This is a really interesting question. I have long wondered whether or not Tableau makes use of GPU resources to enhance rendering times, or perform other computationally-intensive calculations internally.
I haven't found any documentation on this, just a couple of really old forum posts that talk about 2D graphics processing, and which seem to suggest that modern multi-gigabyte gaming-oriented GPU's don't really do much to improve Tableau's performance.
It would be very useful to know from an internal architecture standpoint how Tableau uses CPU vs GPU vs memory to manage performance.
1 of 1 people found this helpful
Tableau licensing is either named user based or based on # of cores on the CPU: Licensing Overview. Typically graphics card is not necessarily the bottleneck to the overall equation, but I'll refer you to an earlier forum discussion on that subject: Hardware for the best Tableau performance. Better GPU might be good for Tableau, but overall you return on investment might not be dramatic.
I get this question a lot around "what hardware is best for Tableau" - and the honest answer is: it depends and there is no common solution. Because dashboard designs, data volume/strategy, usage pattern, volume of automated jobs, etc can all vary dramatically between each use case of Tableau Server, the performance bottlenecks will vary as well. In general, I would monitor and tune three main concepts with Tableau Server if you want to invest in hardware:
- Processor speed and # of cores: This will both help it handle requests faster and potentially help is scale to handle more requests. This does have a licensing impact for any core-based licenses of Tableau Server, but since it directly relates to scalability it makes sense that this is licensed.
- RAM:Tableau does tend to be memory intensive, especially with uses of extract and the data engine. If RAM is not available, extracts are queried on disk, which can be dramatically slower. You want to ensure you have sufficient memory especially with growing extract volumes.
- Disk I/O: This tends to be overlooked but a slow disk storage will result in Tableau take overhead to load workbooks. A better disk can help prevent this bottleneck and we have seen some deployments invest in SSD to ensure better read speeds.
More info might be found here: Collect Data with Windows Performance Monitor. This is really just some of the potential hardware related bottlenecks - keep in mind that Workbook Design/Complexity has a lot to do with the overall load performance, as does network latency. Performance is really something that requires troubleshooting rather than purely a question of quality of hardware; one suggestion would be to consider reaching out to our consulting wing to help go through that troubleshooting to understand what is the best means of improving performance in your use case: Consulting Services | Tableau Software