Another Intel CPU vulnerability

kitfit1 · 16 May 2019 at 19:23

RavenXXX2 said:
You will see a 3% loss in performance.

If you have a cpu with HT and you disable it (as recommended by Intel), you will lose a lot lot more than 3% performance.

ianh · 16 May 2019 at 19:37

There is/will be a microcode patch coming out. You only need to restrict hyperthreading until that patch is applied and then you get the 3% drop, on top of all the other "Intel are unsecure" microcode patches which reduce performance 3-5%, what a bloody joke.

nashathedog · 21 May 2019 at 11:10

Well Intel has once again found itself at the center of another CPU Flaw/Exploit... this time the only way to completely mitigate the threat on a local level is to turn off Hyper-Threading on its CPUs...

It sounds to me like Intel need to move away from the Core I architecture altogether and build a new one from the ground up or is it Intels way of trying to persuade users with older cpu's to move up to the newer sku's? You never can tell with companies like this.

DragonQ · 21 May 2019 at 12:39

kitfit1 said:
If you have a cpu with HT and you disable it (as recommended by Intel), you will lose a lot lot more than 3% performance.

The people quoting these tiny HT figures are almost always talking about gaming, where it tends to be less helpful. It is not, however, a waste of time. Plenty of games choke at 4c/4t compared to 4c/8t.

kitfit1 · 21 May 2019 at 15:46

DragonQ said:
The people quoting these tiny HT figures are almost always talking about gaming, where it tends to be less helpful. It is not, however, a waste of time. Plenty of games choke at 4c/4t compared to 4c/8t.

Of course they choke, even going back to early days. It dosn't show as much now with modern GPU's though, which i think is why most peeps just ignore it or don't realize it.

StarShock · 21 May 2019 at 15:54

DragonQ said:
The people quoting these tiny HT figures are almost always talking about gaming, where it tends to be less helpful. It is not, however, a waste of time. Plenty of games choke at 4c/4t compared to 4c/8t.

Also worth noting that intel charge about £130 quid for the privilege of having hyper threading as least on the 9900k compared to 9700k, so even in cases where it doesn't make a huge amount of difference, its a kick to people who paid out for it.

That being said I think the real issue isn't to the average consumer, its more the servers running hyper threaded systems that are going to be hurting. Chances of a home user actually getting targeted by someone using this kind of exploit is basically zero

nashathedog · 21 May 2019 at 16:07

StarShock said:
Also worth noting that intel charge about £130 quid for the privilege of having hyper threading as least on the 9900k compared to 9700k, so even in cases where it doesn't make a huge amount of difference, its a kick to people who paid out for it.

That being said I think the real issue isn't to the average consumer, its more the servers running hyper threaded systems that are going to be hurting. Chances of a home user actually getting targeted by someone using this kind of exploit is basically zero

This particular issue is only a factor on older cpu's upto 7th gen which means all the 4 core 8 thread i7's up to Kabylake are vulnerable but as you said it's unlikely individuals will be targetted.

I imagine companies who are vulnerable will be quite displeased to learn about this (especially now everyone knows it can be done) & it may give AMD a leg up when it comes time for those firms to upgrade their systems as once again it's not a problem for AMD's Zen ranges.

dragonslayer · 21 May 2019 at 18:09

I can see AMD getting massive sales in the server department this year/ next year and over taking intel quite quickly with all these flaws and vulnerabilities that keep popping up

MaXxBoulton · 21 May 2019 at 20:31

RavenXXX2 said:
You will see a 3% loss in performance.

this is on top of all the other loss's in performance due to the other vulnerability's

an0nym0us · 22 May 2019 at 07:06

This just makes me more excited about the Ryzen 3000 series launch.

Steedie · 22 May 2019 at 09:55

Another "just 3% drop" on top of all the other "just 3% drop" patches. Adding up a bit isn't it...

davida76 · 22 May 2019 at 10:29

Steedie said:
Another "just 3% drop" on top of all the other "just 3% drop" patches. Adding up a bit isn't it...

Aye especially for anyone running an older chip, hard to believe they're suddenly finding all these flaws after nearly 11 years cpu products, the cynic in me says Intel knew about this much earlier and did nothing.

kitfit1 · 22 May 2019 at 10:47

davida76 said:
the cynic in me says Intel knew about this much earlier and did nothing.

I think the cynic in me is a much bigger cynic than you are. The cynic in me says not only did they know about it but deliberately put the vulnerabilities on the cpu's to gain extra performance. Doing it because they didn't think they would be found out.

Rainmaker · 22 May 2019 at 10:56

kitfit1 said:
I think the cynic in me is a much bigger cynic than you are. The cynic in me says not only did they know about it but deliberately put the vulnerabilities on the cpu's to gain extra performance. Doing it because they didn't think they would be found out.

I raise your cynicism with a hypothesis that not only were the branch prediction/MDS 'flaws' intentional, but work alongside the compromised AES-NI and IME to allow easy bypass of encryption (including VPNs etc) for Uncle Sam. If Snowden taught us anything, it's that this is no longer just tinfoil hat stuff. Ask Cisco, or any of the PRISM/Tempora partners.

IT Troll · 22 May 2019 at 14:04

nashathedog said:

0-4% loss in games. 16% loss in Cinebench. <1% loss in Premiere if QuickSync is on. Exactly, why I went for 9700K.

There must be something dodgy in the macOS patch to cause a 40% loss. Oh well, I am sure they will happily sell you a new faster one...

marms · 22 May 2019 at 14:55

IT Troll said:
0-4% loss in games. 16% loss in Cinebench. <1% loss in Premiere if QuickSync is on. Exactly, why I went for 9700K.

There must be something dodgy in the macOS patch to cause a 40% loss. Oh well, I am sure they will happily sell you a new faster one...

But he isn't comparing performance with and without patches installed, just turning off hyperthreading

MaXxBoulton · 22 May 2019 at 17:45

davida76 said:
the cynic in me says Intel knew about this much earlier and did nothing.

An AMD engineer found intel skipping security checks in memory years ago but if anyone ever said anything about it you was called an AMD fan buy and told to bugger off, this is why sandy bridge made a massive step in performance.
Looks like they found a way to exploit the skipped security checks

IT Troll said:
There must be something dodgy in the macOS patch to cause a 40% loss

they turned of hyper threading

Skully178 · 22 May 2019 at 23:52

My poor old 3570k is suffering from all these patches

Stephanie Peterson · 22 May 2019 at 23:58

Might only be a few % in this one ( apart from the turning off of hyperthreading) but add em all up and very soon your 500 quid 9900 is going to be fighting with a 486 for speed.
Ok lets make it far, the 486 Dx4 100...

Grim5 · 23 May 2019 at 03:45

kitfit1 said:
I think the cynic in me is a much bigger cynic than you are. The cynic in me says not only did they know about it but deliberately put the vulnerabilities on the cpu's to gain extra performance. Doing it because they didn't think they would be found out.

If that's true Intel would have a multi billion dollar lawsuit on their hands