Geekbench AI Results

From my observations, GPUs like the RTX 4070 SUPER and the MacBook Pro M3 Max’s GPU excel in AI tasks such as Style Transfer, Object Detection, and Pose Estimation due to their strong parallel processing. The MacBook’s Neural Engine offers a good balance of speed and efficiency, outperforming CPUs in some tasks. While CPUs like the Ryzen 7 5800X handle lighter workloads well, they lag behind in more demanding jobs. I’ve found that quantized models perform consistently well across all systems, making them ideal for faster processing with minimal loss in accuracy. For complex AI tasks, I recommend prioritizing GPUs.

1. Ryzen 7 5800X + RTX 4070 Super | ONNX | GPU Benchmarks

This benchmark evaluates the performance of the NVIDIA GeForce RTX 4070 SUPER using the ONNX framework and DirectML backend, focusing on AI task efficiency with single precision, half precision, and quantized models.

Benchmark Scores

The AI benchmarks were run using the ONNX framework on a GPU backend. Below are the Geekbench AI scores:

  • Single Precision: 37,950

  • Half Precision: 19,608

  • Quantized: 25,495

System Specs

  • Estimated Total Price: ~$1,310

Top AI Tasks

AI Task

Half Precision Score

AI Task

Half Precision Score

1. Pose Estimation

476.0 IPS

4. Depth Estimation

863.3 IPS

2. Style Transfer

880.8 IPS

5. Face Detection

481.6 IPS

3. Image Super-Resolution

2.17 KIPS

6. Object Detection

1.38 KIPS

2. MacBook Pro M3 Max | Core ML | Neural Engine Benchmarks

This benchmark evaluates the performance of the Apple M3 Max using the Core ML framework on the Neural Engine backend, focusing on AI task efficiency with single precision, half precision, and quantized models.

Benchmark Scores

The AI benchmarks were run using the Core ML framework on a Neural Engine backend. Below are the Geekbench AI scores:

  • Single Precision: 13,410

  • Half Precision: 13,896

  • Quantized: 2,457

System Specs

  • Estimated Total Price: ~$4,199

Top AI Tasks

AI Task

Half Precision Score

AI Task

Half Precision Score

1. Pose Estimation

120.1 IPS

4. Depth Estimation

460.6 IPS

2. Style Transfer

191.7 IPS

5. Image Super-Resolution

712.5 IPS

3. Object Detection

755.7 IPS

6. Face Detection

150.6 IPS

3. MacBook Pro M3 Max | Core ML | GPU Benchmarks

This benchmark evaluates the performance of the Apple M3 Max using the Core ML framework on the GPU backend, focusing on AI task efficiency with single precision, half precision, and quantized models.

Benchmark Scores

The AI benchmarks were run using the Core ML framework on a GPU backend. Below are the Geekbench AI scores:

  • Single Precision: 13,358

  • Half Precision: 12,570

  • Quantized: 13,411

System Specs

  • Estimated Total Price: ~$4,199

Top AI Tasks

AI Task

Half Precision Score

AI Task

Half Precision Score

1. Style Transfer

203.2 IPS

4. Depth Estimation

261.7 IPS

2. Face Detection

313.7 IPS

5. Image Super-Resolution

817.9 IPS

3. Object Detection

557.2 IPS

6. Pose Estimation

81.0 IPS

4. Ryzen 7 5800X | OpenVINO | CPU Benchmarks

This benchmark evaluates the performance of the Ryzen 7 5800X CPU using the OpenVINO framework, focusing on AI task efficiency with quantized models.

Benchmark Scores

The AI benchmarks were run using the OpenVINO framework on a CPU backend. Below are the Geekbench AI scores:

  • Single Precision: 5,019

  • Half Precision: 7,593

  • Quantized: 15,148

System Specs

  • Estimated Total Price: ~$710

Top AI Tasks

AI Task

Quantized Score

AI Task

Quantized Score

1. Style Transfer

53.4 IPS

4. Depth Estimation

125.1 IPS

2. Face Detection

180.6 IPS

5. Object Detection

432.3 IPS

3. Image Super-Resolution

299.5 IPS

6. Pose Estimation

17.1 IPS

5. MacBook Pro M3 Max | Core ML | CPU Benchmarks

This benchmark evaluates the performance of the Apple M3 Max using the Core ML framework on a CPU backend, focusing on AI task efficiency with single precision, half precision, and quantized models.

Benchmark Scores

The AI benchmarks were run using the Core ML framework on a CPU backend. Below are the Geekbench AI scores:

  • Single Precision: 5,356

  • Half Precision: 4,472

  • Quantized: 3,841

System Specs

  • Estimated Total Price: ~$4,199

Top AI Tasks

AI Task

Half Precision Score

AI Task

Half Precision Score

1. Style Transfer

37.9 IPS

4. Depth Estimation

75.5 IPS

2. Pose Estimation

19.3 IPS

5. Object Detection

205.0 IPS

3. Text Classification

6.11 KIPS

6. Image Super-Resolution

181.3 IPS

6. Ryzen 7 5800X | ONNX | CPU Benchmarks

This benchmark evaluates the performance of the Ryzen 7 5800X CPU using the ONNX framework on a CPU backend, focusing on AI task efficiency with single precision, half precision, and quantized models.

Benchmark Scores

The AI benchmarks were run using the ONNX framework on a CPU backend. Below are the Geekbench AI scores:

  • Single Precision: 884

  • Half Precision: 3,313

  • Quantized: 3,582

System Specs

  • Estimated Total Price: ~$710

Top AI Tasks

AI Task

Quantized Score

AI Task

Quantized Score

1. Pose Estimation

15.2 IPS

4. Depth Estimation

44.5 IPS

2. Image Classification

728.1 IPS

5. Face Detection

42.2 IPS

3. Image Super-Resolution

116.6 IPS

6. Object Detection

175.9 IPS

Last updated