Recently discovered Intel CPUs processor design flaw

Mentioned in the article
https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

A fundamental design flaw in Intel’s processor chips has forced a significant redesign of the Linux and Windows kernels to defang the chip-level security bug.

Programmers are scrambling to overhaul the open-source Linux kernel’s virtual memory system. Meanwhile, Microsoft is expected to publicly introduce the necessary changes to its Windows operating system in an upcoming Patch Tuesday: these changes were seeded to beta testers running fast-ring Windows Insider builds in November and December.

Crucially, these updates to both Linux and Windows will incur a performance hit on Intel products. The effects are still being benchmarked, however, we’re looking at a ballpark figure of five to 30 per cent slow down, depending on the task and the processor model. More recent Intel chips have features – specifically, PCID – to reduce the performance hit.

Windows 10 computers will be automatically updated at some point, unless we keep it offline, which is impossible for many.
I wonder if Steinberg guys have some recommendation about it, what should we possible expect or do?
maybe other comments?
this situation smells bad…

Yes, rumour has it Jan 9th will be d-day for the MS update. The Mac fix already went live December 6th with MacOs 10.13.2.

Time to put a firewall change in to block my studio PC from getting outside my network! (still need NAS access) that said, I’ve finished all the tracks I’ve been working on lately, so if I took a performance hit right now - it wouldn’t stop me completing an already in-flight project. Not impressed though - should have stuck with AMD! :imp:

Bought my Ryzen 1700 this past Friday :slight_smile:

Having said that though the worry now for many of us is that the patch solving the problem with Intel CPUs actually may cause performance issues for AMD users as well…

Various sources report that AMD also has this security issue.

Intel have cleverly conflated Meltdown (the really serious Intel bug) with spectre (the pervasive but hard to exploit vulnerability) even putting the serious one last calling it “variant 3”, the industry have bitten and the Intel’s view of the world is becoming fact!

Of course kernel vendors will only want one version of a kernel, so if they patch this generically - all CPUs will take the hit. Shameful really because they’ll be slowing AMD CPUs when there’s no need. If there’s any justice in the world after the initial panic, code will be refined to allow the kernel to run in it’s old non-isolated state on AMD machines.

For the spectre attacks on cookies turn first party isolation on in your browser - for firefox do this:
https://www.ghacks.net/2017/11/22/how-to-enable-first-party-isolation-in-firefox/

Looks like the patch for Win 10 was released yesterday at 5pm eastern. I haven’t had a chance to see what the impact is.

I know someone who tried a game benchmark after the patch and the impact on CPU performance was minimal, and in some tests he actually scored better (???). This probably wont affect DAW users much either.

Games are generally tight loops of code with not much i/o going on (once all textures are loaded) - a DAW is running hundreds of separate programs as well as streaming possibly hundreds of channels of i/o - all which will require fresh calls to the OS kernel - I would expect DAWs to be one of the (single user) worse case scenarios for the performance hit by isolating kernel memory from tasks.

This is a concern for me too. How much of a hit will the fix cause, 5%? 50%? And will that hit cause knock-on bottlenecks to real-time processing?

You could keep your studio PC offline - ie: don’t accept the update.

It is easy to block windows update with a firewall like tinywall. It is a front end to windows firewall and it uses no cpu cycles at all.

Well I have just bit the bullet and patched my main DAW.

I ran passmark before and after the patch and there was virtually no difference in the results for all categories, CPU, 2D GFX, 3D GFX, memory or Disk. In fact the disk result was 5% faster after the patch.

I then fired up a couple of Cubase 9.5Pro stress test projects which I made a while ago to exercise the machine and check for weaknesses. One has multiple large VSTi samplers, including 17 instances of Ivory with the largest sample sets all playing concurrently, the other with many, many VST fx. All still play perfectly after the patch, and I could not detect any difference in CPU performance since before the patch.

So from these initial tests, it appears there is nothing to fear from these patches. Considering these will be the initial rushed emergency patch, it is likely that they will if anything get even better when the vendors have had more time to analyse the situation fully.

I was fearful about these patches as only 2 weeks ago I got brand new kick butt hardware especially to run Cubase, and would not have been amused if my considerable investment had been somewhat curtailed. But it seems not to be the case, and I can detect no operational differences after the patch.

It seems that the worst affected people are servers and server farms, so hopefully Cubase users will continue to be OK.

I patched my machine and did a test. My test project with a bazillion plugins hovering at 98% average asio utilization stayed the same and played with no glitches. The project has probably 60 insert plugins and more than a dozen VSTi including 3 Kontakt instances loading instruments from a NAS. Although the project played fine, Samsung Magician reported my SSD IOPS dropped 30%, I haven’t done the exact math but write IOPS went from 51K to 33K. I can believe this, as my project took forever to load. I suspect my network performance has been substantially impacted also, but I haven’t tested that. l I have a dual Xeon system with ancient processors. Old CPUs (with no CPID support) were expected to be impacted more severely than a newer processor would be. In any case, as long as the test projects behaves as before, which it did, I’m OK.

-E

@SledDriver @eli_lilly
thank you for report guys.

I have just updated as well. Compared performance of one of the recent projects. It has one Kontakt instance, one Superior drummer 3 and several audio tracks with lots of plugins. Before patch WIndows CPU meter showed 34-35%useage after patch it’s something around 35-36% not big deal after all.
Will be testing more and report back if something significant comes up.

No change in Cubase performance here after patch

Hippo

I’ve disabled windows update for now. I’ve heard mixed replies from people who’ve had the patch. A common complaint seems to be SSD read performance being hit to a noticeable degree for some. The other thing was a slight increase in asio load and CPU usage (1-2%). Not heard many people experiencing other problems so far.

Well, let’s all hope the performance impact has been over-egged, in theory lots of i/o and lots of threads means a lot of context switches - as I said before, DAWs are all about context switches which is why I was pessimistic about performance, but if in the big scheme of things this only adds a couple of percent of CPU load, most of us probably won’t notice. (fingers crossed!)

Similarly, I updated with the patch earlier today on a Win 7 Pro machine. There was no perceivable performance degradation, in my case with Nuendo version 8.1.

Anyone know what’s the name of the windows update/patch?

Ditto. And is there a way to limit the patch update on my W10 Pro system to security patches only - excluding MS from slipping in things like Fall Creators Edition?