Support Website Contact Support Blog

AMD Ryzen 7 computer for Pix4D

Hi there

Im thinking of buying a PC with the following Specs. Would this work with your program?

AMD Ryzen 7 1800x
MSI AMD Bazooka Motherboard
AMD Radeon Pro WX7100 Graphics Card
64 GB Hyper Fury X RAM

We run data sets of between 200 - 4000 Photos.

Would you guys say this will work fine.

Can you recommend something else?

Regards
Quintin

Have you heard anything back. 

Hello all,

When configuring a computer, the most important is to have a balanced configuration. The 4 key components are: CPU, RAM, Hard Disk and GPU.
In the following links you can find more information regarding the hardware and software recommendations:
https://support.pix4d.com/hc/en-us/articles/202557289
https://support.pix4d.com/hc/en-us/articles/202559159

As a good computer for processing medium and large projects, up to a few thousand images, we could recommend, the computers we use ourselves, which are well tested with Pix4Dmapper with the following specifications:
CPU: i7-3930k (6/12 cores/threads),
RAM: 32-64 GB,
Hard drive: 256GB SSD
GPU: GTX 780 ti - 4GB

Another example of a very good computer we suggest is:
CPU: Intel i7 5960X (8 cores, 16 threads).
RAM: DDR4-2400 64 GB.
SSD: Crucial MX100 512 GB.
GPU: GTX 980, TITAN BLACK.

Regarding your configuration we can make the following comments:

CPU
The AMD Ryzen 7 1800X, has 8 cores at 3.6 GHz and it is ranked high in the CPU Benchmark. It is a powerful enough CPU for processing in Pix4D Desktop and it has been successfully tested in our office with several project.
The more the CPU cores the faster the processing because the more “hands” are working on the computations. We usually use this website to compare CPUs: http://cpubenchmark.net/. However, you should keep in mind that there are some processes in step 1 that are not paralleled and for these substeps the faster CPU will be the one with the highest speed (Hz).

GPU
The graphic card is also an important component. It may improve the processing speed for step 1 and step 2 (if the graphic card is compatible with CUDA (NVIDIA Graphic Cards). Processing time of step 3 is not affected by the GPU. The GPU affects considerably the visualization of the rayCloud.

The AMD Radeon Pro WX 7100 is a good GPU for visualizing the project in the rayCloud. However, you will not take advantage of GPU processing, which takes place mainly during step 1 and less during step 2. To take advantage of the GPU processing you need an Nvidia GPU with at least 2 GB of RAM. You can find more information here: 203405619.

RAM
With 64 GB we expect that you can process medium to large projects (up to 3000-4000 images) without issues.
Hard drive: SSD
An SSD hard drive can optimize your processing time, especially for step 3.

Please keep in mind that the different processing steps use the resources in different ways. Even within a step, it happens that some small instructions need to be completed, which cannot be paralleled and therefore it will not use the full resources.
Hardware components usage when processing with Pix4Dmapper are detailed below: 202559519.
Also, having a balanced computer is of the same importance. If you have a very good CPU, then the amount of RAM and the GPU should also be of the same range to achieve the optimal processing time.

Regards,

1 Like

Hi

Few months ago we decided to upgrade to Ryzen: cost/performance motivated us.

I would like to share our experience, but before our configuration:

CPU: Ryzen 7 1800x (8 cores, 16 threads). 3.6Hz. Actually running at 4.0Hz overclocked (most of the time). buffer memory : 20MB
RAM: DDR4-2400 32 GB.
SSD: Samsung 960 pro 512 GB.
GPU: GTX 1080 8Go GDDR5 X.

I was a Xeon and i7 user before. In general Ryzen is closer to a Xeon rather than a i7 architecture. Something between a Xeon and a i7 6950X.

Ryzen suffers of a small latency problem and it is not as reactive than a intel i7. It’s more like a old Xeon when using MS Office for example. However it is a killing machine when it comes to multicores heavy-duty, just like a multi-processors Xeon, because of its high buffer memory.

