Conversation

Vlastimil Babka

Since the report of 3888.9% performance improvement made it not only to Phoronix (no surprises there) but is now also the subject of at least two youtube videos, where one has 212k views at this point (just wow), I've decided to explain in some detail why the benchmark is in this context completely bogus, and that people missed the very same report also contains a 9% regression in another benchmark (which may be actually less bogus :)
Hope someone finds this useful. http://lore.kernel.org/all/3b09bf98-9bd4-465b-b9c5-5483a6261dc7%40suse.cz
The youtube videos: https://www.youtube.com/watch?v=m9dZkRwWEj8
https://youtu.be/OvLEx6fPVrg
2
16
22

DougMerritt (log😅 = 💧log😄)

@vbabka
Your explanation completely explains what's going on, but more broadly, I'm sure you never believed 3888.9% improvement in the first place.

1
0
1
@dougmerritt of course seeing such an absurd number means the test must be measuring something very weird ;)
0
0
2

@vbabka no, I can't believe that Phoronix is taking a synthetic benchmark as an earth-shattering improvement in computing without understanding its (lack of) impact on real-life loads! It never happened before!

https://flameeyes.blog/2012/06/23/debunking-x32-myths/?mtm_campaign=social&mtm_kwd=mastodon

0
0
1