Okay, so for those who already have the fix installed, what kind of performance hit are you seeing, anons?

Okay, so for those who already have the fix installed, what kind of performance hit are you seeing, anons?

I could run some more significant benchmarks but I can't say I really care either way. Right is obviously post fix.

Other urls found in this thread:

catalog.update.microsoft.com/Home.aspx
catalog.update.microsoft.com/Search.aspx?q=KB4056897
twitter.com/NSFWRedditVideo

here's mine, everything is within margin of error

Can't notice anything on my Core duo

wow its fucking nothing
and you fags were shrieking like a bunch of redditors about this

Its be obvious it was nothing for windows. People have been running this version for weeking with the windows insider program and none of them even noticed

this all stated five or so hours ago on Sup Forums, everyone was aware back then that service providers are the once most affected.

REEEEE normies!!!
This doesn't benchmark the perf impact of syscalls!!!

26 difference on cinebench is massive. You generally get this with a 100-200mhz overclocked.
I'm a 7700k owner by the way, not an AMD shill. Kinda worried.

26 is big in single core. Not multi. Try running it some time and see for yourself the different scores you get

Good idea, use a benchmark that avoids syscalls and I/O.

What do you do that uses lots of syscalls?

>1.20859% difference
reposting in this "aftermath" thread

Important thing is that intel loses stock, fuck their retarded monopoly

Intel still benchmark king

It is big in multi as well, I already tested my 7700k at all frequencies many times over from 4200 to 5000 and noted down all the results and there is never such a big difference.

Ive cinebenched my 8400 and gotten 50+ between high and low scores on multi. Looks normal to me

files, sockets etc

Never had more than 5 or so difference. I have the numbers right there. Pretty odd.

I would like to see a file transfer bench

I saw some user saying the NVMe drives are nearly halved in speed by this, does that have any truth to it? I ordered an intel cpu that shipped hours before I heard this news yesterday, so if it truly affects NVMe drives I'll return it. Otherwise I don't use VMs or anything so I'm fine. And if I did use VMs I would've gotten AMD in the first place.

My cpu is locked and has turbo though, which would have more flux than a fixed multiplier

NVMe are a meme anyway, just get an 850

>posting benchmarks that don't really have any context switching
Nice job shills, keep it up - You just each earned another stock option!

I can tell you are not pleased with the results because you called him a shill. Why are you mad that our pcs still work how we want them to?

At least read please. I'm posting alone as a casual PC user, gamer, some music, some screaming some video (audio mostly) editing etc. You gave me 2 (You)s and called me a shill because I'm not using 4 year old desktop CPU in some crazy science or datacentre work? Sorry, bro I'll leave that to you, OK? Chill

>REEEEEEEEEEEEEE this Exploits was supposed to be the one!!! NOOOOOO!!!!

let me spell it out for you cucks

1. windows benchmarks are not gonna show any change. the 30% number was always for linux not windows. windows has a fundamentally different architecture compared to linux and while the high level solution to the problem is the same, the implementation will likely be different than linux.

2. this issue does not impact many average goyims who use their computers to watch jewtube and porn. this issue impacts datacenters running massive servers for services like redis, databases, etc. this is where the major issues will occur.

try copying 1000 files off an ssd to another

I could see it causing an issue with a lot of small files. I feel doubt it would be enough to actually bottleneck my ssd and certainly not my hhd

Is everyone just testing that Insider build or has the actual update started rolling out? I haven't gotten anything yet but would like to keep an eye out and run some benchmarks too.

What is a rounding error in casual workloads becomes a disastrous problem for large enterprises who have to process a thousand I/Os 24/7. A 5% performance drop basically means your equipment needs an extra hour to do the same work.

>windows has a fundamentally different architecture compared to linux and while the high level solution to the problem is the same, the implementation will likely be different than linux
Are you saying that this patch doesn't address the exploit 3 patch on windows or that windows is fucked less than linux at the moment?

Where is the patch for windows 7?

