Windows 10 CPU Bug Fix Patch Benchmarks (updated)

WareTernal

2018-01-04 20:33

chispy:

It seems on Ryzen systems the performance hit it's bigger than on intel

Sounds like sloppy work by MS to include AMD in this mess. It's a shame if Ryzen is taking a hit for no reason.

tsunami231:

im sorry but those test before and after look like there within "margin" of error tell said people that response able for the flaw and fixing to fix it they have till so and so time to do so or this info will be released

Some of those results are within margin of error, but others have found similar results. If it was margin of error we would expect to see some higher numbers post patch. I think they did wait until patches were available before public disclosure.

Alessio1989:

The wolves are gonna be in cluster virtual machine environments, not for +/- 2% on futuremark.

What he said^. I appreciate the demonstration that this isn't going to affect my games, but what about a VM host with 10 VM's?

#5507360

Primey0

2018-01-04 20:49

Ummmmm.... In order to full secure your PC you also need to update the CPU firmware. That patch that Intel provides will be the one that impacts performance in certain workloads. Did you do this before the benchmarks? Because I think you're only benchmarking the windows 10 patch which isn't the patch that impacts performance

#5507382

SoloCreep

2018-01-04 21:46

8700k here and no difference in FPS.

#5507387

jaggerwild

2018-01-04 21:52

foxx1337:

So I only performed one test on my system, and I would say that I get like 10% worse performance. Core i7 2600K at 4 GHz, samsung 850 EVO ssd. The test is, in a ConEmu terminal under Windows, under Bash (from git-bash), I type this: Weasis compilation: Before patch: 53 seconds After the patch: 105 seconds Weasis is a Java viewer for X-Ray, CT-scan and CAT-scan file formats. This ConEmu + MinGW + bash + Maven + javac scenario probably touches quite a few different aspects of the operating system, from file access, to lookup, writes, compilation, the (few) Weasis tests, etc. I tested by restarting the computer, opening ConEmu and running the clean + install cycle 5-6 times until it stabilized.

Nobody cares!!!!! They already stated the media over rated it to hype it up for hits, also Im not doing shoddy stuff on the internet. The patch could be a spy, good work man!

#5507394

chispy

2018-01-04 22:08

Hilbert if you can , test an AM4 Ryzen or Treadripper system on an SSD or NVMe. It seems Ryzen systems have taken a bigger hit on performance than Intel. Myself and a lot more members can confirm this by our own testing , it's all over the internet by now that degradation in performance on Ryzen systems is worst and more noticable than Intel.

#5507395

Kaarme

2018-01-04 22:15

chispy:

Hilbert if you can , test an AM4 Ryzen or Treadripper system on an SSD or NVMe. It seems Ryzen systems have taken a bigger hit on performance than Intel. Myself and a lot more members can confirm this by our own testing , it's all over the internet by now that degradation in performance on Ryzen systems is worst and more noticable than Intel.

And it shouldn't even affect AMD's CPUs. If I didn't know MS is full of incompetents, I'd tighten my tinfoil hat and say Intel ordered such an outcome. But since I do know MS is a hive of incompetence, I'll just say they simply didn't know what they were doing, as usual.

#5507397

AsiJu

2018-01-04 22:19

MS is incompetents run by Intel :P I am kidding. Or am i...

#5507398

mtrai

2018-01-04 22:24

JUst wanted to point out...it appears that the MS KB patch that can be downloaded only install the update but does not activate it. MS has issued client guidance today with some Power Shell script to see what is and what is not enabled from this patch. MS link here: (Powershell info is at the bottom to check) https://support.microsoft.com/en-us/help/4073119/windows-client-guidance-for-it-pros-to-protect-against-speculative-exe

#5507404

Hilbert Hagedoorn

Administrator

2018-01-04 22:42

mtrai:

it appears that the MS KB patch that can be downloaded only install the update but does not activate it.

There are specific conditions where the patch would not all protections would activate, and that is with certain anti-virus software installed. My test rigs do not have that. The page you refer to only shows a few PowerShell commands to verify the patch is active. New firmware, however, might be required, as mentioned in the article. I check/verify this a bit more tomorrow (it's almost 11pm here).

