Intel processor will loose 17-23%

[MetalFan777]

'Kernel memory leaking' Intel processor design flaw forces Linux, Windows redesign
Other OSes will need an update, performance hits loom

Will this be a crack for Intel?

What is your comment on this.

https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

 

[mutantaxe]

well from another youtube reviewer who did some tests before and after the new code and he found very very slight boost for gaMING BENCHMARKS BUT ONE write benchmark was down more then a little

but thats for gaming

 

[looz]

They're quoting the non-PCID results, that means CPUs older than Haswell.

Actual article:
https://www.postgresql.org/message-id/20180102222354.qikjmf7dvnjgbkxe@alap3.anarazel.de

Note, these results are from PostgreSQL benchmarks. Relation databases are probably the most affected real world use case.

It's certainly major for database servers, but leaving those details out is a bit clickbaity.

 

[Korreborg]

Yes Hardware unboxed have tested it, and it means absolutely nothing to normal people.

 

[AlienALX]

Yes Hardware unboxed have tested it, and it means absolutely nothing to normal people.

Not true though, is it? define "normal people" because I'm almost normal and the massive slap for me is going to be VMware performance. If you mean "Desktop browsing gamer"? then yes, they will be fine. However, those of us who bought threads over speed and IPC for a specific reason will suffer. And lots of people use VMs for all sorts of purposes (mostly trapping scammers these days).

Thankfully my gaming performance won't be affected so that is one thing I guess.

Ed. I want to see everything tested. Not just specific apps to make it look like Intel have escaped. I want to see streaming, video encoding, video editing, VMware and etc. Because I absolutely guarantee it *will* affect some day to day apps that most of PC owners use.

 

[MetalFan777]

Yeah. Can Tom make a real test with before and after results of this. Not only for gaming bur VM, Music production latency and so on.

 

[Korreborg]

VMs and music production is not for "normal" people (Lets call the regular people then)
I have a Unraid server with som dockers, and a couple of VM's, i'm not regular either.

But people go totally mental before anyone knows whats going on, and what the outcome will be.
Myself, and most other "IT enthusiast" including the youtubers, and "tech journalists" are a lot more than regular people, and know a lot of stuff. But we are still just enthusiasts, and really don't know the bottom of stuff like this. But still overflow the internet with articles, blogs and videos about it. The first bench on Linux is, THE FIRST = There might be a degradation of performance, it might be permanent, it might get better.
Nobody knows, but people go ape as soon as someone writes an article.

They need to close the hole NOW! And then they can work into optimizing it.
Recap = relax a little and wait and see.

 

[AlienALX]

You're crazy, right? do you realise how many people these days are into Youtube and uploading video content?

You are seriously trying to simplify this enormous problem. It will affect almost every one, to some degree or another. I do an awful lot of audio work and I can bet it will now take me a lot longer to do that.

Am going AMD next. Epyc possibly or TR.

 

[Korreborg]

You're crazy, right? do you realise how many people these days are into Youtube and uploading video content?

And the point? A lot of people doing stuff does not make the competent at underlaying computer programming.
And if we are talking about it impacting regular people, then there really aren't a lot of people doing it. Well i know the first 100+ gamers. I know of a handful playing with streaming (i have tried it). I know of no one that have continued doing it. And non is doing youtube content stuff. On top, how many has a computer that only uses some kind of browser, and maybe office once in a while.

You are seriously trying to simplify this enormous problem. It will affect almost every one, to some degree or another. I do an awful lot of audio work and I can bet it will now take me a lot longer to do that.

I am not trying to simplify anything. If advocating calming down until the scope is known.
The "media/world/people" are in constant search of breaking news, and the sec someone smell something = BREAKING NEWS - ENORMOUS FIRE BROKE OUT. Instead of looking into the problem, and the deciding what to do.
Class action lawsuit - On the base of exactly what?

 

[looz]

You'll be hard pressed to find any desktop application which hammers syscalls to the point that the performance deficit becomes in any way meaningful. Apparently even compiling (except for Java, which should be patchable) is absolutely fine with patched systems. Video encoding is also unaffected.

The worst case numbers are from mostly arbitrary tasks - no-one has a system exclusively to run du constantly, so while 30% slowdown is major, it's for a command you only use every now and then.