Pix4D Step 1 no surprises, it is a process based on one or two threats. Performance increased a lot when overclocking from 3.6 to 4.0Hz. Some freezing happened because of overheating but only happened twice with Pix4D and it is the fault of the global warming and it is not our water-cooling fault. Solution was to decrease overclocking to 3.8Hz, waiting for cooler days or a new air conditioner.

Pix4D Step 2, using the same project, our configuration (4.0Hz) ran as fast as Xeon 48 threats without CUDA graphic card. Ryzen is a tractor, low reactivity but high performance for heavy duties.

 

Pix4D Step 3 a lot of changes between single-core, multi-core and SSD heavy duty. We do not have conclusions there yet. It depends a lot on the task: if it is index calculation or only DTM or the different export formats etc.

 

Tip 1: after some test Ryzen works better using SMT in AUTO mode. AMDictives know that.

Tip 2: upgrade BIOS often. Ryzen is brand new and it is getting better and better.

Tip 3: intel i9 change everything but it’s at least twice of the Ryzen price but not twice of efficient.

 

I will keep you informed about our Ryzen adventure and test…

 

Regards,

Dan

 

 

 

 

 

 

1 Like

I just spent an hour writing a reply to another thread that became locked and lost my long winded reply.  Essentially, it was essentially very similar to Antigoni’s reply.  However, CPUs are usually benchmarked for gaming, tested by gamers, and voted on by gamers resulting in daily benchmark variations.  Intel CPUs are built for abusive people such as myself who water block and overclock.  However, even without overclocking, CPUs generate heat and heat is the enemy.  Intel CPUs are purposely overbuilt to handle the heat while AMDs…well… they’re built just well enough to handle an amount just prior to their suicide rate.  

The AMD Ryzen 1800X is only capable of handling 95W of power before it begins to degrade due to thermal instability.  

The Intel Xeon E5-2673 v3 is capable of handling 110W of power with 12 cores (as opposed to Ryzen’s 8) at a 4gHz overclock

Here’s where quality comes to play.  Although the Ryzen can Turboboost to 4gHz in the specifications, they fail to tell you that only a maximum of 2 cores are being boosted in single second intervals at any given time while Intel can boost all 12 at full throttle with up to 20 second intervals.  Add an inexpensive Corsair self contained liquid cooler and you can run all 12 at 4gHz full time.

The Ryzen 1800 benchmarks at #54 with an advertised clock of 3.6gHz and the Xeon benchmarks at #32 with an advertised clock of 2.4gHz and both are priced similarly.

But why would a 3.6gHz processor benchmark 50% lower than a 2.4gHz?  The 3.6gHz is never equalized between all the cores. You’re only going to get two cores running at 3.6gHz with the other four running 50% or 1.8gHz.  With the Xeon out of the box stock, all 12 cores are running 2.4gHz and boost as necessary while running only two cores at idle.  Add the Corsair cooler and you’ll get 4gHz on demand with all 12 cores.

I’ve been building my own computers since 1994 when aftermarket parts became readily available.  I have always built mine for optimized gaming.  I purchased one AMD processor and cooked it before I put the case together.

Here are examples of four computers I own that I tested PIX4D with on a 6 1/2 acre earthworks project with 880 photos total divided at 6 different angles:

2011 Model Lenovo Twist i7-3537, 8GB, 512GB Samsung 850:  Removing all of PIX4D’s enhancement capabilities with the lowest possible qualitative result required 34 hours to complete.

2017 Model Microsoft Surface Pro i7-6600, 16GB, 1TB SSD:  Same processing scenario as the Lenovo required 26 hours to complete.

2017 Model MSI GT83VR i7-6820 w/ Dual SLI GTX1070 16GB GDDR5, 64GB DDR4-2400, 3x Samsung EVO 960 M.2, 1x Samsung 850 Pro 1TB required less than two hours to complete the process as described above.  

