• Sign in to Neowin Faster!

    Create an account on Neowin to contribute and support the site.

Sign in to follow this  

Microsoft Kernel Patch CPU Before and After Benchmarks Thread

Recommended Posts

Radium    126

Tasks such as calculating Pi shouldn't show any negative effects since calculating Pi doesn't require any system calls.

The performance hit comes from context switching which you essentially never do when you only calculate/estimate numbers.

Share this post


Link to post
Share on other sites
LimeMaster    15,600

The patch seems a tad glitchy on older CPUs.

Spoiler

Before:

LdfhB9y.png

After:

IPvdvWg.png

 

Share this post


Link to post
Share on other sites
adrynalyne    11,698
23 minutes ago, LimeMaster said:

The patch seems a tad glitchy on older CPUs.

  Hide contents

Before:

LdfhB9y.png

After:

IPvdvWg.png

 

What does that test even do? Performance is not hampered across the board. 

Share this post


Link to post
Share on other sites
LimeMaster    15,600
1 minute ago, adrynalyne said:

What does that test even do? Performance is not hampered across the board. 

I believe it's suppose to benchmark the CPU.

Share this post


Link to post
Share on other sites
adrynalyne    11,698
18 minutes ago, LimeMaster said:

I believe it's suppose to benchmark the CPU.

I’m sure, but what is it actually doing to test it?

Share this post


Link to post
Share on other sites
LimeMaster    15,600
13 minutes ago, adrynalyne said:

I’m sure, but what is it actually doing to test it?

No clue. There doesn't seem to be documentation of what the tests entail.  Both tests (stress and bench) have similar results.

Share this post


Link to post
Share on other sites
adrynalyne    11,698
8 minutes ago, LimeMaster said:

No clue. There doesn't seem to be documentation of what the tests entail.  Both tests (stress and bench) have similar results.

If its just calculating pi, for example, the patch won't affect it at all.

  • Like 1

Share this post


Link to post
Share on other sites
LimeMaster    15,600
12 minutes ago, adrynalyne said:

If its just calculating pi, for example, the patch won't affect it at all.

I figured the tests weren't exactly an accurate representation when comparing the results. It however did have a slight effect on the Multi Thread Ratio. It decreased after the patch was installed.

Share this post


Link to post
Share on other sites
neufuse    3,648

It appears my SQL server 2017 benchmarks on windows actually saw a 2% performance increase after the patch :huh: In Linux SQL Server 2017 shows a 25% DECREASE in performance... ugh

 

SQL Server does a lot of IO and system calls.... doing mass IO bench marking should show any problems with this patch 

  • Like 2

Share this post


Link to post
Share on other sites
LimeMaster    15,600
45 minutes ago, neufuse said:

It appears my SQL server 2017 benchmarks on windows actually saw a 2% performance increase after the patch :huh: In Linux SQL Server 2017 shows a 25% DECREASE in performance... ugh

 

SQL Server does a lot of IO and system calls.... doing mass IO bench marking should show any problems with this patch 

I understand why Linus is angry then.

Share this post


Link to post
Share on other sites
adrynalyne    11,698
Just now, LimeMaster said:

I understand why Linus is angry then.

So here is the thing--is he angry because he was unable to get performance better, vs. Windows?

Share this post


Link to post
Share on other sites
LimeMaster    15,600
1 minute ago, adrynalyne said:

So here is the thing--is he angry because he was unable to get performance better, vs. Windows?

That would be my guess. Of course he blames Intel because everything was technically running smoothly beforehand.

  • Like 1

Share this post


Link to post
Share on other sites
+fusi0n    2,000

Ouch..

 

2018-01-06 22_22_11-PerformanceTest 8.0 Evaluation Version.png

 

I'm building my AMD rig now..

Share this post


Link to post
Share on other sites
Mockingbird    2,552
22 minutes ago, Yogurth said:

Updated Techspot analysis and benchmarks. Up to 5% drop in overall performance aside from massive SSD slowdown. Still no Windows server workload tests apart from the Epic Games login server meltdown if that is a Windows server at all.

 

 

https://www.techspot.com/article/1556-meltdown-and-spectre-cpu-performance-windows/

That's without the microcode update.

Share this post


Link to post
Share on other sites
adrynalyne    11,698
2 minutes ago, Mockingbird said:

That's without the microcode update.

There hasn’t been enough testing with the microcode update to verify it. There is also a screenshot out there showing a massive performance hit after the meltdown patch on an AMD cpu. Just one that I saw. 

Share this post


Link to post
Share on other sites
Mockingbird    2,552
5 minutes ago, adrynalyne said:

There is also a screenshot out there showing a massive performance hit after the meltdown patch on an AMD cpu. Just one that I saw. 

...except that, on Windows, kernel VA shadow is off when using AMD processors.

 

Linux also disables kernel page-table isolation on AMD processors.

Share this post


Link to post
Share on other sites
Yogurth    2,213
19 minutes ago, Mockingbird said:

That's without the microcode update.

Yup, this is Windows patch + BIOS update. The article is a work in progress and Techspot will update it when a new batch of patches arrives.

Share this post


Link to post
Share on other sites
adrynalyne    11,698
34 minutes ago, Mockingbird said:

...except that, on Windows, kernel VA shadow is off when using AMD processors.

 

Linux also disables kernel page-table isolation on AMD processors.

Show me where it is turned off and causing a performance hit please. The specific indicator if you don't mind.

 

 

Linux may disable kpti on AMD systems, but the NT kernel is not Linux.

 

Share this post


Link to post
Share on other sites
Mockingbird    2,552
3 minutes ago, adrynalyne said:

Show me where it is turned off and causing a performance hit please. The specific indicator if you don't mind.

 

 

Linux may disable kpti on AMD systems, but the NT kernel is not Linux.

 

It's turned off, so no performance impact.

 

PowerShell

 

Install-Module SpeculationControl

 

Set-ExecutionPolicy Bypass

 

Get-SpeculationControlSettings

Share this post


Link to post
Share on other sites
adrynalyne    11,698
2 minutes ago, Mockingbird said:

It's turned off, so no performance impact.

 

PowerShell

 

Install-Module SpeculationControl

 

Set-ExecutionPolicy Bypass

 

Get-SpeculationControlSettings

I'll wait for your results patiently. I've got mine here. I want to see where Windows turned off kernel va shadowing post patch.

Thats what you implied with what you said.

Share this post


Link to post
Share on other sites
Mockingbird    2,552
2 minutes ago, adrynalyne said:

I'll wait for your results patiently. I've got mine here. I want to see where Windows turned off kernel va shadowing post patch.

SC.png

Share this post


Link to post
Share on other sites
adrynalyne    11,698
Just now, Mockingbird said:

SC.png

K.

 

Here is mine. Kernel VA shadowing was never enabled, even pre-patch.

 

Capture.PNG

Share this post


Link to post
Share on other sites
Mockingbird    2,552
1 minute ago, adrynalyne said:

K.

 

Here is mine. Kernel VA shadowing was never enabled, even pre-patch.

 

Capture.PNG

Which processor?

 

On Intel processors, Kernel VA shadowing is disabled if incompatible Anti-Virus is detected.

 

On AMD processors, Kernel VA shadowing is disabled.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.