Even in case of servers hosting VMs the performance drop isn't huge on recent hardware, as in Haswell or newer. For large corporations which utilise their servers fully a 5% slowdown will obviously mean that they need more hardware to provide the same services, but most companies have some leeway in the server utilisation.

 

[AlienALX]

I am not trying to simplify anything. If advocating calming down until the scope is known.
The "media/world/people" are in constant search of breaking news, and the sec someone smell something = BREAKING NEWS - ENORMOUS FIRE BROKE OUT. Instead of looking into the problem, and the deciding what to do.
Class action lawsuit - On the base of exactly what?

I couldn't understand the first part of your post to me. So I've skipped it for now.

As for assuming the worst? and the scope being known? the scope was already known. Some D-bag has a back door straight into your kernel and can steal whatever he/she wishes from you. Is that not bad enough? well for me it wasn't, no. I am then told I am going to get a 5-30% performance hit. TBH? even 0.1% would be too much because I did not pay what I paid for a CPU that performs at less than I bought it for.

I think you also misunderstood me. My annoyance at the whole thing is bad enough, without Intel immediately hurling s**t around trying to make it stick to others. First thing they did was drag AMD through the mud with them, even though that was only half of it. And they made at least some of it stick too, right where it counted (with their loyal and devoted fanboys).

It's called an open mind. IE - being open to anything. And yes, that does mean panic when there are two gateways straight into your bloody PC and all of the information on it.

 

[Kaapstad]

With lack of competition companies like intel get fat lazy and incompetent and things like this happen.

All the years intel were pushing out quad core i7s for the highend desktop and offering nothing much more than 5% or 10% performance increase each gen and pocketing the huge profits instead of investing them in proper R & D.

I hope that there is a huge class action law suit against intel, apart from end users getting some compensation it will financially weaken intel and give other companies like AMD the chance to compete on a more even basis.

 

[AlienALX]

With lack of competition companies like intel get fat lazy and incompetent and things like this happen.

All the years intel were pushing out quad core i7s for the highend desktop and offering nothing much more than 5% or 10% performance increase each gen and pocketing the huge profits instead of investing them in proper R & D.

I hope that there is a huge class action law suit against intel, apart from end users getting some compensation it will financially weaken intel and give other companies like AMD the chance to compete on a more even basis.

Well to be blatant all they have done since Sandy, a decade ago, is shrink things.

And they want £180 a time (based on the original I5 2500k price). So yeah, you're spot on. Lazy and terribly complacent IMO.

 

[juandon]

tbh if I was running intel I would be more worried about the fact that any software upgrade for this is going to give any decent hacker the specific point that they need to work on its a plaster that is pathetic

 

[AlienALX]

tbh if I was running intel I would be more worried about the fact that any software upgrade for this is going to give any decent hacker the specific point that they need to work on its a plaster that is pathetic

It's not. They have completely re-routed the data so it does not go near the affected entry point. It's those "traces" if you will that open the door. With them routed out it's effectively shut.

 

[juandon]

but surely that's just a software shut if that makes sence its not a physical shut so you have just told the hacker community that the red door is broken but we are patching that with software that uses the blue door instead, in reality the red door is still broken and can be opened if you make software etc etc

 

[AlienALX]

but surely that's just a software shut if that makes sence its not a physical shut so you have just told the hacker community that the red door is broken but we are patching that with software that uses the blue door instead, in reality the red door is still broken and can be opened if you make software etc etc

It is shut. To gain access again you would need to put an old/infected kernel onto the target PC. That is why it can affect performance, because the instructions are being re-routed. If the target PC already has the update or kernel on it you wouldn't be able to exploit it.

 

[juandon]

seriously hope your right matey but hackers have a strange way of working round things.

 

[looz]

They can no longer use this vulnerability, but I can assure you that a device as complex a computer has plenty of other vulnerabilities they can poke holes through.

 

[AlienALX]

seriously hope your right matey but hackers have a strange way of working round things.

Yeah I know man. Thing is, it was literally an open door. I don't know if you saw the video posted on Twitter I posted here but yeah, was too easy for those in the know.

However, I would imagine a processor is considerably more complex than your average hacker's knowledge. Let's face it, it's been broken like this for a decade and it's only just been exploited now.

Hackers are not like Jim Keller. So I do feel safe with my limited understanding and how they have patched it, even if it means losing a tiny bit of performance.