Brought everything up to “default” as indicated by PIX4D and achieved full processing in 6 hours.  

Bumped up photo size to “Original” and 100,000 Keypoints, Point Cloud Image Scale to "Original Image Size, Optimal Point Density, Medium Mesh Resolution, 9X9 Point Cloud Densification, all boxes checked in DSM / Ortho with Maximum Pixel downsampling and achieved full processing in less than 18 hours.

I have not attempted to do a maximum output quality as I use this in the field.  I’ll perform one this weekend and report back.

Now, for the computer I built to heat half my home last year…

i7-6900K water blocked and clocked to 4.9gHz, 128GB DDR4-2400, dual SLI GTX1080Ti 20GB GDDR5 with factory water blocks, Corsair AX1500i, and more SSD capacity than the NSA.

 

1: Original Image Size (Double Image Size actually distorts high resolution photos)

    1,000,000 Keypoints (Overkill and barely made a difference from 100,000)

2: Point Cloud Image Scale: Original 

    3D Mesh: Octree depth set to 12, custom resolution set to 131072X131072, qualitative set to sensitive, and densification set to 9X9, sample density divider set to 5

3: GSD set to 1cm per pixel, noise filtering: On, Surface smoothing: Sharp, GeoTIFF Raster: Triangulation

I did my best to max out every setting

Before I give the result, Windows defaults the priority of the program to LOW  At this priority, full processing took a little over 30 hours.

WARNING:  If you want to notch up the program priority, perform the following steps:

Above Normal:  Click to process, then remove the batteries from your keyboard and mouse, turn off your monitor and only turn it on to check the progress bar.  27 hours

High:  Turn off every non essential windows process, give the resources an additional 4GB of RAM, take the wife, kids, dog, cat, mouse, and keyboard on an overnight trip.  22 hours

Realtime:  Turn off every Windows process until it crashes Windows, then back off one process while keeping all your Nvidia processes.  Give the resources the same 4GB RAM, turn off all your anti-virus, screen saver, energy saver, EVERYTHING.  Give the wife money to take the kids, dog, cat, mouse, and keyboard on a day trip.  Clear a few shelves off out of the refrigerator, crank it up to 9, and insert computer tower leaving only the power and display cables to extend outside.  Only plug in the monitor to turn it on to check progress.  Place a cooler full of ice and 12 bottles of beer to the right of your easy chair and a fire extinguisher to the left.  Watch Netflix on IPad, drink beer, and enjoy the silence.  17 hours

Not being married with no young children or animals, I have a refrigerator dedicated to my home computer.  It’s actually a wine cooler with a glass door, don’t judge. However, if the computer is disturbed at all with the priority set anywhere above normal, you will either crash the program or the operating system.  Even I switch off my mouse and keyboard right after I start processing.

My next home build will include two Nvidia Quaddro P6000 cards which will be able to compute X4 the CUDA cores than the GTX series with CUDA SDK.

 

 

1 Like

@Daniel: Thanks for your contribution. Your comment about heating of the CPU in step 1 seems to be reflected in James’ comment about power handling. Let us know if you have new insights based on your tests. :slight_smile:

@James: Really sorry for closing the other post while you were commenting! I was trying to reduce the amount of posts on the same subject. Thanks so much for posting your answer again! I really like your insights on the importance of power handling (e.g. 95W vs 110W) of the CPUs, as well as the explanation of the clock speed and active cores. You seem to have a lot of experience in building computers, looking forward to hearing about your next home build with the two Nvidia Quadro P6000. Sounds like this will be a good reason to take the wife, kids, dog and cat on an overnight trip!

