I recently had a Windows PC perform an unauthorized system update in the middle of a big batch of files and gave up and bought my first ever MacIntosh/Apple computer. Possibly because I grew up in somewhat extreme poverty and possibly because I don't particularly like the color silver, these things always seemed like they weren't for me. But they are advertised as very very fast.
It's easy to be skeptical, though, because now they are making their own chips. They can use whatever benchmarks they want. The benchmarks I care about are processing proteomics data!
TESTING TIME! My Apple M2 Pro with 16GB of RAM and 1TB hard drive vs my closest looking Windows PC - same number of cores, but Passmark thinks it is a little slower.
First experiment, performed between when my elderly dog needed to pee at 3am and when my toddler woke up at 5am. Details are fuzzy....
Tools - SearchGUI with MSAmanda 3.0 and SAGE
Random HeLa digest file from an Exploris 480 with FAIMS. 200ng separated over 120 minutes on an EasySpray 25cm x 75um with a 3cm PepMap trap. Converted to MGF and having around 110,000 MS/MS scans. I might have generated it or downloaded it. Not sure. It was on my hard drive from this review I did a while back. If you want the file the MASSIVE download link is in it.
Search tolerance was set at 10ppm MS1 and 10ppm MS2. I used the Uniprot Swissprot for human (9606) that I downloded this morning. 20k entries and I had SearchGUI add decoys. M+oxidation and static carbamidomethylations (these are the SearchGUI defaults).
Windows PC - pretty freaking fast at 3 min and 25 seconds!
2 minutes and 5 seconds! WHAT? Not the way I expected that to go.
Okay - I should also point out here that SAGE is nuts. But the discrepancy is even more insane on the much new MacIntosh chip. 13.4 seconds to search a file vs 2.9 seconds to search a file??
Then it occurred to me that it's probably not fair to run a file on a MacBook on a battery. I plugged it into the wall, deleted the first output file directory and reran the file.
Maybe a tiny bit faster, but within the margin of error here?
It is fair to note that the Windows PC I'm currently typing this on doesn't do much data processing these days. It's certainly perfectly suitable and I have zero reasons to think I'm not going to be using it as my primary office PC for several more years. However, for our main PC "server" which has supported as many as 6 users with SpectroNaut, Proteome Discoverer and Compound Discoverer we've been using a Ryzens 9 (I think this is the chip on that box below) - which aren't even all that fast now! The big EPYCs and Threadrippers are pushing 3x this benchmark - but you've got to be ready to drop as much as $10k just on your processor. Good time to be a consumer if you need PC power!
No comments:
Post a Comment