#5507405

mtrai

2018-01-04 22:58

No problem...I was just pointing out a bit more confusion on this in hope you or others would be able to check. Anyhow my Skylake system is enjoying a vacation and I am on my Ryzen build and I am on Win 10 Insider build 17063 so did not get the update nor can I run that powershell command it seems. Get some rest.

#5507416

DashD

2018-01-04 23:28

Hilbert Hagedoorn:

There are specific conditions where the patch would not all protections would activate, and that is with certain anti-virus software installed. My test rigs do not have that. The page you refer to only shows a few PowerShell commands to verify the patch is active. New firmware, however, might be required, as mentioned in the article. I check/verify this a bit more tomorrow (it's almost 11pm here).

Are you running latest bios 1003 for the NVME tests? I've heard a report that the patch is not active before that version, but I don't know if that's true. My own tests show a 20% drop on the 4k read/write 🙁

#5507428

chispy

2018-01-05 00:35

DashD:

Are you running latest bios 1003 for the NVME tests? I've heard a report that the patch is not active before that version, but I don't know if that's true. My own tests show a 20% drop on the 4k read/write 🙁

PC specifications ? That info might help a lot. 20% drop it's a lot.

#5507436

DashD

2018-01-05 01:07

chispy:

PC specifications ? That info might help a lot. 20% drop it's a lot.

Intel 900p, 8700K, Maximus X Hero

#5507440

The Goose

2018-01-05 01:14

Given that bench marking an ssd can give very random scores, i know my Samsung pro 960 can vary from 2200 -2500mbs seq read so I think its safe to say that the difference in the bench marks is within margin of error and everyone can stop panicking for the sake of it.

#5507443

tensai28

2018-01-05 01:43

Primey0:

Ummmmm.... In order to full secure your PC you also need to update the CPU firmware. That patch that Intel provides will be the one that impacts performance in certain workloads. Did you do this before the benchmarks? Because I think you're only benchmarking the windows 10 patch which isn't the patch that impacts performance

This. We are probably drawing conclusions too soon.

#5507444

Aitortxo

2018-01-05 01:44

Thanks Hilbert for the tests but as someone already commented, I believe it's a little early to come to any clear and definitive conclusion. We are still missing the firmware patches from the CPU and who know what else... Also, a test on Windows 7 with more down to earth hardware would be appreciated. Take some rest and keep up the good work!

#5507490

Darkiee

2018-01-05 05:02

I´ll say this here too, using 2600K on W10 and i found nothing to be worried. I got same thing aswell, TimeSpy gave abit better result.

#5507494

sykozis

2018-01-05 05:29

H83:

Nice to see Hilbert already posting some benchies about the patch. But i have to say i´m surprised with the results after all the talk that performance would tank and the difference is around 2% in the end... Did people overblown the issue or should we wait for further updates/patches that may cripple the performance severely??? Thanks for the quick nice work Hilbert!

Not everyone is going to see noteworthy performance impact. The large performance impacts are very conditional. Some database software is showing performance impact in the 20%+ range according to Phoronix. It's possible that updates to the most heavily affected software might mitigate some of the performance loss. In any event, security should be more important than performance.

#5507496

Carfax

2018-01-05 05:55

DashD:

My own tests show a 20% drop on the 4k read/write 🙁

I just tested that on my PC, and I had no performance hit at all for 4K reads and writes. In fact, writes were slightly higher than before the patch.

#5507500

sykozis

2018-01-05 06:55

jaggerwild:

Nobody cares!!!!! They already stated the media over rated it to hype it up for hits, also Im not doing shoddy stuff on the internet. The patch could be a spy, good work man!

The media didn't "over rate" the impact. The impact is very workload dependent, as stated in every report I've read. Phoronix has noted performance impacts upwards of 20%+ in some database software. Just because it doesn't affect a certain subset of users, doesn't mean that the performance impact, as a whole, is overrated.