Performance Impact on Windows Systems with Spectre and Meltdown Mitigations

[biffzinker]

Quoting this from the source link at the bottom of my post.

In general, our experience is that Variant 1 and Variant 3 mitigations have minimal performance impact, while Variant 2 remediation, including OS and microcode, has a performance impact.

Performance impact summary:

• With Windows 10 on newer silicon (2016-era PCs with Skylake, Kabylake or newer CPU), benchmarks show single-digit slowdowns, but we don’t expect most users to notice a change because these percentages are reflected in milliseconds.
• With Windows 10 on older silicon (2015-era PCs with Haswell or older CPU), some benchmarks show more significant slowdowns, and we expect that some users will notice a decrease in system performance.
• With Windows 8 and Windows 7 on older silicon (2015-era PCs with Haswell or older CPU), we expect most users to notice a decrease in system performance.
• Windows Server on any silicon, especially in any IO-intensive application, shows a more significant performance impact when you enable the mitigations to isolate untrusted code within a Windows Server instance. This is why you want to be careful to evaluate the risk of untrusted code for each Windows Server instance, and balance the security versus performance trade-off for your environment.

For context, on newer CPUs such as on Skylake and beyond, Intel has refined the instructions used to disable branch speculation to be more specific to indirect branches, reducing the overall performance penalty of the Spectre mitigation.

Source: Understanding the performance impact of Spectre and Meltdown mitigations on Windows Systems

 

[er557]

I see no performance deficit, dual haswell-ep xeon, but I dont think the full mitigation is already enabled. Even with windows update 16299.129 and intel microcode update 8th january, the powershell script says only meltdown is mitigated, and not due to lack of OS or hardware support, so I dont think they fully enabled that. Performance is the same, no reboots, all is well.

 

[Splinterdog]

There's a handy checker available and you can read about it here.

 

[silkstone]

I get a ~30-40% drop in mining speed while cpu mining XMR.
This is only after the microcode update.

 

[HTC]

There's this dude @ anandtech's forum that has an i7-860, an i5-8600k and a 2500k and he did a "few" benchmarks:

post #1

post #2

post #3

post #4

Dude didn't make the benches on the 2500k yet because it seems he's using that particular PC for something, after which, he'll do them benches.

This is a user's performance changes pre VS post spectre / meltdown security patches, as opposed to a "professional review".

 

[John Naylor]

Wow ... wish he would hav eposted a summary ... the reader risks RDI with that clickfest,

 

[HTC]

Wow ... wish he would hav eposted a summary ... the reader risks RDI with that clickfest,

Dude's limited with the amount of pics he can post in a single post, which is why it's spread across 4 posts. That said, @ least the dude made an effort to show us, from an user's perspective, how these patches affected his CPUs, which span a few generations, and the 2500k is still missing it's benchmarks.

EDIT

From the same topic @ Anandtech, in the next page, from a different user:

VMware has pulled all patches that apply Intel's microcode updates, and recommends that users do not install them. If they have already been installed, you can't reverse the process, so they recommend disabling speculative execution entirely on every affected ESXi Host and rebooting every VM in that environment, then waiting until Intel releases more microcode updates, applying those, undoing your configuration change on your ESXi Host, and then rebooting every VM in your environment again. So everyone with VMware who's already applied patches is now looking at 2 guaranteed separate, and manually performed total VM outages as a reward for applying Intel's microcode updates.

 

[Hugis]

Ok these are my personal tests and as such I wont be hassled about it to much by you lot

This is the update i downloaded from Microsoft

the one at the bottom for me 64bit intel/amd


Specs of my PC:

System Name Office/Gamer Mark IIProcessor i7 4770K @ 4.2GhzMotherboard MSI Gaming 5 Z97Cooling Raijintek Aidos, MX-4Memory 16GB Crucial Ballistix Tactical BLT2K8G3D1608ET3LX0Video Card(s) EVGA GTX 780 Ti ClassifiedStorage Samsung 850 Evo 120Gb,SAMSUNG HD502HJDisplay(s) LG W2261VP @ 1920 * 1080Case Corsair Carbide 200RAudio Device(s) Realtek ALC1150 (On Board)Power Supply Corsair TX750WSoftware Windows 10 Pro 64Bit


I also found this, a great little util to check your system!

https://www.grc.com/inspectre.htm

First things first i have only just installed the patch and as i can tell at the mo, it seems similar in speed to before.

Samsung magician

Pre

post

make of that what you will, as said this is my own system so who knows what you will get!

next up Unigine Valley with an Oc'd 780ti (thank you TPU!)

pre

post

Well again who the feck knows what to make of that mabe my rooms cooler today.....


and i know this is sumwhat taboo but i did test my system on a online tester the other day just to check and

pre

UserBenchmarks: Game 77%, Desk 84%, Work 63%
CPU: Intel Core i7-4770K - 89.6%
GPU: Nvidia GTX 780 Ti - 79.1%
SSD: Samsung 850 Evo 120GB - 94.6%
HDD: Seagate Momentus 2.5" 500GB - 33%
RAM: Crucial BLT8G3D1608ET3LX0. 2x8GB - 75.8%
MBD: MSI Z97 GAMING 5 (MS-7917)



post

UserBenchmarks: Game 77%, Desk 82%, Work 62%
CPU: Intel Core i7-4770K - 86.5%
GPU: Nvidia GTX 780 Ti - 79.1%
SSD: Samsung 850 Evo 120GB - 92.6%
HDD: Seagate Momentus 2.5" 500GB - 33.5%
RAM: Crucial BLT8G3D1608ET3LX0. 2x8GB - 76%
MBD: MSI Z97 GAMING 5 (MS-7917)


Again its seems it should be slower but.......my guess its a small hit and as i person that has "worked on 386s+/amigas/st's" i do know a few onions


I dunno if ill get a bios update on this Z97 - might be an unoffical hack who knows, like the ip35e ahci sata hack on my board from a few years back-ill see what happens..


