More Intel leaks.. this one is not good though

[Princess Luna]

4 minutes ago, Labeled said:

. 

That is why we have 2 step verification systems on just about every thing nowadays though.

[Arika]

20 minutes ago, cj09beira said:

one specific task that heavenly uses the type of command that would have this exploit

but my benchmark numbers QQ

[Silentprototipe]

Now I dont feel bad for skipping on Intel all the time . Ryzen & Llano life FTW

[TidaLWaveZ]

14 hours ago, Princess Cadence said:

No need for alarm just yet, there always is plenty of exaggeration and drama around this kind of news on release.

Nah, intel is probably filing for bankruptcy as we speak.

 

I mean every time there's an issue they're doomed, so there's no way they'll manage to survive this one!

 

 

[cj09beira]

Just now, TidaLWaveZ said:

Nah, intel is probably filing for bankruptcy as we speak.

 

I mean every time there's an issue they're doomed, so there's no way they'll manage to survive this one!

 

 

although it isnt that bad for intel, its great to undermine intels "reliability" which might mean more epyc sales for amd

[TidaLWaveZ]

Just now, cj09beira said:

although it isnt that bad for intel, its great to undermine intels "reliability" which might mean more epyc sales for amd

Kudos for caring that much.

[LAwLz]

44 minutes ago, TVwazhere said:

And here I was hoping 2018 would be a good year. We aernt even three days in and I wake up to this shit.  Hopefully the real world daily use performance hit is more like 5% rather than 30%

8 minutes ago, bcredeur97 said:

this is insane. wow. 

 

 

30% performance hit is *general* or specific to a type of task?

The average performance hit seems to be around 5%. In a lot of cases it is lower than that (for example in Phoronix gaming benchmarks they were all within margin of error). The ~30% numbers people are reporting are specific tests which rely heavily on syscalls, and does not take advantage of PCID. It is basically the worst case scenario. For your average program expect a 0 to 10% performance decrease.

 

 

17 minutes ago, cj09beira said:

edit2: i am super confused now, what does the - and + mean, is it that those lines were removed? 

and the others added?

Yes, the minus means the part was removed, and the plus means it was added.

That is just a pull request though, proposed by AMD. It is not included in the kernel yet.

@DoctorWho1975

 

23 minutes ago, Stefan Payne said:

The Problem is not imaginable but at least a Fukushima or Tschernobyl on the desaster scale...

You're overreacting...

 

9 minutes ago, Eibe said:

The peak was 30%. It usually hovers between 5 and 30%.

Actually, it usually hovers between 0 and 9 percent, not 5 and 30. At least according to this mail thread about KAISER, which I think would protect against exploits based on this security hole. The authors of the actual white paper states an overhead of 0.28%, but that seems too optimistic. It seems like 5% is the average impact, not by any means the smallest.

[Stefan Payne]

5 minutes ago, LAwLz said:

You're overreacting...

 

I was just citing someone who worked on DOOM and other stuff. And other "Server Guys" who see it in a similar way...

[cj09beira]

5 minutes ago, LAwLz said:

The average performance hit seems to be around 5%. In a lot of cases it is lower than that (for example in Phoronix gaming benchmarks they were all within margin of error). The ~30% numbers people are reporting are specific tests which rely heavily on syscalls, and does not take advantage of PCID. It is basically the worst case scenario. For your average program expect a 0 to 10% performance decrease.

 

 

Yes, the minus means the part was removed, and the plus means it was added.

That is just a pull request though, proposed by AMD. It is not included in the kernel yet.

@DoctorWho1975

 

You're overreacting...

 

Actually, it usually hovers between 0 and 9 percent, not 5 and 30. At least according to this mail thread about KAISER, which I think would protect against exploits based on this security hole. The authors of the actual white paper states an overhead of 0.28%, but that seems too optimistic. It seems like 5% is the average impact, not by any means the smallest.

hopefully they added it soon, (they probably want to test it themselves first)

[Nocte]

10 minutes ago, LAwLz said:

Actually, it usually hovers between 0 and 9 percent, not 5 and 30. At least according to this mail thread about KAISER, which I think would protect against exploits based on this security hole. The authors of the actual white paper states an overhead of 0.28%, but that seems too optimistic. It seems like 5% is the average impact, not by any means the smallest.

I see. My bad for understanding it incorrectly.

[LAwLz]

5 minutes ago, cj09beira said:

hopefully they added it soon, (they probably want to test it themselves first)

Yeah, it seems like the Linux devs are on the causious side for now, by applying it to all processors. Then they will figure out if it affects other processors or not.

 

Also, you read the code correctly. It's just that it was a change request rather than something that had been implemented.

That request will probably be denied though, since, like you said, it applies the fix to all non-AMD processors. It's better to detect and apply the fix to all processors which are affected, rather than apply it to all processors except AMD ones.

 

 

Just now, Eibe said:

I see. My bad for understanding it incorrectly.

Not your fault. A lot of news sites are saying the same thing you said. Even the quote in the OP says 5% to 30%.

[Notional]

8 hours ago, mr moose said:

Is there any evidence Intel knew about this for the last ten years or do we just have the typical armchair experts again pontificating Intel's ineptitude and evilness?

 

There's no way Intel would have known about this and not fixed it in the numerous iterations of new architecture over the years. However, I do believe the CEO knew about this some weeks/months ago. The stock sales are VERY dubious and should be looked into. How can that be legal?

