About this intel bug

According to Ryan Shrout, who is probably the most reliable source for this kind of thing, consumer workloads don't appear to be affected. If you're running a DB or a ton of IO you might have a little trouble.

twitter.com/ryanshrout/status/948411662046060544

Other urls found in this thread:

theregister.co.uk/2018/01/02/intel_cpu_design_flaw/
twitter.com/SFWRedditVideos

Who cares if they "don't appear" to be affected or you'll have "little trouble", Intel lied and scammed their customers. Say whatever you want, your processor is being gimped. If you stand behind Intel you're an actual retard.

The draw calls will absolutely love it.

Maybe you don't care to know if that affects you or not but I do

Shut the fuck up and play along dumbass. You really want intel to hog the CPU market forever? Just play along with the fake news and spread it far and wide. Watch intel crash and burn. We doing it for the lols

Imagine being a VPS host right now.

>consumer workloads don't appear to be affected
This only pertains to the fix, not the vulnerability itself.

>AMD consumer brain power at work

Is ESXi affected? KVM?
Does it expose keys in memory to userspace applications? How does the patch work and can it be bypassed?
I don't give a shit if user358729352's laptop is 30% slower

these digits tell you all you need to fucking know

I hope EPYC is waiting to catch this rebound HARD because intel fucked up and they're officially in the dog house for all future server purchases I oversee until I understand these patches are not slapped together half-assedly

tfw Ryzen

As far i understood about the fix, system calls in general will get slower.
And calls to the vulkan/DX/Open GL API are well system calls.

Will be fun to see the updated gaming benchmarks.

I'm also dying to see the rematch benchmarks post doping scandal

kekd

Seems only people about to get fucked are server providers

Actually I'm more worried about the Intel atom SoC devices like security cameras and NASes and AP controllers where patching will be more annoying and they can sit on networks for years unpatched

however to be fair exploiting this will take some remote execution

Implying Sup Forums has any kind of social connections. This is just going to stay relativity within Sup Forums and have no major impact on the market.

>And calls to the vulkan/DX/Open GL API are well system calls.
Nope

fixed

Omg I may get 110 fps instead of 120... how will I survive!

No-one buys intel for servers anyway, Epyc was already a no brainer.

Are they loaded on the same memory space as the program, and somehow have access to the driver etc?
In some point of this chain the kernel must be called as far i know, you can't make direct hardware access from a user mode context.

Gaming and video decoding will not even be affected.

if this affects VMs and VPS this is going outside of Sup Forums

also if all the low end laptops suddenly can't play a youtube video at 1080 or if gamers suddenly drop 10% FPS people will howel. and loud minorities seem to drive this stuff.

>Ryan Shrout
I don't give a shit what that shill thinks. I'll wait until Level 1 Techs chime in on this issue.

Media reporting that everyone will be slowed down, will be interesting seeing how low the stock falls.

>Level 1 Techs
>the mall santa amd shill
lol

I didn't really look into the issue and don't know what exactly will cause performance drop, but API call is not a system call. APIs implementations can use system calls, that's true. But due to existence of this intermediate API layer no context switch is required.
Still I don't know what exactly will hit the performance. But since games on WINE aren't affected I suppose it will be same for windows.

What media?

You may be right, I went back and re-read and I'm not understanding where in the pipeline this is actually located.

If it's really a double context switch for every kernel space access it will hurt like the EPT/NPT issues with VMs. (10%)

I think I'll take advice and get off this hype train until I get a deeper writeup.

everyone seems to be citing this at the moment:
theregister.co.uk/2018/01/02/intel_cpu_design_flaw/


However if you read the kernel patch linked from the story there's some more info

wouldn't things like Netflix get fucked since they make an enormous amount of syscalls?

I was wondering about things like HDCP and audio processing

Underrated

If my anime player and fightan games performance won't tank then all's well with the world

Most online servers you are playing on will take a hit.

>Who cares
Funny mentality from someone who's as affected as everyone else

>fell for the 8700k meme
hold me lads

...

>Program has a glitch like all programs do

WTF U LIER SCAMMER PIECE OF SHIT AHAHGFMSMGKTKSNJA

Are you retarded?

>hurr durr
ladies and gentlemen: the average IQ of an intel consumer

>Sell a product
>Advertise its performance
>Due to a bug in the product, it performs worse than advertised

Id call this false advertisement.

I can't say much but in internal testing it does expose keys. Entirely speculative but Intel may or may not be facing lawsuits after the embargo is lifted, it's believed to have cost multiple billions in damages. Go figure.

Nigger this is Intel's pipeline pre-emptively executing instructions without making the security checks beforehand. You could say that their performance gains were a lie all along.

I wouldn't.
Not unless they were sitting on that bug, knowing it would have to be addressed at the time they brought the product to market.

Reminds me of the Volkswagen fuel efficiency scandal.

Wasn't that more like device makers overclocking their devices when they knew benchmarks were running?

>10 years
hmmmmmmm

How does this affect prebuilts sitting on Walmart shelves and the like? Até those going to be promptly recalled or what?

