Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1766
Article Thumbnail

Pix4D Performance: Intel Core 10th Gen vs AMD Ryzen 3rd Gen

Written on May 20, 2020 by William George
Share:

TL;DR: Intel Core 10th Gen vs AMD Ryzen 3rd Gen for Pix4Dmapper

Intel's new 10th Gen Core processors, especially the Core i9 10900K and i7 10700K, are faster per dollar than AMD's Ryzen 3rd Gen chips in Pix4D. However, that statement is somewhat misleading because there is a more expensive Ryzen model that outperforms them, and even higher-end processors which outpace both the Core and Ryzen lines. For the best performance in Pix4D, the AMD Threadripper 3970X and 3960X are clear winners.

Introduction

Intel has just launched its next generation of consumer desktop processor: the 10th Gen Intel Core family. Historically, this product line from Intel has offered fantastic clock speeds with moderate core counts, but recently Intel has been steadily increasing the number of cores included in these CPUs. There are quite a number of chips launching today, but to gauge what potential this series have we are going to look at the very top-end Core i9 10900K, Core i7 10700K, and Core i5 10600K models.

Pix4D Intel Core 10th Gen vs AMD Ryzen 3rd Gen

For this article, we are looking at how the new Intel 10th Gen Core processors compare to other currently-available models and the preceding 9th-gen Core series. The focus is on photogrammetry processing in Pix4D, for which we have developed internal benchmarks here in our lab. More info about those tools and how we use them are available in the Test Methodology section.

If you would prefer to skip over our test setup and various benchmark results, feel free to jump straight to the Conclusion.

Looking for a Pix4D Workstation?

Puget Systems offers a range of powerful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!

Test Setup

Listed below are the specifications of the systems we will be using for our testing:

Intel 10th Gen Test Platform
CPU Intel Core i9 10900K ($488)
Intel Core i7 10700K ($374)
Intel Core i5 10600K ($262)
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z490 Vision D
RAM 4x DDR4-2933 16GB (64GB total)
(Set to 2666MHz for 10600K testing)
AMD Ryzen Test Platform
CPU AMD Ryzen 9 3950X ($749)
AMD Ryzen 9 3900X ($499)
AMD Ryzen 7 3800X ($399)
AMD Ryzen 7 3700X ($329)
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte X570 AORUS ULTRA
RAM 4x DDR4-2933 16GB (64GB total)
Intel 9th Gen Test Platform
CPU Intel Core i9 9900K ($488)
Intel Core i7 9700K ($374)
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z390 Designare
RAM 4x DDR4-2666 16GB (64GB total)
AMD Threadripper 3rd Gen Test Platform
CPU AMD TR 3990X ($3,990)
AMD TR 3970X ($1,999)
AMD TR 3960X ($1,399)
CPU Cooler Noctua NH-U14S TR4-SP3
Motherboard Gigabyte TRX40 AORUS PRO WIFI
RAM 4x DDR4-2933 16GB (64GB total)
Intel X-10000 Series Test Platform
CPU Intel Core i9 10980XE ($979)
Intel Core i9 10940X ($784)
Intel Core i9 10920X ($689)
Intel Core i9 10900X ($590)
CPU Cooler Noctua NH-U12DX i4
Motherboard Gigabyte X299 Designare EX
RAM 4x DDR4-2933 16GB (64GB total)
Shared PC Hardware/Software
Video Card NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit (version 1909)
Pix4Dmapper Professional 4.5.6
Puget Systems Pix4D Benchmark

A Note about Power Limits

One thing that our product development team found in their testing, prior to this round of benchmarks in our lab, was that the pre-launch motherboards we received from Gigabyte (and some other manufacturers) were not defaulting to Intel's specified power limits in the BIOS settings. This is something we've seen in past launches as well, but now that Intel is being more aggressive about adding cores and pushing clock speeds higher we are seeing much higher power draw (and therefore temperatures) than we expect from a processor rated at 125W. In this generation, it has gotten to the point that temps can spike quickly - often getting as high as 100C in a matter of seconds under heavy loads. Because of this, we decided to manually set the PL1 and PL2 power limits in the BIOS. Following Intel's specifications, we used a value of 125W for the PL1 (long term) setting on all three Intel 10th Gen Core CPUs we tested, along with the following PL2 (short term) limits:

Core i9 10900K: 250W
Core i7 10700K: 229W
Core i5 10600K: 182W

Setting these power limits ensures that our Noctua NH-U12S is more than enough to keep these CPUs properly cooled and lines up with our philosophy about prioritizing stability and reliability over raw performance in our workstations. Please note, however, that not all reviewers share this perspective - and some may not even be aware of the issues at all! If you see variances between our benchmark results and those on other websites, different BIOS settings and default behaviors may well be the reason.

