I want to see benchmarks on how this affects actual gameplay. That’ll be the real story.
From what I’ve seen the authors of the papers have listed the zenbleed mitigation impact as “statistically insignificant”.
Which was in stark contrast to Intel’s “up to 50%” performance hit for the Downfall fix.
Which is clearly not correct. Take a look at eg. this benchmark; many workloads take a sizeable hit. Even plain 'ol glibc
sinandcostake about 8% longer, and the most pathological hit was the MariaDB workload which took almost 200% longer. Looks like many tasks related to math or heavy-duty string processing will be at least 10-20% slower, but it’s hard to say yet what this’ll do to games. I’d expect CPU-heavy games to be affected
.
Removed by mod
Bit of a silly mitigation to have on the Deck, honestly. Hopefully there’ll be a way to disable it; it’s probably going to kick performance down by tens of percents, and I really don’t need that on a gaming device where this particular attack is just incredibly unlikely and most likely wouldn’t lead to jack shit anyhow.
People have proven that the Steam Deck is pretty much just a daily desktop driver. People do all kinds of things on it. It also seems like its performance downgrade is insignificant.
And I’m not one of those people so I’d like the option to not take the performance hit. Based on benchmarks I’ve seen it’s definitely not going to be insignificant, but it’s hard to say what the effect on games will be. See eg https://openbenchmarking.org/result/2308071-NE-2308068NE86, where math operations can be seen to take about 10% hits and some tasks go way above that. Not terrible but it’s not going to be nothing either, which is why I’d like to be able to opt out
Edit: corrected the link