5 hours ago, porina said:

https://www.phoronix.com/scan.php?page=news_item&px=x86-PTI-Initial-Gaming-Tests

Gaming tests (under Linux) don't show any change, so gamers rejoice!

 

Well is the game is not CPU limited, then you wouldn't see any differences. If it is CPU limited, then you probably will see a difference.

[jagdtigger]

2 minutes ago, LAwLz said:

Yeah, it seems like the Linux devs are on the causious side for now, by applying it to all processors. Then they will figure out if it affects other processors or not.

 

Also, you read the code correctly. It's just that it was a change request rather than something that had been implemented.

That request will probably be denied though, since, like you said, it applies the fix to all non-AMD processors. It's better to detect and apply the fix to all processors which are affected, rather than apply it to all processors except AMD ones.

Most of the world runs on it so im not surprised they opted for a solution that is a response to the worst case scenario...

[iiNNeX]

So this happens just as I am about to order my 8700k system.... hmmm. I ll wait for some benchies after the patches and see how this goes. I want ze frames!!!

[Princess Luna]

Just now, Nicholatian said:

I don't think you understand the magnitude of this exploit, so let's put it in a different way. Taking the mystification of technology out of it for a moment to help break down in layman's terms why it's very very bad for everyone from a security standpoint.

I would very much appreciate if you'd quit the condescending tone when addressing me every time because I am reading the situation from the mainstream end user perspective only.

[NumLock21]

10 hours ago, leadeater said:

From what I understand anything that supports virtualization extensions is affected i.e. VT-x and VT-d

I had mine turn off. So I'm not affected!

 

[leadeater]

Just now, NumLock21 said:

I had mine turn off. So I'm protected!

 

lol you wish . That's just when the hardware changes were introduced, it's sort of unrelated but an easy way to tell if an Intel CPU might be affected.

[Natsoup]

15 hours ago, Egg-Roll said:

Over all the smarter people won't have much to worry about unlike those who think leaving their phones and computers unlocked without passwords and use them for personal stuff and keeping passwords on them for said things.

I leave my home PC unlocked. I don't think that it will ever be a problem for me and if anything I have on my computer is comprimised it will almost definitely be from the inside rather than out. 

 

As per the article, I'm not freaking out yet because this information seems broad and panicky. I might invest in AMD though just to make a few bucks.

[LAwLz]

14 minutes ago, Notional said:

There's no way Intel would have known about this and not fixed it in the numerous iterations of new architecture over the years. However, I do believe the CEO knew about this some weeks/months ago. The stock sales are VERY dubious and should be looked into. How can that be legal?

Hanlon's razor...

 

It is actually not uncommon for some issues to go unnoticed for many years.

EternalBlue (SMB exploit used by WannaCry and others) was something like 15 years old when it was made public, and it was most likely ~10 years or older when first discovered.

 

 

5 minutes ago, Nicholatian said:

With that in mind, I'd be happy to continue the discussion to answer any questions you have or anything you might be confused over, if I know. I'm not a security engineer, but I am a software engineer professionally and have learned a lot about computers, even studying to the level of assembly with ARM CPUs.

What question can you even answer? From what I know, the details of the issue are not public yet. Seems like it's mostly speculation at this point.

[Tiberiusisgame]

NAS's with Intel processors... that's really going to hurt with Synology's recent announcement to be able to host virtual systems on their devices.

 

I guess now is a good time to run benchmarks so we can all start making comparisons for ourselves after the patch... and find out if a decent overclock helps.

[porina]

51 minutes ago, Notional said:

Well is the game is not CPU limited, then you wouldn't see any differences. If it is CPU limited, then you probably will see a difference.

Of course we will have to wait for more data, particularly Windows data on more titles than they tested. Still, in that limited testing they did on Linux, most of it was at lower resolutions, and pushing some frames which is one scenario where CPU becomes more influential. There was still no significant difference in those cases.

[cj09beira]

2 minutes ago, Tiberiusisgame said:

NAS's with Intel processors... that's really going to hurt with Synology's recent announcement to be able to host virtual systems on their devices.

 

I guess now is a good time to run benchmarks so we can all start making comparisons for ourselves after the patch... and find out if a decent overclock helps.

good for qnap because they use ryzen cpus 

[goodtofufriday]

1 minute ago, cj09beira said:

good for qnap because they use ryzen cpus 

Only their newest equipment. Plenty of qnap devices doing virtualization on intel cpus. 

[ItsMitch]

Watching the stocks rn, Intel is having a tough day while AMD are having a glorious one

[Mira Yurizaki]

5 hours ago, Smoofie said:

So, any Americans feel like suing Intel if the loss  in performance is great enough :P? Even though Intel most likely didn't hide this fact, it is still at fault for what's likely some performance drops in cloud services and if I read it right, also for things like Blizzard WoW servers.

Except nothing will come out of it from, as far as we know, the results of a genuine mistake. And if you can successfully sue on the grounds of that... oh boy howdy. I mean nobody sued Intel over the FDIV or F00F bugs. And I don't think anyone sued AMD over Phenom's TLB bug. Besides that, people still have the option to not apply the patches and live with the risks. It's not like Intel said "this processor gets 8000 bungomarks" when it physically cannot.