Intel bug incoming? Meltdown and Spectre exploits

KentMan · 3 Jan 2018 at 18:48

Intel is already facing a fight in the server space from not only AMD but other companies. They probably need this like a hole in the head right now.

Desktop users won't give a monkeys though if the performance hit is negligible right now, what will be interesting though is if AMD is not affected and their new CPU close the gap on Intel and AMD keep their more cores for less money stance too, Intel will have a bigger fight on their hand.

Right now in the desktop space Intel has an IPC and clockspeed lead, if they both reduce and AMD catches up within reasonable distance, Intel will have problems.

new boy · 3 Jan 2018 at 18:52

Nasher said:
Hmm, it says the fix will significantly slow down CPU performance. I guess we need to avoid that patch on gaming PCs. It will be something really specific that only effects certain tasks that 95% of people won't be doing. Yet the update will be pushed on everyone.

But let me guess. The next generation will have the fix and be just as fast?

The security risk is a danger to everyone, its the performance hit that the patch causes that might not effect the majority of tasks too much.

Perfect_Chaos · 3 Jan 2018 at 18:53

KentMan said:
Right now in the desktop space Intel has an IPC and clockspeed lead, if they both reduce and AMD catches up within reasonable distance, Intel will have problems.

Biggest difference will come with Zen 2 i guess but at least Zen+ will be interesting to see what they get out of it.

ultralaser · 3 Jan 2018 at 18:54

mattyfez said:
Insightfull. Source?

So it apparently effects every call into the kernel (so every syscall), and the reports are that every syscall will take a 50% performance hit. Now some applications will get an absolute hammering (virtualization like you say and anything that does a LOT of IO). But that no way what so ever should be equated to 'it won't really effect' home users.

JasonM · 3 Jan 2018 at 18:58

This is looking like a bigger mess up since Professor Nicely found the floating point issue in the Pentium 1.

faceman123 · 3 Jan 2018 at 18:59

I really really hope there's a way to opt out of this update, essentially this 5% is what each tick has given us a 30%loss in anything is disastrous

Nasher · 3 Jan 2018 at 18:59

KentMan said:
Intel is already facing a fight in the server space from not only AMD but other companies. They probably need this like a hole in the head right now.

Desktop users won't give a monkeys though if the performance hit is negligible right now, what will be interesting though is if AMD is not affected and their new CPU close the gap on Intel and AMD keep their more cores for less money stance too, Intel will have a bigger fight on their hand.

Right now in the desktop space Intel has an IPC and clockspeed lead, if they both reduce and AMD catches up within reasonable distance, Intel will have problems.

But only if AMD keeps undercutting them on price. They are only doing that now because they don't have the lead.

gavinh87 · 3 Jan 2018 at 19:00

From 5-50% now, so much hyperbole from so little information.

mattyfez · 3 Jan 2018 at 19:00

ultralaser said:
So it apparently effects every call into the kernel (so every syscall), and the reports are that every syscall will take a 50% performance hit. Now some applications will get an absolute hammering (virtualization like you say and anything that does a LOT of IO). But that no way what so ever should be equated to 'it won't really effect' home users.

This site suggests games and desktop apps won't really be affected.

translate.googleusercontent.com/translate_c?depth=1&nv=1&rurl=translate.google.com&sl=auto&sp=nmt4&tl=en&u=https://www.hardwareluxx.de/index.php/news/hardware/prozessoren/45319-intel-kaempft-mit-schwerer-sicherheitsluecke-im-prozessor-design.html&usg=ALkJrhg--IiIqA7LOGozMMSS3ccCy0y9cQ

ultralaser · 3 Jan 2018 at 19:01

mattyfez said:
This site suggests games won't really be affected.

In order for that to be true, no games ever make syscalls then?

David Bisset · 3 Jan 2018 at 19:04

faceman123 said:
I really really hope there's a way to opt out of this update, essentially this 5% is what each tick has given us a 30%loss in anything is disastrous

You want to voluntarily leave a vulnerability that compromises everything on your computer, that so far looks to not really cause a significant performance loss for gaming? Why?

Nasher · 3 Jan 2018 at 19:06

But we've had the vulnerability for 10 years, so why bother now

Plus it sounds more like something that would be used in a targeted attack on host machines, not random desktop users.

mattyfez · 3 Jan 2018 at 19:07

ultralaser said:
In order for that to be true, no games ever make syscalls then?

Well the frames per second speaks for itself, there's a small hit, but it is small.

jigger · 3 Jan 2018 at 19:07

gavinh87 said:
From 5-50% now, so much hyperbole from so little information.

I thought you couldn't give a rat's.

mattyfez · 3 Jan 2018 at 19:08

Nasher said:
But we've had the vulnerability for 10 years, so why bother now

Plus it sounds more like something that would be used in a targeted attack on host machines, not random desktop users.

Because no one knew about it, now the cat is out of the bag, it needs to be patched before exploits are developed.

jigger · 3 Jan 2018 at 19:10

Nasher said:
But we've had the vulnerability for 10 years, so why bother now

Plus it sounds more like something that would be used in a targeted attack on host machines, not random desktop users.

If it ain't broken... no wait if it is broken... I'm confused

Nasher · 3 Jan 2018 at 19:11

mattyfez said:
Because no one knew about it, now the cat is out of the bag, it needs to be patched before exploits are developed.

OFC people knew about it, it's been around for at least 10 years. They just didn't tell anyone.

nixd2012 · 3 Jan 2018 at 19:12

Nasher said:
No word on if AMD are affected or not yet...

This basically says AMD are not affected. (See my sig for a level of smugness).

Spyhop · 3 Jan 2018 at 19:14

WOW

Intel to give everyone a free Optane PCIe SSD as a goodwill gesture

https://www.anandtech.com/show/14201/intel-response-to-cpu-bug-more-generous-than-anyone-expected/

ultralaser · 3 Jan 2018 at 19:15

mattyfez said:
Well the frames per second speaks for itself, there's a small hit, but it is small.

So with this bug it's going to be very different depending what application (be that a game or not) is doing (is it reading from disk, is it using network IO etc). I would not take any benchmark seriously at this point apart from the synthetic ones demonstrating the raw problem...