'Kernel memory leaking' Intel processor -- a serious cpu bug!?!

[Guenon]

The problem with the % voices "loss" they report of course is we may be seeing something that only affects the audio driver, not the plug-in.

So it's very hard to say. A DAW has many different components interacting together, and an audio-driver update or a different audio device might eliminate this entirely.

Yes, certainly. This of course makes it relevant for the ones who need that last 3-9% in their current projects, for example (as given the above, it's not certain how and in which part of the chain the measured loss will be remedied). So if anyone's running a stable and dependable Kontakt workstation and is, at this point in time, interested in the possible losses before applying BIOS patches, it's naturally nice to have this sort of data available, regardless of what specifically causes the decreased performance after patching.

[aciddose]

Definitely, I agree. If people are running Kontakt 4 with the same libraries and configuration, 10% seems like it's genuinely being observed.

Speculation: with 6-cores at play it brings in to question whether Kontakt is running all those voices on a single core which should only suffer minimally, or using IPC (inter-process communication) which would suffer most from the context switching issue. Very complicated.

That doesn't mean though that it isn't something specific about that configuration like the audio driver, or a future Kontakt 4.01 patch might "fix" the problem. We're not able to narrow down the actual problem without having all the source-code available for every component of the system to allow us to isolate every component and provide an accurate scientific "control" measurement to compare against.

[Guenon]

Definitely, I agree. If people are running Kontakt 4 with the same libraries and configuration, 10% seems like it's genuinely being observed.

I would imagine they are using the test methodology of DAWBench VI as they declare, but with the latest version of Kontakt. Doesn't say

But yeah, from the actual user's (the actual musician's/producer's) point of view, "having all the source-code available for every component of the system to allow us to isolate every component and provide an accurate scientific "control" measurement" is not what they should be worried about at all, like you have pointed out yourself, of course. The relevant information is: is it likely that the system I use for my work will function worse, and is that decrease big enough to matter in my use case. Information like this can be gathered separate from the technical analysis of what exactly causes it and how it can be engineered into a better state.

[aciddose]

I doubt that the bench-mark results accurately portray a realistic use-case for the hardware or software. So at best this should inspire people to always give themselves extra "CPU headroom". So it only takes a tiny nibble out of your headroom and doesn't affect your actual work at all.

10% is extreme though, so hopefully such numbers encourage people to bench-mark this themselves using their actual projects and actual systems. Jumping in to applying a microcode patch ("bios") is a crazy thing to do no matter what you're using a system for and it looks like these results apply only to that particular action taken on that particular CPU.

Best advice: don't take actions you don't need to take without having evidence demonstrating their necessity and cost vs. benefit first! A firmware patch is: going nuclear!

As far as I can find, there is no forced (via windows update) microcode patch. It seems the microcode modification was only applied on boot and only temporary until the next boot. Confirmation of that would be nice, otherwise people better start wearing their tin-foil hats if they want to worry about this issue at all.

[Guenon]

Best advice: don't take actions you don't need to take without having evidence demonstrating their necessity and cost vs. benefit first!

For a DAW user, this is the most important thing to gather out of all this, in my opinion too. Somewhat disagree on the general benchmark doubt, as I've found that the DAWBench scores have pretty closely followed what I've seen in real world working scenarios on different setups, but that's anecdotal, heh.

Data like this will likely increase over time, and it's also likely that the situation keeps changing, the impact varies between system configurations, and so on. However, variations aside, if there's a demonstratable effect of this kind, then in my opinion, for someone whose work depends on using a DAW in this manner, it's perfectly okay to be asking some questions and maybe even worrying a bit about this

It's exactly the case of "don't take actions you don't need to take without having evidence demonstrating their necessity and cost vs. benefit first" -- and only by asking those questions and having motivation to find out about the situation (a certain amount of worry often being a catalyst for that) such an informed decision can be made.

[metamorphosis]

If you're running Win7 and you have installed the Jan/Feb 2018 updates which prevented Meltdown/etc - just make sure you've updated with the March update - otherwise...
https://blog.frizk.net/2018/03/total-meltdown.html?m=1

[Zombie Queen]

ROTFL

[standalone]

For all of you processor vulnerabilities lovers, stay tuned for the next big hit: Hyper Threading (mainly Intel) bugs that lead to cache timing attacks, like the Spectre ones. It's under embargo, I don't know for how long.

[standalone]

https://www.techspot.com/news/75240-res ... l-cpu.html

Intel downplays the first hyper-threading vulnerability that has been presented. Let's see what happens after Black Hat in August.

[camsr]

https://www.itwire.com/security/83347-o ... u-bug.html