 

Not sure I agree with James. I regularly poor 115 watts through my overclocked Ryzen 1700 without any issues. TDP doesn’t mean anything except to OEM manufacturers. It’s a number AMD/Intel provide to allow for factory system builders to make sure there cooling solution will provide adequate heat dissipation for the CPU. You can blow right past the TDP with overclocking without damaging the CPU. The trick is to run safe voltages, I’m at 1.35 volts, and make sure your CPU cooling solution has enough headroom for the amount of heat your CPU is producing. I run a H110v2, nothing special by any means, and I’ve never seen over 55 c in Pix4D projects with 1k-2k of images and after 24 hours in Prime95 I’m around 65 c. Well below the “cooked,” range.

As for Intel versus AMD construction Intel is by far using the cheaper processes. Cheaper thermal compounds requiring overclockers to pull apart the chips, particularly with anything on the X299 platform, have plagued Intel with roasted chips. AMD is using an overclockers dream on their chips, indium solder. While you can count on Ryzen to dissipate the heat shown in a HW monitor type program efficiently you cannot count on the same from an Intel i9 at this point, not without a lot of work anyways.

AMD Ryzen isn’t a miracle, really a Zen core alone is very similar to the Haswell-E core. Similar clocks and similar IPC. However Ryzen uses less power and costs about half what Intels current matching offering does when you compare multicore performance IRL. I mean Intel didn’t drop the price of the new i9’s by 40% just because and the reviews out today on Ryzen Threadripper appear to be very bad for Intel, unless of course $/performance is not a metric you care much for.

AMD has come along way since the 1990’s and Ryzen is a god send in terms of breaking the Intel price monopoly on enthusiast and workstation level hardware. It finally gives normal people like me the ability to run Pix4D in something reasonable of a time span. I tried running Pix4D on a Skylake i7. I was not impressed.

One huge complaint I have yet to solve though has to do with random crashes in Pix4D with my current rig. I can run Prime95 till the cows come home, MemTest shows zero issues with my ram, and I could stream/game/chat/type all at the same time without any issues but for whatever reason I still get random crashes with Pix4D. Originally they blamed it on my PSU, which I had tested and then replaced with a 1.3 KW unit to prove a point, and I’m still having issues. At this point the only thing that hasn’t been swapped out of my machine is the MB. I’m getting to the point where everything is pointing to something wrong with the software. I hope I hear back from the Pix4D support team soon. Right now I handicap my computer by cutting out one core and 2 GB of memory during processing, which seems to fix the errors, but I’d really like to get rolling again at full speed.

Oh, and here’s my very modest do everything rig:

AMD Ryzen 1700 OC’d to 3.9 GHz on AIO loop

32 GB HyperX DDR4 2666

Asus Strix GTX 1070 on air (strix needs no water loops)

Samsung 1 TB 960 Pro SSD

 

1 Like

I’m about to build a rig to process 5000+ photos for an 800 acre site and will give the AMD a shot.  Thanks for everyone for posting your results.  Planning on the 1700x with 32 GB RAM, 512 Samsung PRO SSD, and a Nvidia 1070 card.

My only concern is the reports of random crashes…I’ll post an update once I get it going.   @sterling - please let us know if you resolve you crash issues.

Good luck and keep us posted.

Think of something like this myself.

Only consider being at 64GB RAM or higher…and 6-8GB of VRAM is not enough either to really open up the full power of all the Pix4D settings.

The Intel vs. AMD debate has been going off/on for decades and will never stop.  My 2 cents is that both are pretty equal right now, I am running the Intel i9 with great success.

Oh, getting the NVMe M.2 drives makes a big difference at different steps as well.

2 Likes

It’s just bang for buck is all. I’m not band loyal.

So I ended up getting the following parts:

Ryzen 1800X - will OC to 4.0 Ghz

64 GB RAM

512 GB Samsung EVO NVMe M2 - I wanted the PRO but interestingly the EVO has higher rated speed according to Samsung’s specs on their webpage.

Nvidia GTX 1070 FTW with 8 GB 

Is it really that much better to get a GTX 1080?  Video card prices seem to be high right now.

