Technology: Tech stack & Key Info on the Engine

We often get asked about the performance and capabilities of the engine. It’s rapidly advancing – so please don’t take this as gospel but we’ve recorded some KPIs and indicative data and examples for reference. You can find links to other technological capabilities and features below:


Polaron by the numbers

Some quick key stats on how Polaron performs and some key indicators of it’s performance.


23
People

Have already signed up to our Early Access Programme and more join all the time

200 km2 +

Simultaneous modelled area in real-time
Based on 1 ? scale model running in real-time

Ten concurrent users

Concurrent editing users interfacing with a single mapped simulation (2D & 3D).


Circa
£3-5,000

Cost of PC / Edge based server to Run Polaron based on an i9 & 3090ti or better plus SSD etc.

112 seconds

Is the time it takes to “unpack” the entire UK Open Street Map DB
Inc Error Correction. Based on 1 ? real-time model and 50km2 tiles

152 minutes

Is the time it takes to “unpack” the entire EU Open Street Map DB
Inc Error Correction. Based on 1 ? real-time model and 50km2 tiles


350 milliseconds

Typical time to ingest and render point cloud from a typical Velodyne Lidar (determined by read/write of Motherboard / CPU / GPU)

150 milliseconds

For 50km2 tile retrieval and render
permitting seamless real-time scrolling over very large AOIs.

EU-Horizon proven

Partly developed within a Horizon project focussing on real-time simulation over cellular networks.


Minimum & Recommended Spec

This will likely be revised upwards for some of the commercial applications, whereas for gaming and historical simulation we want to ensure the biggest audience can enjoy these products.


CPU / GPU requirements

PlatformMin SpecRecommended SpecComment
Portable / EdgeLaptop – i7 / 3070tiLaptop – i9 / 3090tiSmaller scenes run acceptably
Fewer agents / Ai on lower spec
Consumer PCPC – i9 / 3090ti PC – i9 / 4090tiSmaller editable scenes on lower power devices
Nvidia CloudTBCTBCNot a problem – remains untested*
*NB – we use NVidia GPU hardware – but it could run on most GPU / CPUs we just haven’t tested it yet on AMD or Ryzen based systems, but the software is interoperable.

RAM & Memory Requirements

Please note that due to the very high data volumes involved – Polaron really runs best on solid state memory and as fast a RAM array as is available in order to avoid bottlenecking the CPU / GPU.

ComponentMin SpecRecommended SpecComment
SSD256GB500GBAs fast as you can go!
RAMSolid state @robert-sugarurbanhawk-space Solid state @robert-sugarurbanhawk-space As fast as you can get!
Network connection Wifi-6EthernetLower latency leads to fewer issues and hosting of games / collaborative simulations
*NB – we use NVidia GPU hardware – but it could run on most GPU / CPUs we just haven’t tested it yet on AMD or Ryzen based systems, but the software is interoperable.

How fast?

Typical frame rates from testing (i9 – 32GB RAM, 500GB SSD & Nvidia 3090Ti)

*Please note that these tend to get faster all the time! Please don’t treat these as gospel as it’s updated as frequently as we can. Performance enhancements come in a range of methods – therefore some are cumulative and others are reworks of pipelines to enhance efficiency.

We will release more finite benchmarks to track progress in Q1 2024.

PlatformMin Frame rateTypical Frame RateMax frame rates
Landscape engine (2.5 / 3D)35fps40-70fps140fps
Volume engine (3D)29fps45-70 fps140fps
Physics single thread (C++ / CPU)18-20fpsBy slowing the physics in games
we can “tick” over the sim and gameplay isn’t affected.

For serious sims – FPS is at or closer to real-time or higher – depending on the application.
OSM Unpack – UK2 mins45s
OSM Unpack – EU2hrs1hr 45
200GB Lidar import2mins3minsDepends on packing etc. Once saved as scene it’s an instant load. Can be real-time fed or chunked.
*NB – we use NVidia GPU hardware – but it could run on most GPU / CPUs we just haven’t tested it yet on AMD or Ryzen based systems, but the software is interoperable.