Advice Request Microsoft Defender can Significantly Impact Intel CPU Performance, We have the Fix

Please provide comments and solutions that are helpful to the author of this topic.
Any AV slows down any computer and must have an impact on the computer resources. The issue reported in this thread is very small compared to the overall impact of Defender's real-time protection. If one has a problem with one particular AV then it is possible to simply change it to another.
 
That's a likely behaviour for third party antivirus. The bigger problem is Windows's integrated antivirus which slows PC the same or even more than third party solutions. It's harder to control. You can't be sure what's it doing in background and on low end (or older) machines it sometimes makes a disaster in terms of CPU and disk usage leaving those machines practically unusable.
I noticed such behavior on the low-resource computer when the machine was updating the system or Microsoft Store apps. But, after finishing the updates and .NET optimization, the machine is mostly usable (with an Intel Atom processor).
The performance of Windows Defender (now Microsoft Defender) is well known and well tested. Some time ago I made a thread that seems to be still actual:

Over the years I could change two things:
  • The false positives rate of Microsoft Defender is one of the best, for several months (starting from the year 2021).
  • The Defender's detection is now slightly better (as compared to 3rd part AVs) than it was in the year 2018.
 
Last edited:
This issue affects AMD as well, contraty to what the article states, this is a normal issue on how WD works.

Difference is AMD CPUs are better tailored for multi-tasking, so the performance drop is not noticeable like on Intel.
That is true. I did not notice it.
 
That's a likely behaviour for third party antivirus. The bigger problem is Windows's integrated antivirus which slows PC the same or even more than third party solutions. It's harder to control. You can't be sure what's it doing in background and on low end (or older) machines it sometimes makes a disaster in terms of CPU and disk usage leaving those machines practically unusable.
Agreed. In my case the third party AVs I've used in past few years namely Kaspersky IS and free, Eset IS and F-secure safe all had significantly better performance compared to WD. Among them Kaspersky IS isn't known to be low on resources and yet it outperformed WD significantly for me. There has been enough reports that its not just small number of users but a good number of them facing this issue so people shouldn't just dismiss it. They have made great progress on protection but the performance is just sad. I don't have enough money to buy a new device just so my device wouldn't struggle running Windows defender.
 
This is absolutely true and TechPowerUp's counter control utility does fix the issue for me as well on i9 10900K system. My Cinebench score improved by almost 1000 points after resetting counters.

Before resetting counters on counter control:
Screenshot 2022-11-27 190813.png


After resetting counters on counter control:
Screenshot 2022-11-27 191011.png


So yes, the fix does work, absolutely. Anyone here saying otherwise has clearly not tried it themselves, or owns a CPU that's not affected (likely an AMD CPU). I also noticed when testing that before I reset the counters, Defender used 2% of my CPU during Cinebench run. After I reset the counters, Defender only used 0.1% of the CPU during the Cinebench run. So quite clearly it is Defender causing the issue and this fix does indeed fix the aforementioned problem.

Quite pathetic that Microsoft and Intel haven't fixed this issue by now. This bug must've been there for quite long time before it was even found. Light users probably have never noticed it, but under heavy stress its been noticeable for me on many Intel based systems. Its why I kept using ESET, but after this fix I've been using Defender again and its now actually very light.

You can download counter control from here
 
Cinebench R23 multi core bench:
1. Disabled all MD modules - 4780 points.
2. Enabled all MD modules - 4785 points.

NovaBench results:
1. Disabled all MD moduler - 2131 point (cpu score 820)
2. Enabled all MD modulels - 2198 points (cpu score 906)
edit: Tested with InSpectre too (enabled/disabled) - same close results (+/- few points).

Screenshot (39).png
Screenshot (40).png
 
Last edited:
This optimization guide recommends disabling Cloud-delivered protection, Automatic sample submission and Tamper Protection. Will Defender be much worse after doing this or will it still be decent with common sense?

 
  • Like
Reactions: [correlate]
This optimization guide recommends disabling Cloud-delivered protection, Automatic sample submission and Tamper Protection. Will Defender be much worse after doing this or will it still be decent with common sense?
You'll castrate Defender and essentially be without any meaningful protection.
 
They don't provide any kind of official documentation for the "bug", only test info, so I take this as a kind of scaremongering.
Kevin Glynn is totally legit. Developer of ThrottleStop. Dude has been carrying users for years. Users are such freeloading cheapskates that he couldn't even get donations despite his utility being crazy popular.

The fixes are in the article. He does not provide step-by-step how-to instructions. It's for power users.