Best setup for Alteryx - newer processors

[bminuk]

With all the new buzz about the new AMD Threadripper processors, is there anyone who knows about any limitations in using one of these massive processors for running Alteryx? I know that sometimes too many threads can be a performance issue. Also, does memory bandwidth or larger caches help much?

Also, does anyone have experience using Alteryx with the newer Intel processors with Performance and Efficiency cores? Do you require Windows 11 for them to work properly in Alteryx?

I do flows involving large amounts of geographic data and non geographic data with record counts in the millions and data sizes that can range into the gigabytes.

[Raj]

When using AMD Thread ripper processors for Alteryx, consider that too many threads may not always lead to linear performance gains. Factors like clock speed and architecture matter.

Larger caches and higher memory bandwidth can help, depending on your workflows. For Intel processors with Performance and Efficiency cores, check Alteryx's system requirements and release notes for compatibility. Windows 11 may bring improvements but ensure its compatible with your Alteryx version. For workflows involving large geographic data, processor features influencing spatial operations should be considered.

[bminuk]

The reason I ask about Windows 10 is that I am aware that many are stuck on it due to IT policy. Is Windows 11 a requirement for newer processors?

What sort of processor features affecting spatial operations are you referring to? Would this include AVX512 (not supported on the Intel regular desktop processors)? Does Alteryx use Intel or AMD specific features?

[Raj]

As of My Knowledge windows 11 is not a compulsory requirement. 

Also, regarding processor features affecting spatial operations, these can include SIMD (Single Instruction, Multiple Data) instruction sets like AVX (Advanced Vector Extensions) or AVX512. These instruction sets can enhance the performance of certain mathematical and data processing operations, including those related to spatial data.

hope this help.

[bminuk]

It would be useful if the company would post what special features (AVX512, AI extensions, etc.) that Alteryx takes advantage of, in order to aid in selection of a processor.

It would also be good to know if any of the common benchmarks (SpecInt, Passmark, Geekbench, etc) are good predictors of performance.

[Raj]

I Agree to the point and there might be something present on community around same.

[bminuk]

For the record there is a document on thread scaling AMP Engine Best Practices . It says to have 6-8 logical cores per running flow. So if 2 flow operation is anticipated, it says 16 logical cores and 32 GB memory. However, measurement beyond 14 cores do not seem to have been made.

There has been some excellent work done in the post Hardware Matters , however this is dated around 2019-2021, and much has changed since then. A repeat of this work with newer, larger machines/instances may be useful. Also, what are the effects of E cores on scaling? The newer Intel desktop (non Xeon) processors lack AVX512. Does this hamper things on Intelligence Suite?

[bminuk]

A have a question on the "AMP Engine Best Practices" document about the Benchmark Results for a Typical CPU Heavy Predictive or Machine Learning Task . It appears that having two E2 workflows shows less benefits from additional threads, with marginal improvements out to 12 cores for single E2, while a bottleneck appears to occur at 6 threads with two simultaneous workflows (something similar is shown in the data prep example, with bottlenecks at 8 and 6 respectively). Can there be issues with cache size or memory throughput?