VOOZH

URL: https://www.phoronix.com/review/amd-epyc-4345p/3

⇱ AMD EPYC 4345P 8-Core CPU Performance Review - Phoronix

👁 Phoronix

Articles & Reviews
News Archive
Forums
Premium
Contact
Categories

Computers Display Drivers Graphics Cards Linux Gaming Memory Motherboards Processors Software Storage Operating Systems Peripherals

AMD EPYC 4345P 8-Core CPU Performance

Written by Michael Larabel in Processors on 9 June 2025 at 11:00 AM EDT. Page 3 of 8. 11 Comments.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Llama-3.1-Tulu-3-8B-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

For those interested in edge AI inferencing or similar low-power AI server deployments, the Xeon 6300 series really struggles to compete and comes in well behind even the prior generation EPYC 4004 series. The EPYC 4005 series benefits from the faster DDR5-5600 memory, full 512-bit data path AVX-512, and the other Zen 5 architectural advantages to really help with CPU-based AI inferencing. For Llama.cpp with Llama 3.1 8B, the EPYC 4345P delivered 1.73x the performance of the Xeon 6369P, 2x the performance per Watt, and 3.2x the performance per dollar of the Xeon 6369P flagship.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 512. EPYC 4565P was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 512. EPYC 4565P was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 512. EPYC 4565P was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: granite-3.0-3b-a800m-instruct-Q8_0, Test: Prompt Processing 512. EPYC 4565P was the fastest.

With the Granite 3B model tested with Llama.cpp, the EPYC 4345P provided 1.47x the performance of the Xeon 6369P at 1.91x the performance per Watt and 2.69x the performance-per-dollar.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

👁 Llama.cpp benchmark with settings of Backend: CPU BLAS, Model: Mistral-7B-Instruct-v0.3-Q8_0, Test: Prompt Processing 512. EPYC 4585PX was the fastest.

The AMD EPYC 4345P with a 65 Watt TDP really shines if wanting a low-power edge AI inference platform while also being economical.

👁 Whisper.cpp benchmark with settings of Model: ggml-medium.en, Input: 2016 State of the Union. EPYC 4585PX was the fastest.

👁 Whisper.cpp benchmark with settings of Model: ggml-medium.en, Input: 2016 State of the Union. EPYC 4585PX was the fastest.

👁 Whisper.cpp benchmark with settings of Model: ggml-medium.en, Input: 2016 State of the Union. EPYC 4585PX was the fastest.

👁 Whisper.cpp benchmark with settings of Model: ggml-medium.en, Input: 2016 State of the Union. EPYC 4585PX was the fastest.

👁 PyTorch benchmark with settings of Device: CPU, Batch Size: 16, Model: ResNet-50. EPYC 4585PX was the fastest.

👁 PyTorch benchmark with settings of Device: CPU, Batch Size: 16, Model: ResNet-50. EPYC 4585PX was the fastest.

👁 PyTorch benchmark with settings of Device: CPU, Batch Size: 16, Model: ResNet-50. EPYC 4585PX was the fastest.

👁 PyTorch benchmark with settings of Device: CPU, Batch Size: 16, Model: ResNet-50. EPYC 4585PX was the fastest.

With other AI benchmarks as well the EPYC 4345P was far outpacing the Xeon 6369P flagship and delivering very sizable gains over the prior generation EPYC 4344P.

11 Comments - Next Page

Page: 1 2 3 4 5 6 7 8 Next Page

RISC-V CPU Performance Up 8x In Five Years: SiFive HiFive Unmatched To SpacemiT K3

ARM Linux Server Performance Up More Than 7x Geo Mean In 8 Years, As Much As 15x With NVIDIA Vera CPU

AMD EPYC 8635P "Sorano" Benchmarks: Significant Upgrade Opportunity For EPYC 8004 Servers

Intel Xeon 6+ & Intel Ethernet E835 Launch

AMD Announces Radeon RX 9070 GRE, Ryzen AI Max PRO 400 Series

NVIDIA Vera CPU Benchmarks: Olympus Cores Delivering The Best Performance Ever Seen On ARM

Support Phoronix
While Having Ad-Free Browsing,
Single-Page Article Viewing

Legal Disclaimer, Privacy Policy, Cookies | | Contact
Copyright © 2004 - 2026 by Phoronix Media.
All trademarks used are properties of their respective owners. All rights reserved.