Intel CPU performance-loss by security-patch?!?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
MikeB
Posts: 3467
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Intel CPU performance-loss by security-patch?!?

Post by MikeB » Thu Jan 04, 2018 7:10 pm

Milos wrote:
MikeB wrote:But the avg prices he sold out is only $.30 higher than today's price. The price ran up right after he sold it. For these guys $.30 on 250,000 shares is chump change.
Intel is 4% down today only (while DOW is 0.6% up) and that's just the beginning. So I'd say you are greatly downplaying it.
In the same time AMD up over 5% just today ;).
Well I’m not playing anything. My point was that he did not make a killing before the news came out. The news is out now - buyer beware.

syzygy
Posts: 4455
Joined: Tue Feb 28, 2012 10:56 pm

Re: Intel CPU performance-loss by security-patch?!?

Post by syzygy » Thu Jan 04, 2018 9:43 pm

MikeB wrote:
zullil wrote:
Probably just a coincidence. :roll:
it was only $39 million - sold all the shares he was allowed to sell (Intel requires CEO to maintain 250,000 shares

This article was written before the latest debacle was publicly known and even then it was suspicious.
Motleyfool
But the avg prices he sold out is only $.30 higher than today's price. The price ran up right after he sold it. For these guys $.30 on 250,000 shares is chump change.
Now that the news has "officially" come out, it seems there is not that much that was not publicly known when he sold the shares. What is now called the "meltdown" attack has been publicly discussed on the Linux kernal mailing list for a couple of months. The "spectre" attack or attacks might be new, but are apparently even more difficult to get working.

I thought all the sudden secrecy and hurry was a sign that someone had discovered that the problem was more serious than it was initially thought, but that doesn't seem to be the case. (So the problem is serious, but it seems that so far no one has managed to exploit it in the real world.)

MikeB
Posts: 3467
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Intel CPU performance-loss by security-patch?!?

Post by MikeB » Thu Jan 04, 2018 10:31 pm

syzygy wrote:
MikeB wrote:
zullil wrote:
Probably just a coincidence. :roll:
it was only $39 million - sold all the shares he was allowed to sell (Intel requires CEO to maintain 250,000 shares

This article was written before the latest debacle was publicly known and even then it was suspicious.
Motleyfool
But the avg prices he sold out is only $.30 higher than today's price. The price ran up right after he sold it. For these guys $.30 on 250,000 shares is chump change.
Now that the news has "officially" come out, it seems there is not that much that was not publicly known when he sold the shares. What is now called the "meltdown" attack has been publicly discussed on the Linux kernal mailing list for a couple of months. The "spectre" attack or attacks might be new, but are apparently even more difficult to get working.

I thought all the sudden secrecy and hurry was a sign that someone had discovered that the problem was more serious than it was initially thought, but that doesn't seem to be the case. (So the problem is serious, but it seems that so far no one has managed to exploit it in the real world.)
Just for the record , although it may look suspicious , in the United States , large publicly held companies such as Intel are governed by SEC rules. Large companies have blackout dates which prevent Officers from the company from excercisng stock options or selling stocks during the the blackout period. Most companies that report on earnings as of 12/31 are currently in that blackout period. The exception to this rule is that officer may set a plan in advance , picking the date and the number of shares he would either excercise or sell way in advance - ( it would be set in a non blackout period , typically right after earnings release. So although it looks very suspicious , it may be one of those transactions. The lawyers are very good at making sure officers follow the rules in those companies that have good governance. I’m sure the SEC will be looking at those transactions since they have gardnered so much press attention.

syzygy
Posts: 4455
Joined: Tue Feb 28, 2012 10:56 pm

Re: Intel CPU performance-loss by security-patch?!?

Post by syzygy » Mon Jan 08, 2018 11:13 pm

syzygy wrote:Now that the news has "officially" come out, it seems there is not that much that was not publicly known when he sold the shares. What is now called the "meltdown" attack has been publicly discussed on the Linux kernal mailing list for a couple of months.
I have now realised that this is not quite true.

What was publicly known is that information could be obtained about the (randomised) layout of the kernel in memory, even though this was not supposed to be possible. The KAISER patches that have been discussed were an attempt to fix this on Linux. At least officially.

What was not publicly known is that it was actually possible for a user space process to read arbitrary kernel data (on Intel cpus). This is a much more serious problem. The KAISER patches (now called Kernal Page Table Isolation (KPTI) patches) happen to prevent this attack (and the attack was probably the real reason those patches were developed in the first place).

It seems quite a few people were already suspecting that something was being hidden from the public, because there appeared to be an unusual haste to get the KAISER patches integrated into the Linux kernel. And at the same time similar patches were popping up for MacOS and Windows.

https://lwn.net/Articles/741878/
KPTI, in other words, has all the markings of a security patch being readied under pressure from a deadline. Just in case there are any smug ARM-based readers out there, it's worth noting that there is an equivalent patch set for arm64 in the works.

MikeGL
Posts: 900
Joined: Thu Sep 01, 2011 12:49 pm

Re: Intel CPU performance-loss by security-patch?!?

Post by MikeGL » Tue Jan 09, 2018 3:38 am

syzygy wrote:It seems quite a few people were already suspecting that something was being hidden from the public,
By connecting the dots, I think this is related to the recent Kaspersky vs.
NSA contractor who is of Vietnamese origin. (but already a US citizen)

The said NSA guy brought all types of exploits imaginable to his home laptop
(including this secret intel bug), but those powerful exploits including compressed source code were siphoned
and copied by someone in Russia (thanks to Kaspersky Anti-virus malware
auto-upload to cloud). These exploits made everybody vulnerable from powerful secret
agencies (Russia, China, probably others). To undo this, it is
required that US tech giants release a patch for these bugs asap.

Far fetched, but that's what I think on why there's a haste to applying patches to these, very difficult to find, bugs.

User avatar
Nordlandia
Posts: 2453
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Intel CPU performance-loss by security-patch?!?

Post by Nordlandia » Tue Jan 09, 2018 9:15 am

I notice 2.5 - 5% performance decline on sandy-bridge 2500K.

Jouni
Posts: 2005
Joined: Wed Mar 08, 2006 7:15 pm

Re: Intel CPU performance-loss by security-patch?!?

Post by Jouni » Tue Jan 09, 2018 12:17 pm

I see now, that SF8 bench is always below 2 Mnps previuosly it was about 2 100 000 (Corei5) if I recall correctly so 5% slower :( .
Jouni

Jouni
Posts: 2005
Joined: Wed Mar 08, 2006 7:15 pm

Re: Intel CPU performance-loss by security-patch?!?

Post by Jouni » Tue Jan 09, 2018 1:16 pm

Never mind - no slowdown so far.
Jouni

schack
Posts: 136
Joined: Thu May 27, 2010 1:32 am
Contact:

Re: Intel CPU performance-loss by security-patch?!?

Post by schack » Tue Jan 09, 2018 7:33 pm

How is it affecting tablebase reads on SSDs?

APassionForCriminalJustic
Posts: 415
Joined: Sat May 24, 2014 7:16 am

Re: Intel CPU performance-loss by security-patch?!?

Post by APassionForCriminalJustic » Tue Jan 09, 2018 11:57 pm

Jouni wrote:I see now, that SF8 bench is always below 2 Mnps previuosly it was about 2 100 000 (Corei5) if I recall correctly so 5% slower :( .
Is this patch really necessary? I mean - is someone really going to hack a personal PC? Doesn't seem worth any performance loss...

Post Reply