Thanks to the cryto-currancy mining folks, but its a good time to sell some slightly older GPU’s.

Okay, I just ordered another upgrade. Like I said I’m down to either my issues with Pix4D crashes being a motherboard or software issue so I decided to upgrade after a conversation with my boss.

Just as an aside if I setup Pix4D to run without using all of my cores on my Ryzen 7 1700, IE run using 16 out of 18 threads, I have no issues. So if anyone needs a work around that’s it.

I might be the first guy running Pix4D on a Threadripper 1950X!

-Gigabyte X399 AORUS Gaming 7 (rev. 1.0)

-Ryzen Threadripper 1950x (16 cores, 32 threads)

-Corsair 110iV2 AIO cooler

All other specs the same. Once everything is built I’ll do some testing and let you know how it turns out. The Threadripper is the most powerful CPU available to our market out of the box outside of the Xeon range until the last two i9 series CPU’s come out. I’m very interested to see how it handles Pix4D.

About the NVMe drives I run the 1 TB Evo version. Andy is correct, right now with the way the memory controller is setup to cache in the Pro and Evo series the Evo actually tests a bit faster. However it does wear out a bit faster too do to it’s cheaper memory, that’s just the price of running SSD in my opinion though. You know almost exactly when it will die.

Only comment I have about AMD versus Intel is that the last thing Intel fans should be is mad about anything good AMD does like Ryzen. Ryzen caused the E-series i5/i7/i9 range prices to drop by over 40%, thus giving anyone who wanted to buy one a price drop like we’ve never really seen before in one generation from Intel. When AMD succeeds, it’s gets Intel motivated to stop ripping people off and actually price things close to fair.

 

As for video cards, with AMD Vega launching right now there is some hope that nVidia will drop their prices slightly. However, as was said, the market is seeing record demand for the 1080 and 1080Ti due to crypto currency mining. It might be best to buy whatever is in your price range and wait for the better cards. 

I have been watching this thread and decided a couple weeks ago to build a ryzen pc. Here is a little background. I bought a eBee SQ this past winter and I built an Intel core i7 computer with the following specs.

Intel Core i7 6700k overclocked to 4.2Ghz
32 GB Ram running at 3200
512GB m.2 SSD – Samsung 950 Pro
Nvidia 1070 with 6GB Ram
Two 4TB Spinning drives in raid 1 for data backup
Windows 10 Pro

I have been using this machine all summer learning and gathering data for my ag drone business that we will be working our way into next year. While this machine was working well and comparing processing speeds to my dealer’s laptop speed I was feeling happy.

With all the hype around ryzen and specifically the lower cost over Intel’s 8 core offerings I decided to build the fastest machine I could based on the 1800x.

Here is what I built.

AMD Ryzen 1800x overclocked to 3.9Ghz
32GB Ram running at 2400
256GB m.2 SSD – Samsung 960 Evo
Nvidia 1080Ti
Windows 10 Pro

I built the machine and started testing out the speeds rendering three different projects. I Testing the following different configurations.

Intel 6700k with no GPU
Intel 6700K with 1070
Intel 6700k with 1080ti

AMD 1800x with 1070
AMD 1800x with 1080ti

The three projects were as follows.

Project Name: Timber 2
Images: 1072
Keypoints: 10,000

Project Name: Leos
Images: 4564
Keypoints: 500,000

 

Project Name: Darwin Home
Images: 3756
Keypoints: 500,000

 

The projects Leos and Darwin Home were later in the season are both 120 acre fields that were flown at 85% overlap.

Here are the results that I gathered from the reports that pix4d produced. I put all this info into a spreadsheet and I will just give you the results of my findings. The numbers are the seconds to run thru steps 1, 2 and 3.