Insider build has been out for weeks. Release came out today, but it wasnt pushed to my pc yet and i wasnt going to install it manually

The update is rolling out for Windows 10 but if you don't want to wait or using windows 7/8 just get it from here

catalog.update.microsoft.com/Home.aspx

He's saying the Windows kernel is fundamentally different from Linux and just because Linux saw a performance drop of (up to) 30% doesn't mean Windows would be affected in the same way. It could be better, it could be worse, it could go either way depending on the particular use case and so on.

M$ has been pretty calm during this whole debacle. Maybe they actually have things under control for once.

no. the windows patch addresses the same shit as the linux patch (which should be either 2 or 3 of the attack variants).

i am also saying windows isn't linux and their implementation of the mitigation is not going to match how the open source neckbeards did it. because the implementation is different, the performance hit will also be different. consequently, the 30% number you've heard thrown around will not apply to windows.

average kikebergs like you and me don't run tasks that use a shitload of syscalls. the performance dramatically drops when syscall usage becomes extremely large.

tl;dr: all the numbers and benchmarks you're seeing in this thread are entirely expected results and are useless. wait for the datacenter benchmarks to roll in. OR, wait for someone to do a real benchmark that involves context switching + syscalls (which is what kills the performance of the cpu).

I feel like i'm crawling on a 7700k. fresh windows.
what other shit is going wrong..

My AMD FX 8320 is unaffected, cba to get screenshots but I got 530cb before and after update.

Yeah, fuck moving files, right intelbros?

when might we start seeing these kind of syscall benchmarks?

Well, if you just installed the patch right now, it installs new .NET libraries after you boot, wait for it to finish before you run benchmarks.

>Not helpful: the post

I'm gonna run a test on my 7700HQ then do the update and test again. Any other tests I should run other than geekbench, 3dmark, Doom, PUBG, crystaldisk and a RAM test?

dx12 DXMD benchmark is the same 65.4FPS vs 66.8, heavily GPU cucked

dx11 DIRT Rally

old 100.26 average 130 max
new 98.37 average 128 max

gpu cpu?

im not a fucking psychic but my guess is over the next 2-4 weeks. these patches have been pushed pretty quickly. gotta give the datacenters time to roll out the distros to their nodes.

So I applied the new Windows update, and it just led me to a black screen, and I have to mess around with command prompt just to get access to my programs. Thanks Winblows. For goodness sake, I had to open my browser by running it from task manager after ctrl+alt+deleting to the options menu.

1070 stock
4790k stock

how do i install the fix on ubuntu?

do i select the intel micro code optional driver under additional drivers section of the software menu?

Where the fuck can I even find the fix? I've found 4 different articles linking me to the Win10 version when I need the Win& version, and the Windows website itself is obtuse as fuck. What ever happened to selecting Windows 7 from a dropdown list and it then displaying potential updates? Christ.

Well, it looks like difference for end users are minimal, good news.

Thanks for the info. I'm fine with waiting for it to pop up, just didn't know it had been released.

They had it in a November build of Insider, so they were probably just waiting to release it with everyone else.

Is this still under error margin?

Don't you love how the update notes don't actually address this issue in plain language at all?

is it out for windows?

Perfectly.
One went up, one went down.

With a locked mobile cpu? Certainly. Let it cool down and try it again lol

yes you can run it again and get different number

I don't use my machine for such workloads, it's mostly for gaming and running CAD/BIM/3DSmax. I knew the performance drop wasn't going to mean much to me, the thread is just to share impressions.

Oh fuck you guys. I actually sold all my shit with intel CPU already and ordered AMD replacements. I'm never going to trust you ever again.

Seriously, where the fuck are the Win 7 versions? All I can find are links to the Win10 version.

Yes. You can download manually.

catalog.update.microsoft.com/Search.aspx?q=KB4056897

No one uses them anymore, wait fro patch Tuesday if you can't google update yourself.

baka

DELETE THIS

...

what's the worse that can happen if I just opt out of the update?