Incucks in full damage control. This is going to be brutal.

Right about when Intel took lead BTFO AMD.

it affects all intel cpus from the last decade.

Volkswagen met the emissions standards during testing but in the wild, their emissions were like something 40 times higher. What I'm getting at is that I wouldn't be surprised if Intel allowed this problem to exist for some time to get that performance increase. Volkswagen had been lying for years about their performance.

Nigga that's not what I asked.

Their fate.

why the fuck would they recall anything retard. it affects shit a DECADE OLD

New CPU line macaco-lake.

DirectX and OpenGL yes. They try to batch syscalls as much as possible (OpenGL succeeds to a higher degree), but all the benchmarks thus far have been on Linux because the Windows patch isn't out yet. There's more than Direct3D calls too; some games have diffuse constant I/O to disk and to network. The extra latency in I/O could be troublesome.

Anything within the last decade.

That's just the range of CPUs affected, isn't it?

Yeah, their cars detected the test conditions and ran differently for the tests.

I ran some phoronix tests (MP3 encoding, FFmpeg, AIO-Stress, SCi-mark, linux kernel compilation) on my machine (intel q8300 running ubuntu 17.10). now i wait for the patched kernel and i'll re-run the tests to see if it impacts me.

any idea when the linux kernel update will be available in the repos?

>ton of IO
That's all people do with computers unless they're playing solitaire. What decade is your expert living in?!

i suggest ya all running linux to install the phoronix test suite and do the same:

"sudo apt-get install phoronix-test-suite"

then run "phoronix-test-suite gui"

No, the performance was legit, the security of the processor/kernel was a lie.

Please don't tell me my i3 will get even slower, I can't even play some 5yo games already.

The performance isn't legit. In order to get that performance they skipped crucial security.

Just don't update your kernel.

So can I just sit on my ass and let my system install the update for me or what

>any online game will take a massive hit
Oh shit I didn't even thought about that.

And get rekt by 4chin hacker

Someone give me the rundown of this bug in semi-laymans terms.

Apparently there were no checks of the privilege level when speculative execution was accessing the hardware page-table and so kernel memory was open any other VM instance running on the same machine? The work-around is a kernel patch that disables the hardware page table?

>Intel cuts corners to get performance advantage
>Intel dominates market for 7 years
>AMD comes back, IPC the same as intel without cut corners
>Massive intel bugs discovered, ME completely broken into and 5-60% performance hit to keep your memory secure
>Both linux and windows devs are bribed to cripple AMD needlessly as well

>Be American
>Be 56%

>Be American CPU
>Be 65%

>tfw still using Intel qx9650
Get fucked you upgrade fags. I'm lmao'ing at your lives as I type this.

You're also affected.

retard, it impacts all intel core processors. you're affected but since your processor is crippled in the first place it will almost become unusable with a 30% hit on performances.

>Both linux and windows devs are bribed to cripple AMD needlessly as well
I understand Microsoft taking the bribe.
But commie Linux devs cucking out for a big corp. Looks like money still talks eh? What have Linux devs got to say for themselves after being so up their own ass all these years.

tfw q6600

Also affected.

>it impacts all intel core processors. you're affected but since your processor is crippled in the first place it will almost become unusable with a 30% hit on performances.

rinse, repeat...

Also affected. Of you want to patch it might want to consider a phenom, lmao

Article I read says Intel processors from the past decade. How do you know for certain which models are affected?

Are you such a retard that you actually believe that mouth breather when he spouts shit like that?

So how much did Intel pay you then to sellout your beliefs?

For gaming, it may be affected for database-heavy games like Civilization

Can I green-flag certain applications that I know to be safe like Civ V. So they won't be affected by this kernel fix?

>database heavy
You mean games with lots of different properties attached to a "unit"? I don't know the current civ but the last one had less than half of the amount of that when compared to rts games like Starcraft or Dawn of War, and those are in real time.

Everything from Pentium 3 and up could be affected

No. Because if your user account has the right to launch processes with the protection off, then so does malware. Malware creates (or exploits/injects itself into) an unprotected process, and then this flaw gives it root. Game over.

The kernel address mapping has to be off for every process all the time now. Actually the security people wanted that for quite a while, doing this eliminates a lot of chances for kernel information disclosure at a stroke. Linus wouldn't hear of it though, because of the performance regression. You could tell this was a big deal because we're in rc4 and he pulled a major virtual-memory patch in a big hurry that has known performance regressions. He never fucking does things like that.

Blue bucks for lin cucks

Looks like Nvidia is going to be affected as well

What about emulators? Thinks like Cemu are constantly writing the shader cache and shit.

The late merge is for embargo reasons

How about Denuvo games?

Cemu is dead already.

>security bug causes performance problems in malware
kek

Nothing wrong with that, unlike that racist dumpster-fire AMD puts out which is only capable of producing 1/10th of the performance (not to mention, normally lacks a working iGPU)

It affects ALL Intel CPU's including those sold a decade ago AND those sold TODAY.

WHERE CAN I FIND MORE 2B EMOJIS

North Korea