topic Re: Benchmarking two CPUs in SAS Programming

Benchmarking two CPUs

michelconn — Thu, 27 Jun 2019 00:50:54 GMT

I'm trying to convince management that getting a better CPU will increase the performance of SAS but they need concrete numbers before I can push things farther.

Luckily my coworker has an I7 16gb ram and I have I5 8gb ram. My plan was to run two identical programs overnight and see which computer finishes first.

Does anyone have an idea of some good code that could test CPU performance? What SAS operations stress the CPU the most?

Re: Benchmarking two CPUs

koyelghosh — Thu, 27 Jun 2019 01:23:13 GMT

Will this link be of use to you (https://www.lexjansen.com/nesug/nesug10/ma/ma09.pdf)?

Re: Benchmarking two CPUs

Reeza — Thu, 27 Jun 2019 03:07:11 GMT

RAM isn't a huge deal usually with SAS, but disk types can be - my SSD is about 10x as fast as my non-ssd drives.

Re: Benchmarking two CPUs

koyelghosh — Thu, 27 Jun 2019 04:06:10 GMT

@Reeza You are right. Nice point about SSD and non-SSD drives. However in some cases you can choose to do in memory computation (also called in-memory analytics) as advertised here (https://www.sas.com/en_in/solutions/in-memory-analytics.html).

In such a case having more memory might be better.

Re: Benchmarking two CPUs

Kurt_Bremser — Thu, 27 Jun 2019 06:46:56 GMT

If you want to pinpoint CPU types as the performance driver, you need to isolate that fact. Meaning that everything else (main bus/bridge, I/O and network controllers, storage) must be identical.

But from my SAS experience, it's not the processors, but mainly your storage where most performance can be gained. Switching from, say, a single spinning metal drive to multiple SSDs will feel like being on a different planet.

What is your current SAS setup? Client/server, or desktops only? If the second, how many desktops? If the first, what kind of hardware and which operating system on the server?

Re: Benchmarking two CPUs

michelconn — Thu, 27 Jun 2019 12:17:06 GMT

I am desktop only but most of the data I pull from is stored on a server. I would say a majority of my work is done on my local machine.

Re: Benchmarking two CPUs

Kurt_Bremser — Thu, 27 Jun 2019 13:42:06 GMT

Then you might want to concentrate your SAS computing power on the server. Add cores there, and set up a BI server (if not yet present), and drop the desktop installations. A centralized setup will need more attention one time while setting up, but be much easier to maintain later on.

Servers are usually better equipped with regards to storage.

Re: Benchmarking two CPUs

Reeza — Thu, 27 Jun 2019 14:33:06 GMT

I'd say if you have access to a server you should have no problem with speed. Why are you working locally?