I hope that helps someone in our community somewhere, but as an average joe with semi old hardware i thought i would run a few test just to see what if anything happens.

 

[eidairaman1]

@biffzinker, there seem to be 2 threads relating to the same exact thing, i wonder if they should be combined

 

[RejZoR]

@Hugis
The "InSpectre" tool doesn't seem to allow me to enable "Enable Spectre protection", button is unclickable... Am I missing something here or is that just a placeholder or something?

 

[Hugis]

@Hugis
The "InSpectre" tool doesn't seem to allow me to enable "Enable Spectre protection", button is unclickable... Am I missing something here or is that just a placeholder or something?

Unfortunately mines grayed out

 

[HTC]

To my understanding, these problems affect more datacenters as well as those programs (games included) that use disk communication very often.

This reminds me of something: SSD speeds are affected in some of the benchmarks but, does anyone know if and how are "regular" HDDs affected, or is this just an "SSD thing"?

 

[Hugis]

To my understanding, these problems affect more datacenters as well as those programs (games included) that use disk communication very often.

This reminds me of something: SSD speeds are affected in some of the benchmarks but, does anyone know if and how are "regular" HDDs affected, or is this just an "SSD thing"?

just doing my own personal test here @ home on old-ish computer

 

[HTC]

just doing my own personal test here @ home on old-ish computer

My reply wasn't directed specifically @ you.

Please look @ the CrystalDiskMark results from the 2nd to 5th pics of this post: 2nd and 4th pics are after patches while 3rd and 5th pics are pre patches.

As you can see, that's quite a hit, specially on the 850 EVO.

I was wondering it this problem also affect regular HDDs or only SSDs are affected.

 

[Ahhzz]

... the 2500k is still missing it's benchmarks.

..:

waiting for that....

.... If they have already been installed, you can't reverse the process...

..:

That's 25% of the last 30% of the reason I didn't update mine.

 

[HTC]

waiting for that....

Dude said this, right in his 1st post:

My 2500k will have to wait for a while, because I am running some other projects at the same time and I need to finish up with the i5-860.

How long will it take? No idea!

 

[Steevo]

To my understanding, these problems affect more datacenters as well as those programs (games included) that use disk communication very often.

This reminds me of something: SSD speeds are affected in some of the benchmarks but, does anyone know if and how are "regular" HDDs affected, or is this just an "SSD thing"?

The whole I/O system is effected, as speculative prediction requests information from any memory address range, so every disk transaction SSD, or HDD, are going to see worse performance as each transaction now has to be verified to make sure its not reading out of bounds. That takes CPU cycles as the "host controller" is using the system CPU, but even those with RAID cards will have some slowdown as the verification still has to take place, its impact will be slightly less.

 

[Vario]

Do you trust the Spectre/Meltdown verification software? I am really hesitant to run this kind of thing, it would be nice if Microsoft released a tester that wasn't a Powershell script. When I tried running their script it said the module wasn't readable or missing files or similar, I do not remember the exact error.

 

[RejZoR]

Unfortunately mines grayed out

To my knowledge, this just looks like a tool to disable the patches on demand. Meltdown patches have been released so you have a button for that, Spectre wasn't yet, which is why it's greyed out.

 

[Hugis]

To my knowledge, this just looks like a tool to disable the patches on demand. Meltdown patches have been released so you have a button for that, Spectre wasn't yet, which is why it's greyed out.

I disagree but each to there own, it tells me if there will be a impact to my system and what is covered by what ive done..

 

[FireFox]

@Hugis
The "InSpectre" tool doesn't seem to allow me to enable "Enable Spectre protection", button is unclickable... Am I missing something here or is that just a placeholder or something?

Here is mine:

I can disable or enable Meltdown and Spectre Protection

To my knowledge, this just looks like a tool to disable the patches on demand. Meltdown patches have been released so you have a button for that, Spectre wasn't yet, which is why it's greyed out.

I guess you are right:

 

[BMfan80]

My 4k writes on my 960 Evo and Pro went from 174 to 141.
The 4k reads went from 54\53 to 51\50.

I was using W10 with all the update and the microcode update.

 

[Ahhzz]

Dude said this, right in his 1st post:



How long will it take? No idea!

I understand that, and didn't ask

 

[RejZoR]

@Knoxx29
How did you make Spectre thingie not being vulnerable to?

 

[FireFox]

@Knoxx29
How did you make Spectre thingie not being vulnerable to?

Only thing i did was to install the Windows's update and the Bios that Asus released for the motherboard.