CPU Benchmark Performance: AI and Inferencing

As technology progresses at a breakneck pace, so too do the demands of modern applications and workloads. With artificial intelligence (AI) and machine learning (ML) becoming increasingly intertwined with our daily computational tasks, it's paramount that our reviews evolve in tandem. Recognizing this, we have AI and inferencing benchmarks in our CPU test suite for 2024. 

Traditionally, CPU benchmarks have focused on various tasks, from arithmetic calculations to multimedia processing. However, with AI algorithms now driving features within some applications, from voice recognition to real-time data analysis, it's crucial to understand how modern processors handle these specific workloads. This is where our newly incorporated benchmarks come into play.

As chip makers such as AMD with Ryzen AI and Intel with their Meteor Lake mobile platform feature AI-driven hardware within the silicon, it seems in 2024, and we're going to see many applications using AI-based technologies coming to market.

We are using DDR5 memory on the Core i9-14900K, Core i7-14700K, Core i5-14600K, and Intel's 13th Gen at the relative JEDEC settings. The same methodology is also used for the AMD Ryzen 7000 series and Intel's 12th Gen (Alder Lake) processors. Below are the settings we have used for each platform:

  • DDR5-5600B CL46 - Intel 14th & 13th Gen
  • DDR5-5200 CL44 - Ryzen 7000
  • DDR5-4800 (B) CL40 - Intel 12th Gen

Note: As we are running with a completely refreshed CPU test suite, this means we are currently re-testing other processors for our data sets. These will be added to the below graphs as soon as we have more results, and these will also be added to our Bench database. Thanks for your understanding.

(6-1) ONNX Runtime 1.14: CaffeNet 12-int8 (CPU Only)

(6-1b) ONNX Runtime 1.14: CaffeNet 12-int8 (CPU Only)

(6-1c) ONNX Runtime 1.14: Super-Res-10 (CPU Only)

(6-1d) ONNX Runtime 1.14: Super-Res-10 (CPU Only)

(6-2) DeepSpeech 0.6: Acceleration CPU

(6-3) TensorFlow 2.12: VGG-16, Batch Size 16 (CPU)

(6-3b) TensorFlow 2.12: VGG-16, Batch Size 64 (CPU)

(6-3d) TensorFlow 2.12: GoogLeNet, Batch Size 16 (CPU)

(6-3e) TensorFlow 2.12: GoogLeNet, Batch Size 64 (CPU)

(6-3f) TensorFlow 2.12: GoogLeNet, Batch Size 256 (CPU)

(6-4) UL Procyon Windows AI Inference: MobileNet V3 (float32)

(6-4b) UL Procyon Windows AI Inference: ResNet 50 (float32)

(6-4c) UL Procyon Windows AI Inference: Inception V4 (float32)

Digesting performance in the latest addition to our CPU test suite for 2024, it's clear that the extra L3 cache on the Ryzen 7000X3D processors has a clear benefit in ONNX when using the INT8 model. Even in ONNX with Super-Res-10, the AMD Ryzen CPUs seem to do a better job of inferencing than Intel. It does appear that more L3 cache can benefit performance in this area, and even in TensorFlow with VGG-16, the AMD Ryzen 9 7950X seems to perform the best. Outside of that, the Core i9-14900K is next best, just marginally ahead of the Core i9-13900K. 

The Core i7-14700K performs respectively, and while the Core i5-14600K can certainly handle AI and inferencing, it doesn't offer quite the grunt of the Core i9 and Core i7 series chips.

CPU Benchmark Performance: Science And Simulation Gaming Performance: 720p And Lower
Comments Locked

57 Comments

View All Comments

  • Gastec - Friday, October 20, 2023 - link

    Maybe they do it through a proxy app, part of the overall package of Windows' Telemetry?
  • pookguy88 - Tuesday, October 17, 2023 - link

    you didn't have a 13700k to test against?
  • shabby - Tuesday, October 17, 2023 - link

    Yup pity, that would show us what those 4 e-cores can actually do.
  • TheinsanegamerN - Tuesday, October 17, 2023 - link

    I mean they still dont have a GPU test bed going on 3 years post fire. I wouldnt expect much.
  • nandnandnand - Wednesday, October 18, 2023 - link

    I recommend everyone go to Tom's Hardware if they are missing something here. They'll have the reviews, decent ones IMO, and are owned by the same company as AnandKek.
  • TheinsanegamerN - Thursday, October 19, 2023 - link

    Tom's Hardware was caught shilling for Nvidia eons ago. They're another dinosaur of the tech space.

    Techspot, Techpowerup, and reviewers like gamers nexus are the new hotness.
  • wrosecrans - Tuesday, October 17, 2023 - link

    Some motherboards will let you just set a power limit. I'd like to see a benchmark where the power limit is set to only the advertising number (125 W) and see what it can do with that constraint. 400+ watts just seems insane. My laptop is currently suffering terrible battery life because the CPU throttles up and gets hot and cooks the laptop because of exactly this Power Be Damned philosophy. I want a quiet desktop that isn't going to cook me if I'm sitting next to it, and isn't going to just cook the motherboard components and fail after a few years.

    I was expecting the new chip to be slightly more power efficient with a year of design tweaks and improvements. (And you'll note Intel wants you to think this because they kept the 125W marketing power usage on the box.) I am kinda baffled how Intel is executing so poorly. Nobody had a gun to their head forcing them to release this product. There's some deeply broken structural inertia in the organization to just keep pumping out products and not disrupting the flow of new model numbers. Somebody in Intel should have been screaming and said the plan wasn't working, rather than just keeping their head down to deliver a new model number for no reason.
  • TheinsanegamerN - Tuesday, October 17, 2023 - link

    If you want low power, get a ryzen. The 7800x3d tops out at just 50 watt.

    Performance loss, if anything like raptor lake (which this is) will be 15%+ down at 125 watt, more if they make heavy use of P cores.
  • schujj07 - Tuesday, October 17, 2023 - link

    Andandtech did this with the 13900k vs 7950X at different TDP/PPT. Basically the Ryzen at 65W TDP or 88W PPT was faster than Intel at 125W TDP. Once the Ryzen was set to 105W TDP or 142W PPT the Intel needed 253W TDP to be faster. In fact the scaling on the Ryzem dropped off quite quickly over 105W TDP.
  • mode_13h - Wednesday, October 18, 2023 - link

    This: https://www.anandtech.com/show/17641/lighter-touch...

Log in

Don't have an account? Sign up now