Test Methodology

For benchmarking photogrammetry applications we now have four image sets that we own the rights to, covering both smaller and larger size Model and Map projects. All of these image sets are available in our public Pix4D benchmarks, split up to allow quick or extended tests, which you can download and run if you want to compare your system's performance to what we measured for this article.

  • Rock Model - 82 photos at 20 megapixels each
  • School Map - 51 photos at 18 megapixels each
  • School Model - 278 photos at 18 megapixels each
  • Park Map - 810 photos at 18 megapixels each

Benchmark Results

There are a lot of processors in this round-up, so we are color-coding the results to make them easier to sift through:

  • Light blue with glow = New 10th Gen Intel Core CPUs
  • Light blue = Previous-generation Intel consumer CPUs (9th Gen Core)
  • Dark blue = Intel HEDT CPUs (Core X 10th Gen)
  • Light red = AMD consumer CPUs (Ryzen 3rd Gen)
  • Dark red = AMD HEDT CPUs (Threadripper 3rd Gen)

For those who want to dig into the weeds of how these CPUs performed in different processing steps, here is a full table of the results:

Pix4D 4.5.6 Processor Performance Table

Click to enlarge

Analysis & Conclusion

In Pix4D, the new 10th Gen Intel Core processors are faster than AMD's Ryzen models at similar price points. The top-end Core i9 10900K rivals or beats the more expensive Ryzen 9 3950X in three of our four tests, though it doesn't keep up as well on larger projects.

However, both of those families of processors fall behind AMD's higher core count Threadripper models in most of our benchmark tests. Particularly when working with maps these chips perform the best, with the 32-core Threadripper 3970X being the fastest CPU on average in Pix4D - and the more affordable 24-core TR 3960X not far behind. Intel's Core X series isn't competitive here, unfortunately.

Is the Intel Core 10th Gen or AMD Ryzen 3rd Gen better in Pix4D?

Intel's new 10th Gen Core processors, especially the Core i9 10900K and i7 10700K, are faster per dollar than AMD's Ryzen 3rd Gen chips in Pix4D. However, that statement is somewhat misleading because there is a more expensive Ryzen model that outperforms them, and even higher-end processors which outpace both the Core and Ryzen lines. For the best performance in Pix4D, the AMD Threadripper 3970X and 3960X are clear winners.

Looking for a Photogrammetry Workstation?

Puget Systems offers a range of poweful and reliable systems that are tailor-made for your unique workflow.

Configure a System!

Labs Consultation Service

Our Labs team is available to provide in-depth hardware recommendations based on your workflow.

Find Out More!
Tags: Intel 9th Gen, Intel X-series, AMD Ryzen 3rd Gen, Intel 10th Gen, Intel X-10000, AMD Threadripper 3rd Gen, photogrammetry, Pix4D
Sam Rossback

Do you guys have any guesses as to why the 3990X didn't work as well as the 3970X?

Posted on 2020-05-28 21:46:17
Sam Rossback

Like, is it purely the difference in clock speeds?

Posted on 2020-05-28 23:00:30

I can't speak for Pix4D in particular, but it isn't at all unusual for the 3970X to out-perform the 3990X. Heck, the 3960X often out-performs either of those more expensive CPUs depending on the application.

Part of it is the lower per-clock performance - if the application doesn't scale extremely well, it is really hard for it to benefit from 32 cores, let alone 64. There is also a number of issues I've heard about where even if the app can use all those cores, you can run into conflicts with the RAM or CPU cache becoming a bottleneck to the point that it results in overall worse performance. It is almost impossible to guess how a specific CPU will perform in the real world, which is one of the main reasons why we are so dedicated to doing this kind of real world, application-specific testing.

Posted on 2020-05-28 23:13:46

My best guess? Likely just a lack of utilization for more than ~32 cores. It looks like >24 are used, since the 3970X outpaces the 3960X (slightly) - but the 3990X does not, so the extra cores it has are not helping and the slightly lower clock speed then hurts overall performance. But that could be wrong, as there are so many factors that can affect performance "under the hood". Matt gave several good examples in his reply. It is nearly impossible to know for sure what the exact cause is, but real-world testing like we do can at least show what performs the best - even if we aren't certain why :)

Posted on 2020-05-29 04:26:12
Andrew Milanes

These data sets aren't large enough for the higher core count of the 3990X to overcome the higher core speed of the 3970X. As the number of images and megapixels per image increase, the 3990X will outperform the 3970X. 810 18MP photos is on the small side for drone mapping projects. I routinely have well over 1,000 42MP images for my projects.

William tested a project for me with 1,974 42MP images. The 3990X was 4% faster than the 3970X.

Posted on 2020-06-01 14:11:41

This is also a very good point! I'm not very impressed by 4% faster speed for $2000 higher cost... but it does look like Pix4D (at least in the version we had tested) gained more from higher core counts as the image count increased. There are also other factors, like SMT: turning that *off* on the 3990X improves performance as well: https://www.pugetsystems.co...

Posted on 2020-06-01 18:29:23