Software and Application Benchmarking

[TitsNPickles]

Not sure where to start, but I plan to build my first pc. I plan on doing some light video editing, and 3D modeling for a future (hopefully) 3d printer and various metal fab and carpentry projects that I do occasionally. 

 

But a little context first.

 

But most importantly, I am a sub contractor for precision agronomy company and I develop fertilizer and seeding strategies, prescriptions, and various mobile nutrient plans for my customers (farmers) here in the Midwest. Think like ArcGIS and shapefile development, primarily layering several polygons and/or surfaces on a single tract of land (for different nutrients), that I load into equipment rate controllers, that a retailer, co-op, or farmer uses to variable rate apply nutrients and seed.

 

The software that I use is a variation of the customer version of AgStudio by Mapshots (https://www.mapshots.com/products/, $5,000 a license!), but is tailored for reps like myself (domain and data management through AWS, talking to other Granular products and equipment manufactures, ect.)

 

Since this software is pretty obscure to probably most in this community, compared to say, SolidWorks or Adobe Premier, where there is always plenty of component testing and results with these workstation type programs, how do I go about determining what this program needs hardware wise to perform better?

 

I know it is a memory hungry program, and I'm thinking more cores/threads would help from looking through the resource monitor, but are there other sources/monitors that could tell me other details, like if it favors single threaded or multi threaded performance? Memory speed? Hardware acceleration? Anything else that could help guide me through component choices?

 

I have contacted support for answers to questions like these, but I get the minimum requirement speech that isn't very helpful.

 

I am currently running a stock Dell E7470 with a Core - i7 6600U, 8gb of RAM, and a 256gb SSD (not much, but that's what the company set me up with)

 

Any insight would be helpful, as I don't want to spend more money on components than is necessary

Edited December 31, 2019 by LogicalDrm
Link removed

[LogicalDrm]

RAM and cores are key with anything requiring calculation. If you are working for more 3D stuff, then GPU comes a factor, but only if you are working with point clouds or really complicated architechtural stuff. Otherwise you can do it with integrated GPU. This based on my experience with ArcGIS, QGIS, Microstation and GEMS/Surpac (mine planning software).

 

The companies I've worked for, mining company Nordkalk and Geological Survey of Finland, both had laptops as work machines. With GSF giving me cores and RAM with just iGPU. And that all I needed to work with big ArcGIS databases. iGPU was bit laggy rendering 300 drillholes with Surpac. Nordkalk had Quatros on those laptops, but as I was there on temp contract, I had older model which had bit trouble rendering large triangulations and point clouds. There too, having 16-32Gb of RAM was much more important. I'm bit sad that I don't have screenshot of software using over 7Gb of RAM. I have one where 32bit software version caps RAM usage.

 

So for CPU you want more cores and threads. TR is something to look at. With 32Gb of RAM. For GPU things are bit tricky. Many professional software are optimized for Quatro/Firepro series GPUs. Usually with list of supported GPUs on software's website. Or supported driver versions if you are forced of using consumer-grade GPU.

 

However, looking at the software FAQ, does it do rendering/calculation locally or on the cloud? If its latter, then you really don't need more than what i7/R7, 16gigs of RAM and maybe normal gaming-grade GPU for rendering images. So stuff that makes browsing around Google Earth smooth.

[TitsNPickles]

Thank you for the response.

 

I would say that the 70% of the rendering is done locally, with the other 30% is done through AWS. Whenever I initiate a render, some of the functions give me the option to send it to AWS for calculation.

 

But, since much of my work is seasonal, some times of the year are slower to go that route and just calculating locally is easier. Not only that, being located outside of town, my ISP only provides 7/10 Mbps Up/Download speeds at the best of times (they did just receive a federal grant to improve rural internet access, and there is a fiber optic cable buried along my road, so hopefully that will improve anyway).

 

I would prefer to do as much as possible locally, just for the sake of reliability and timeliness. 

[LogicalDrm]

12 hours ago, TitsNPickles said:

Thank you for the response.

 

I would say that the 70% of the rendering is done locally, with the other 30% is done through AWS. Whenever I initiate a render, some of the functions give me the option to send it to AWS for calculation.

 

But, since much of my work is seasonal, some times of the year are slower to go that route and just calculating locally is easier. Not only that, being located outside of town, my ISP only provides 7/10 Mbps Up/Download speeds at the best of times (they did just receive a federal grant to improve rural internet access, and there is a fiber optic cable buried along my road, so hopefully that will improve anyway).

 

I would prefer to do as much as possible locally, just for the sake of reliability and timeliness. 

If you go by benchmarking scroed, look at Blender, Handbrake and other such scores. Gaming scores won't be really beneficial as many of professional software use different drivers, some older, some specially made.

 

Otherwise my previous post is good start.