GeekRadar / The comparison desk

COMPARE.

GPU

NVIDIA

NVIDIA RTX PRO 6000 Blackwell Server

Graphics processor

Video memory: 96 GB^*
Memory bus: 512 bit^*
Rated memory bandwidth: 1,600 GB/s^*

Benchmark results

LLM Inference: Geekbench AI: Geekbench AI Half Precision (fp16) · version unreported · score53,322^* pointsPlatform unreported Compare results
LLM Inference: Geekbench AI: Geekbench AI Quantized (int8) · version unreported · score28,264^* pointsPlatform unreported Compare results
LLM Inference: Geekbench AI: Geekbench AI Single Precision (fp32) · version unreported · score37,299^* pointsPlatform unreported Compare results
LLM Inference: MLPerf: llama2-70b-99 (fp4) · version unreported · score3,250.74^* tokens/sPlatform unreported Compare results
LLM Inference: MLPerf: llama2-70b-99.9 (fp4) · version unreported · score3,250.13^* tokens/sPlatform unreported Compare results
LLM Inference: MLPerf: llama3.1-8b (fp4) · version unreported · score5,757.50^* tokens/sPlatform unreported Compare results
LLM Inference: MLPerf: mixtral-8x7b (fp8) · version unreported · score3,766.79^* tokens/sPlatform unreported Compare results

Processing & platform

Architecture: Blackwell 2.0^*
Process node: 5 nm^*
GPU cores: 24,064^*
Graphics interface: PCIe 5.0 x16^*
Reported chip die area: 750 mm²^*
Transistors: 92,200 million^*
GPU L2 cache: 128 MB^*
Graphics processor: GB202^*
Texture mapping units: 752^*
Raster output units: 192^*
Streaming multiprocessors: 188^*
Tensor cores: 752^*
Ray tracing cores: 188^*
Chip foundry: TSMC^*
Card or module format: Dual-slot^*
Declared GPU TDP: 600 W^*
Suggested system PSU: 1,000 W^*
External power connectors: 1x 16-pin^*
DirectX / feature level: 12 Ultimate (12_2)^*
OpenGL: 4.6^*
OpenCL: 3.0^*
Vulkan: 1.4^*
CUDA compute capability: 12.0^*
Shader model: 6.9^*

Memory & storage

Memory options: 96^*
Video memory: 96 GB^*
Video memory type: GDDR7^*
Memory bus: 512 bit^*
Rated memory bandwidth: 1,600 GB/s^*

Power specifications

Thermal design power: 600 W^*

System configuration

Detailed specifications

Specifications 1 · 11 fields

GPU architecture — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: Blackwell^*
Theoretical peak: FP32 non-tensor dense per GPU — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: 120^* TFLOPS
Physical configuration — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: PCIe, FHFL, dual slot, passive (4.4 in H x 10.5 in L)^*
Native hardware support for GPU interconnect — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: No^*
PCIe interface declaration — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: PCIe Gen 5 x16^*
Maximum board power — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: 600^* W
Peak memory bandwidth per GPU — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: 1,597^* GB/s
Memory interface width per GPU — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: 512^* bit
Memory capacity per GPU — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: 96^* GB
Memory technology — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: GDDR7^*
GPU model — NVIDIA RTX PRO 6000 Blackwell Server Edition — 96 GB GDDR7: NVIDIA RTX PRO 6000 Blackwell Server Edition^*

Specifications 2 · 52 fields

proc transistors

92,200 million^*

proc die size

750^* mm²

cores

24,064^*

tensor cores

752^*

rt cores

188^*

sm count

188^*

base clock

1,590^*

boost clock

2,617^*

memory clock

1,750^*

ram

96^*

name

NVIDIA RTX PRO 6000 Blackwell Server^*

test name Graphics benchmark result

LLM Inference: MLPerf: llama2-70b-99 (fp4)^*

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: Geekbench AI: Geekbench AI Half Precision (fp16)^*

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: Geekbench AI: Geekbench AI Single Precision (fp32)^*

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: llama3.1-8b (fp4)^*

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 (fp4)^*

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: Geekbench AI: Geekbench AI Quantized (int8)^*

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: mixtral-8x7b (fp8)^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

ram type

GDDR7^*

bus width

512^*

bus interface

PCIe 5.0 x16^*

ram bandwidth

1.79 TB/s^*

fp16 performance

126^*

fp32 performance

126^*

fp64 performance

1.968^*

score Graphics benchmark result

37,299^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

3,250.125^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

53,322^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

28,264^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

5,757.5^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

3,766.7875^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

score Graphics benchmark result

3,250.7375^*

Scope and conditions

benchmark_context

suite: Per-row third-party test name^*
test_configuration_verified: No^*
comparable: No^*

tdp

600^* W

suggested psu

1,000^* W

slot width

Dual-slot^*

power connectors

1x 16-pin^*

api cuda

12^*

is retail board

No^*

gpi value

57.47^*