A Guide to Getting the Best Performance with Large Datasets

Issue link: https://geospatial.trimble.com/en/resources/i/1415711

Contents of this Issue


Page 8 of 12

Trimble Software: Working With Large Datasets 9 Addendum A: PC Details and How They Relate to Trimble Software One of the first questions that arises after being selected to manage a project involving large amounts of data is whether your existing computers are up to the task. To help determine the answer, this section provides descriptions of key desktop PC components that exist in the market today and the configuration choices that provide the best processing of scan data using Trimble software. There are many variables and complex interactions that must be considered to ensure that the system holistically provides good software performance. Computer Harmony The key to optimizing computer performance is balance. One must remove as many bottlenecks as possible to get the best end-to-end performance. For example, if you have slow drives and a fast CPU, the CPU may not be fed data fast enough to perform at its highest potential. Similarly, slow RAM may not be able to feed a high- performing CPU fast enough. On the other hand, using RAM and drives that are high-speed with a relatively slow CPU can cause an input overload situation that hinders optimum computation speed. The Central Processing Unit (CPU) The quickness of a manufacturer's instruction set for their CPU has a significant impact on overall execution speed. Today, people following advances in computing technology will note that AMD has an instruction set that executes faster than Intel's, so for the same clock speed you will see better performance from the AMD processors. This is only part of the story. Higher clock speeds when using Intel CPU's can typically overcome the AMD IPC improvements, especially with single threaded operations. Thus Trimble recommends Intel CPUs as the preferred option for use with our scanning software. Random Access Memory (RAM) For Trimble scanning software, the speed of the installed RAM has a slightly greater impact on performance than the amount of RAM. Thus it is important to consider the speed of added RAM and not just its quantity. The recommended minimum amount of RAM to have on any computer that you use with Trimble scanning software is 32 GB. The software can use a good fraction of that while processing large datasets. If you process data in Trimble RealWorks from 500 scans or more, it is possible to hit a limit with 64GB of RAM. This would negatively impact performance, as slower virtual memory on a disk would need to be accessed. However, it still is generally better to have 64 GB of really fast RAM than to have 128GB of slower RAM. It is often observed that the speed of RAM goes down as the amount of available RAM grows larger. There is a fixed frequency at which the system runs. Therefore the speed at which the memory is accessed is inversely proportional to the amount of memory available. Also the BIOS for some motherboards cannot run as quickly with all of the memory slots full of RAM chips. This is manufacturer dependent. It is instructive to look at your motherboard's website. It should have a list of supported memory modules. Typically the site would include a list of the brands, SKU's, and speeds that have been validated in the manufacturer's own internal testing. Only use validated RAM for their particular motherboard to avoid potential problems.

Articles in this issue

view archives of Whitepapers - A Guide to Getting the Best Performance with Large Datasets