Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1663
Article Thumbnail

Pix4D 4.4 CPU Performance: AMD Threadripper 3990X 64 Core

Written on February 7, 2020 by William George
Share:

Introduction

AMD has launched a new top-end CPU in its Threadripper product line, equipped with a whopping 64 cores: the 3990X. We have put this new chip to the test in a wide range of applications, from photography to rendering, but this article is looking specifically at Pix4D. Keep reading to find out how it stacks up to the other Threadripper models as well as Intel and AMD's various other desktop processors.

AMD Threadripper 3990X Tested in Pix4D

For this article, we are primarily interested in how this new AMD Threadripper 3990X compares to other current-generation processors. The focus is on photogrammetry processing in Pix4Dmapper, using a benchmark tool we developed here at Puget Systems. More info about this utility is available in the Benchmark Details.

If you would prefer to skip over our test setup and benchmark sections, 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 Hardware

Here are the specifications of the different platforms we used for our testing:

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-U12S 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)
AMD Ryzen Test Platform
CPU AMD Ryzen 9 3950X ($749)
AMD Ryzen 9 3900X ($499)
AMD Ryzen 7 3800X ($399)
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 ($499)
Intel Core i7 9700K ($385)
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z390 Designare
RAM 4x DDR4-2666 16GB (64GB total)
Shared 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 4.4.12
Puget Systems Pix4D Benchmark

A quick note about memory specifications: the latest CPUs from AMD and Intel support varying speeds of RAM depending on how many modules are installed. In past articles, we've sometimes tested at multiple speeds, but to help keep things more uniform going forward we are standardizing on the speed that each processor/platform supports when all memory slots are used (even if we aren't filling them completely for our testing). Along with using the same total RAM capacity, that should help to keep results fair and consistent.

Additionally, it is worth pointing out that the CPU cooler we used for testing the new TR 3990X is somewhat borderline for how much heat this processor can potentially put out. In the systems we build with this chip going forward, we will be utilizing a larger heatsink and fan to ensure maximum performance even under extended loads - but in the open-air testbeds we use here in our lab, we found that the cooler we used did not lead to throttling.​​​​

Benchmark Details

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 testing, 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

A lot is going on in the chart below, so before getting into it here is a reference key for the color scheme:

  • Light blue = Intel consumer CPUs (9th Gen Core series)
  • Dark blue = Intel HEDT CPUs (X series)
  • Light red = AMD consumer CPUs (Ryzen 3rd Gen)
  • Dark red = AMD HEDT CPUs (Threadripper)
  • Dark red with glow - NEW AMD Threadripper 3990X

For those who want to dig deeper into how each CPU performed in the various processing steps within Pix4D, here is a full results table:

Pix4D 4.4.12 Processor Performance Table

Click image to expand for better readability

Analysis

AMD's high core count processors, from the 16-core Ryzen 9 3950X up to the 64-core Threadripper 3990X, reliably took the top four performance spots in every data set within our Pix4D benchmark. That alone is quite impressive, considering Intel's performance dominance for most of the last decade. The 32-core 3970X appears to be the sweet spot, as it led the pack (or tied, in one case) across the board... but the less expensive 3960X was not far behind (never more than about 5% slower). Unfortunately, the additional cores in the new 3990X didn't help enough to offset its slower clock speeds, so while it performed just fine it couldn't justify its price tag in Pix4D.

Is AMD's Threadripper 3990X Good for Pix4D?

AMD's new 64-core Threadripper 3990X performs similarly to the lower core count 3960X and 3970X in Pix4D - so it is good in terms of raw speed, but a poor choice from a price/performance standpoint. Going with one of the lower core count models in this family will give you similar results for thousands of dollars less. The Ryzen 9 3950X is also a good choice for those on a more limited budget.

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, Intel vs AMD, AMD Ryzen 3rd Gen, AMD Threadripper 3rd Gen, Threadripper, CPU, Pix4D, photogrammetry, Threadripper 3990X, Intel X-10000
kermi

Could it be possible to test the 3990X with Windows 10 pro for workstation version? Anandtechs testing shows hugely improved performance vs normal pro edition.
Pix4D isn't processor group aware, so it can only use half of the threads on the 3990X.

Posted on 2020-02-08 07:40:56

Already planning on it, along with testing with SMT off. I read that Anandtech article as well :-)

Posted on 2020-02-08 07:47:50
Jonathan Kubiak

One thing I don't see you testing is running parallel processing.

I work on a large project, performing up to 10 drone models a day. I generally run 2-4 Pix4D instances at once.
One answer that I'd like is whether multiple graphics cards would be utilised in multiple instances of Pix4D, even if it is only one card per instance.
Same with the processors - 32core had the edge over 64core in single instance processing but would 3 instances of Pix4D running at once give the 64core the edge?

I also process large datasets on a monthly basis (5500 images). I'd love to see you benchmark these datasets (and can provide the data if you like).

Thanks

Posted on 2020-02-23 03:54:35

I have had a few other people ask about running concurrent instances of Pix4D (and other photogrammetry applications) as well, but the benchmarks I have developed do not behave properly if more than one instance of them (or the test software) is running at a time. From what I can tell, the tools I use for automation (primarily AutoIt) don't seem to have ways to deal easily with identifying and separately handling multiple windows with the same name... though I admit there is still a lot about programming and working with tools like this which I do not know, so maybe I'm missing something which would allow it.

Maybe at some point I can run a more manual test to see what happens with 2 or 3 instances of Pix4D on one of these big processors. I'll start mulling that over.

As for the larger data sets, the main problem becomes time. In order to run 2 or 3 iterations of each image set, we are already looking at 12 to 24 hours for all of our image sets on just *one* of these photogrammetry programs on *one* set of hardware. Even larger projects would exacerbate that, and it doesn't seem (from some individual testing I've done with a contact at Pix4D) that behavior changes much with larger projects, other than needing more RAM. I do appreciate your offer to provide datasets, though, and may be in touch if we decide to do some limited testing with bigger stuff than we have available now :)

Posted on 2020-02-24 19:11:25