X*
- Graphics Card Bus Interface
- PCIe 5.0 x16*
- Clock Speeds Memory Clock
- 2000 MHz 8 Gbps effective*
- Memory Memory Size
- 288* GB
- Memory Memory Type
- HBM3e*
- Memory Memory Bus
- 8,192* bit
- Memory Bandwidth
- 8.19 TB/s*
- Render Config TMUs
- 1,024*
- Top GRAPHICS PROCESSOR
- Aqua Vanjaram*
- Render Config ROPs
- 0*
- Theoretical Performance Pixel Rate
- 0 MPixel/s
- Theoretical Performance Texture Rate
- 2,457.6 GTexel/s
- Board Design Slot Width
- OAM Module*
- Board Design TDP
- 1,400* W
- Board Design Suggested PSU
- 1,800* W
- Board Design Outputs
- No outputs*
- Graphics Features OpenCL
- 3*
- source file
- amd2025.json*
- Graphics Processor Foundry
- TSMC*
- Graphics Card Predecessor
- FirePro Data Center*
- Graphics Processor Density
- 150.4M / mm²*
- Top TMUS
- 1,024*
- Graphics Processor Chip Package
- MCM*
- Clock Speeds Base Clock
- 1,000* MHz
- Clock Speeds Boost Clock
- 2,400* MHz
- Top CORES
- 16,384*
- Render Config Shading Units
- 16,384*
- Render Config Compute Units
- 256*
- Render Config L2 Cache
- 16* MB
- Top MEMORY SIZE
- 288* GB
- Theoretical Performance FP32 (float)
- 78.64* TFLOPS
- Theoretical Performance FP64 (double)
- 78.64 TFLOPS (1:1)
- Render Config L1 Cache
- 16 KB (per CU)*
- Graphics Card Announced
- June 11th, 2025*
- Theoretical Performance FP16 (half)
- 629.1 TFLOPS (8:1)
- Top MEMORY TYPE
- HBM3e*
- Render Config L3 Cache
- 256* MB
0 selected Compare 0
* Provisional data