My older machine

This is kind of wrong. The Windows and Linux patches will basically be the same, because there's only one workaround for the bug: to remove all the kernel mappings from a user process's address space. The OS internals have no real impact on the bug or the workaround. Within some fudge factor, the same workloads will see similar drops in performance in both Linux and Windows.

If you use your Intel CPU to compile large codebases, run database software, or host some heavily loaded web service, you're going to get dicked. Otherwise, you probably won't notice any difference.

Someone in the other thread said you would see it's effects more on an M.2 SSD compared to a SATA3 SSD.

I only did before and after test on the M.2 and should have run a test on the other SATA 850 evo as well.

Maybe you shouldn't run other CPU intensive tasks while doing the benchmark.

You get v& when the NSA access your CP collection.

Is there a way to manually download this for Windows 10? I'm using LTSB, but it's not showing up when I check for updates.

WAIT I THOUGHT INTEL WAS SUPPOSED TO GET FUCKED
WHY DID EVERYONE TELL ME INTEL WAS GOING TO GET 30% PERFORMANCE DROPS
I WAS HAVING SO MUCH FUN WATCHING INTELFAGS REALIZE THEIR HARDWARE WAS BECOMING USELESS BEFORE THEIR EYES
WHAT HAPPENED BROSSSSSSSSS????????

NOO why aren't using it for a server workload jobs??? IT wAS ouR tImE AMDBROS!!

Why? then I'll only blast the AMD kids out the ball park.

You can try downloading it from the windows update catalog, but it never works for it. It says "incompatible" or some junk. But I'm sure you'll have better luck.

I don't know why it even gives me that error message. Just to be sure I download and try each one. Not a single one works.

Crystalmark is not a good benchmark tool in this case, it doesn't test I/Os

i dunno bud, i thought context switching was a key element of this slowdown. wouldnt the different os' handle that using different code?

Post above. Pre below.

>That IOPS loss
Fak.
Windows isn't very good with I/O to begin with but that's a notable chunk.

0x80240037 WU_E_NOT_SUPPORTED The functionality for the operation is not supported.

Is it time I update to windows 10? Did they get me?

>160 burgers more expensive than ryzen 5

Serious question, if people have money for SSDs why aren't they buying optane ones? I mean at least it won't die in a year or two

>77 dollars for 32 gigs

>Oy vey, why aren't you buying more quality Intel™ products? You aren't poor, are you?

Ive been using my 840 pro for like 5 years now and its still just fine.

7% loss in sequential read.
12% loss in random read.
11% loss in random write.

Shiiittt.

In comparison i was also getting 500mb/s read and write. I also don't have RAPID mode enabled.

You using it?

Optane is a meme. The same performance of Sata 3 SSDs without the storage to justify.

The entry code and a lot of implementation details are different but the mechanics are similar. The context switching mechanism is provided by the CPU arch, not the OS. What will shit on syscall-heavy workloads is inherent to the steps needed to protect the kernel memory (TLB flush).

20TB written so far

>if people have money for SSDs
850 Evo were like $109 for 500gb last week.
The 375gb optane was $1500.

The 850 loads windows and games fast enough.

Dude, you have no idea what you're talking about. Stop.

It's a HARDWARE bug. As a result, the impact is largely independent of the operating system. Perhaps the fixes might be somewhat different, but unless you're implying that the Linux developers or Linux kernel are preposterously worse than Windows, there is absolutely no reason to think that similar workloads would operate much differently. In fact, for the particular workloads being discussed, Linux works just as well as windows does-- which is the datacenter setting. Actually, Linux probably works better.

It's a meme that works though. My friend had 2 SSDs die on him, he's a photographer. Not a hipster, he's making a living with pro-photoshoots.

Skipped this post because it REEKS of reddit.

What I don't understand is why Intel cannot be held accountable

>Sorry, here is the patch
>Patch cripples performance, product is no longer what I bought. Money back
>No lol

How are they not being driven into the ground?

Been here for over ten years, newfag. If you're so familiar with the reddit style, you should stick to that site.