Just the Messenger

Recently, while running some performance benchmarks, on a lark, I decided to run them on a faster PC. It was a lark, because, I figured faster would be faster and would not provide me with more insight. But 35% is a lot faster and is worthy of further consideration.

So for your consideration; the first machine, "S1", I ran the benchmarks on has dual 2.2GHz Xeon processors and 1GB RAM (RDRAM). The second machine, "F1" has dual 3.2GHz Xeon processors and 3GB RAM (DDR). The benchmark I was running is a "Dependency Check" on a target configuration with 536 components that resolve down to a final configuration of about 1300 components. The benchmark on S1 took an average time of 11 minutes and 39 seconds. F1 performed the benchmark in an average of 7 minutes and 35 seconds, or, about, 4 minutes faster.

Right now you’re thinking, "50% faster processor or 3x more RAM, which is it?"; which was my initial thought. Well, I pulled 2GB of RAM out of F1 and the benchmark timing didn’t change significantly. OK, +14 seconds isn’t significant in my book. The configuration did appear to load faster (think "snappy") with more RAM, but, that wasn’t my benchmark.

I also, as a side note, looked at what affect running MSDE versus the full SQL Server package had on performance. Here I found that on both S1 and F1, the benchmark was unaffected by swapping between these two database engines.

So, you can take this benchmark to your boss, who’s probably been bugging you about your build times, and point out that it’s your 5 year old PC he’s stuck you with. If he still yells at you, please remember, I’m just the messenger.

- Jim