 

Timber 2 Project

|   |   |   |   |   |

Difference

|

increase

|

Winner

|
|

No gpu to 1070

|   |

1325

|

1120

|

205

|

15.5%

|

1070

|
|

No GPU to 1080Ti

|   |

1325

|

1083

|

242

|

18.3%

|

1080ti

|
|

Intel 1070 to Intel 1080Ti

|

1120

|

1083

|

37

|

3.3%

|

1080ti

|
|

Intel 1080Ti vs AMD 1080 Ti

|

1083

|

1019

|

64

|

5.9%

|

AMD

|
|

Intel 1070 to AMD 1080Ti

|

1120

|

1019

|

101

|

9.0%

|

AMD

|

 

 

Leos Project

|   |   |   |   |   |

Difference

|

increase

|

Winner

|
|

Intel 1070 to Intel 1080ti

|

17866

|

12208

|

5658

|

31.7%

|

1080ti

|
|

Intel 1070 vs AMD 1070

|

17866

|

14545

|

3321

|

18.6%

|

AMD

|
|

Intel 1080ti vs AMD 1070

|

12208

|

14545

|

-2337

|

-19.1%

|

Intel 1080ti

|

 

 

Darwin Home Project

|   |   |   |   |   |

Difference

|

increase

|

Winner

|
|

Intel 1070 to  Intel 1080Ti

|

12507

|

8798

|

3709

|

29.7%

|

1080ti

|
|

Intel 1070 vs AMD 1080Ti

|

12507

|

9484

|

3023

|

24.2%

|

AMD

|
|

Intel 1080ti vs AMD 1080Ti

|

8798

|

9484

|

-686

|

-7.8%

|

Intel 1080ti

|

 

My conclusion is that the AMD 1080x was faster on smaller projects with less images where stitching was easier. When I got into the larger projects where I had to up the keypoints to make the image stitch without leaving gaps the 1080ti made the biggest difference in speed.

My feeling is the slightly higher clock of the 6700k and the overclocked ram helps a little with the speed difference. I am now watching the new i9 processors as they will run at a turbo boost level equal the newer 7700k when full processing speed is not needed then back down a little when all cores are utilized.

I ended up keeping the 1080ti in my Intel machine and disassembled the ryzen pc and returned all the parts. I will probably build a i9 setup this next spring before we get back into the fields.

 

If you have any questions please feel free to ask. Hope this helps someone.

Brian

3 Likes

Hello everyone, 

This post is very interesting! @Brian, thank you for sharing the results! 

We have performed some experiments in our office and we would like to share the results with you. 

You will see in this article, how different configuration affects processing time. Check it out here: https://support.pix4d.com/hc/en-us/articles/115003928846 

Looking forward to hearing your feedback,

Cheers.  

1 Like

Major thanks to Brian for the detailed and scientific testing.  I haven’t tried AMD but I am very happy with my Intel i9 and GTX1080Ti setup compared to my old Intel i7 and GTX1060.

It is great to see Pix4D publishing some of their own testing.  However, in today’s world doing a “large” project at just under 20 billion pixels isn’t good enough in my opinion.  My projects are higher than the average user but my current work is on 192 billion pixels making 1 point cloud.  Plus a large, well known company like Pix4D should also be testing on the latest hardware, like the Intel i9 and NVidia GTX1080Ti.  This whole discussion and that article is for people looking to buy a NEW computer…not something 1-2 years old.  Hopefully the QA department at Pix4D will be getting computer upgrades quicker and more data can be added to those charts in the future.

Hello Adam, 

We are working on publishing more results. Stay tuned :-) 

It would be nice to see some modern numbers. Too be honest I’m not really interested in seeing an i9 and Ryzen 7 comparison since the price of a Ryzen 7 is less than half of that of an i9. I want to see Ryzen Threadripper comparisons. I have a new Threadripper build and I would like to know what graphics card is my sweet spot for processing the 135 acre projects I have been working on.

I feel a bit like I was mislead by older posts from Pix4D staff that a card above a 1070 was unnecessary, or at least not a good balance, with a high end processor but now I can see I may need to upgrade to the newest Titan cards.