Make ArcGIS Pro Faster (Benchmarked by ArcMap)

[JeffBishop]

Simple idea. 

GeoNet is flooded with complaints of how slow ArcGIS Pro is. 

64Bit Architecture was supposed to bring speed. 

Simple operations such as opening, saving, selecting, etc are sloooow. 

Geoprocessing takes forever and makes my PC sound like it's going to fly. 

I have compared the performance between Pro on my workstation grade laptop to my crappy admin laptop using Map and Map wins out solidly on almost everything. 

Pro has a nice UI and flow to it, but suffers from MAJOR performance issues on even the most powerful of machines. 

Working with local datasets has also not seemed to help at all. 

For reference, my workstation running pro has: 

8 core (16 Threads) Xeon 2.4 GHZ

32 GB of RAM

NVIDIA Quadro RTX300

[JeffBishop]

See this closing comment for details: https://community.esri.com/t5/arcgis-pro-ideas/make-arcgis-pro-faster-benchmarked-by-arcmap/idc-p/1090903/highlight/true#M16886

[EmmaHatcher]

I am finding that ArcGIS Pro 2.7 is even slower than previous versions. Just like the original post, every selection, edit, save, etc. takes at least 30 seconds to process, and my computer fan makes so much noise doing each of these things. I'm working in a project right now with one layout, and one map. I'm nowhere near out of space, so that can't be the issue. What is going on??

[KoryKramer]

Hi @EmmaHatcher Waiting 30 seconds for a selection, edit, save isn't expected so will require troubleshooting.  While it could be helpful to take a look through Troubleshooting Performance Issues in ArcGIS Pro it may be helpful to investigate with technical support.

I hope this helps.

[TamirNof]

Hi,

I did not change anything but it works better now - do not know why.

I have a workstation with Asus X470 motherboard, CPU from AMD (2700X), GPU
from AMD (RX590 8GB DDR6), RAM 64GB 3200 MHz, and storage from Samsung (SSD
970 EVO Pro) - so regarding hardware, I'm in a more than average shape.

What I did change is the 'Parallel Processing Factor' to 75% (I have 16
cores so should put 12 cores to work). It really helped in tedious
geoprocessing missions.

Sincerely, Tamir

[JWickham]

I also experience very slow processing.  Right now I'm running an "Extract MultiValues to Points."  The point file has 11.7 million observations (points) and the raster has 30-x-30 m pixels.  Only 1 raster map was used. It's been running for 7 hours and is only 29% complete.  You might say that is a lot of points, but the task manager indicates that only ~4% of the cpu and 4% of the GPU are being utilized.  Even if I could get 50% capacity there would be something like a 12x increase in speed and the process would take a day or more to complete.  The same function took 5+ days when included 3 raster data sets.

I'm running ArcPro 2.6.2 on a machine with 2 physical chips with 8 cores per chip and 64 GB of RAM.  The clock speed is a little slow (2.10 GHz), but still ...  The pointid attribute is indexed (Attribute Index function). 

I know "Extract MultiValues to Points" is not yet multi threaded (no parallel processor field in the environments tab) but even for the functions that are CPU and GPU percentage usage only increases to ~10%.

Working with large datasets is taxing in ArcPro.  I can do joinfields in SAS on the same 11.7 point file and it finishes in less the 2 minutes!

[YuWang-Jade]

Hi @JWickham Thank you for your input for "Extract MultiValues to Points" tool. We'd like to narrow this issue down and have a solution for it, but there are some extra questions we'd like to know first. Would you prefer to continuing discuss in this thread or you can send me a private message for us to exchange emails.

[JWickham]

Am I replying to an ESRI employee or an ESRI community member?

[YuWang1]

@JWickham I am a Product Engineer from Esri Spatial Analyst team. Previous comment was posted from a different account. Sorry about the confusion. 

[JWickham]

Thanks for replying. What are your other questions?

[YuWang1]

@JWickham You mentioned in your about the raster data you used, when you say "raster has 30-x-30 m pixels", do you mean the raster has 30-million rows x 30-million column? If not, could you tell us what are the numbers of the rows and columns? And we are also interested in the data type of the raster.

Also, we'd like to know what are the exact details of your workflow (type of feature class, were selection sets used, exact gp environments in effect, using local disk or network access, etc).

Another question would be how long does it take to run copyfeatures on that same machine using the same point fc as input, copying to the same type of feature class?

At last, did you have a chance to try it at 2.7? 

Moreover, we are wondering if it possible for us to get an anonymized version of your data?

[YuWang1]

@JWickham As for the "Extract MultiValues to Points" tool, we can't reproduce the performance issue. If the problem persists, please contact our technical support team or provide a screen recording showing the issue from setting everything up at the beginning and demonstrating the performance problem. Please let me know if you have any further question.