Initial Fedora 32 vs. Fedora 33 Beta Benchmarks Point To Slightly Higher Performance

[phoronix]

Phoronix: Initial Fedora 32 vs. Fedora 33 Beta Benchmarks Point To Slightly Higher Performance

In addition to Fedora Workstation 33 switching to Btrfs, there are a number of key components updated in Fedora 33 as well as finally enabling link-time optimizations (LTO) for package builds that make this next Fedora Linux installment quite interesting from a performance perspective. Here are some initial benchmarks of Fedora Workstation 32 against the Fedora Workstation 33 Beta on an Intel Core i9 10900K system.

http://www.phoronix.com/vr.php?view=29546

[wizard69]

Well a least they didn't go completely backwards performance wise.

In any event is there any intention to update Mesa before actually shipping?

[Britoid]

Well a least they didn't go completely backwards performance wise.

In any event is there any intention to update Mesa before actually shipping?

If it isn't updated before release it will likely get updated mid-release.

[spykes]

The filesystem change seems to bring some performance regressions... I'll keep ext4 for a while so.

[qarium]

I use Fedora 32 and Btrfs really Looks Like a Sapontage in the Performance

Btrfs should Not be the Default

[crowen]

Well a least they didn't go completely backwards performance wise.

In any event is there any intention to update Mesa before actually shipping?

https://copr.fedorainfracloud.org/coprs/che/mesa/

[skeevy420]

Well a least they didn't go completely backwards performance wise.

Like Michael pointed out, BTRFS caused some of those slowdowns and that there are config tweaks to speed up certain operations. All that says to me is that Fedora BTRFS needs a more advanced setup by default.

It's like running ZFS on Root where the stock settings suck ass and /lib, /etc, /bin, /var all need different default mount options and file system features used. Like using gzip on /usr/share/docs and lz4 on /bin or setting primarycache=metadata on a db or sync=off for /tmp.

In any event is there any intention to update Mesa before actually shipping?

That X 1000

[JackLilhammers]

Something doesn't seem right.
In nearly every chart where there is a sizable difference between F32 and F33, F32 wins, but in the end F33 wins the geometric mean.
How's that possible? Maybe the labels are reversed?

[skeevy420]

Something doesn't seem right.
In nearly every chart where there is a sizable difference between F32 and F33, F32 wins, but in the end F33 wins the geometric mean.
How's that possible? Maybe the labels are reversed?

The article doesn't have all the tests included while the Geometric does.

FWIW, That was my first reaction as well.

[Michael]

Something doesn't seem right.
In nearly every chart where there is a sizable difference between F32 and F33, F32 wins, but in the end F33 wins the geometric mean.
How's that possible? Maybe the labels are reversed?

See the linked OpenBenchmarking.org page in the review for all the data points, it wasn't practical including all ~200 benchmarks especially for the ones where there is virtually no change.