Jump to content

This dont look good, Intel design flaw reduce performance

SvAF/F16_Goblin
 Share Followers 1

Recommended Posts

ShamrockOneFive

ShamrockOneFive 3736

Posted
Posted

There's still a lot of hype flying around about this.

 

PC World had a good statement about gaming performance:

 

Will my games get slower?

Maybe not. Phoronix also tested Dota 2, Counter-Strike: Global Offensive, Deus Ex: Mankind Divided, Dawn of War III, F1 2017, and The Talos Principle on a Linux 4.15-rc6 machine with a Core i7-8700K and Radeon Vega 64. None saw a frame rate change outside the margin of error range.

None of those run on Microsoft’s DirectX technology though, which integrates deeply with the Windows operating system. It remains to be seen how DX games perform in the wake of the forthcoming patches.

 

https://www.pcworld.com/article/3245606/security/intel-x86-cpu-kernel-bug-faq-how-it-affects-pc-mac.html

 

There's not much we can do about it right now but if you read the whole article above it does seem like much older gen CPUs are most affected and even then it may not be a dramatic change for anything but specific use cases.

 

I expect this will be more of an issue in the server room in clustered environments and less of an issue for home PC owners.

 

Wait and see is all we can do.

  • Upvote 1
Link to post
Share on other sites
SCG_Fenris_Wolf

SCG_Fenris_Wolf 1415

Posted
Posted

From official statement by MS and the Windows patch today you can pull following information:

 

 

Older than skylake processors are affected, e.g. i7 6700k and older

All of AMD's processors are affected

 

 

People with younger Kabylake processors such as i7 7700k and the later i7 8700k are not affected

 

 

 

So, I guess everyone who thought AMD is not affected should not have been so gleefull. They are all affected, even their latest Ryzen models. You're only safe with Kabylake and newer right now. It's still a very bad issue, and noone should have to suffer from a thing such as this.

 

 

@OP You might want to correct your Thread title, since it carries false information since today's final official statement. 


However, as stated before, individual reports and benchmarks changing after the patch should be drafted for the actual seriousness of the results for each chip, be it older Intel or the AMD/ARM processors.


Sources:

https://newsroom.intel.com/news/intel-responds-to-security-research-findings/

 

https://www.intel.com/content/www/us/en/architecture-and-technology/facts-about-side-channel-analysis-and-intel-products.html

 

https://twitter.com/CNET/status/948679344855158786

 

https://googleprojectzero.blogspot.de/2018/01/reading-privileged-memory-with-side.html

 

https://www.amd.com/en/corporate/speculative-execution

Link to post
Share on other sites
BeastyBaiter

BeastyBaiter 412

Posted
Posted

From official statement by MS and the Windows patch today you can pull following information:

 

 

Older than skylake processors are affected, e.g. i7 6700k and older

All of AMD's processors are affected

 

 

People with younger Kabylake processors such as i7 7700k and the later i7 8700k are not affected

 

 

 

So, I guess everyone who thought AMD is not affected should not have been so gleefull. They are all affected, even their latest Ryzen models. You're only safe with Kabylake and newer right now. It's still a very bad issue, and noone should have to suffer from a thing such as this.

 

 

@OP You might want to correct your Thread title, since it carries false information since today's final official statement. 

However, as stated before, individual reports and benchmarks changing after the patch should be drafted for the actual seriousness of the results for each chip, be it older Intel or the AMD/ARM processors.

Sources:

https://newsroom.intel.com/news/intel-responds-to-security-research-findings/

 

https://www.intel.com/content/www/us/en/architecture-and-technology/facts-about-side-channel-analysis-and-intel-products.html

 

https://twitter.com/CNET/status/948679344855158786

 

https://googleprojectzero.blogspot.de/2018/01/reading-privileged-memory-with-side.html

 

https://www.amd.com/en/corporate/speculative-execution

 

There are 3 exploits, Ryzen is only subject to 1 of them, Intel is subject to all 3. Source: those links you gave.

Link to post
Share on other sites
dburne

dburne 2583

Posted
Posted (edited)

From official statement by MS and the Windows patch today you can pull following information:

 

 

Older than skylake processors are affected, e.g. i7 6700k and older

All of AMD's processors are affected

 

 

People with younger Kabylake processors such as i7 7700k and the later i7 8700k are not affected

 

 

 

I am not sure that is correct.

 

From the Intel page you linked to:

 

The following Intel-based platforms are impacted by this issue. Intel may modify this list at a later time.

Please check with your system vendor or equipment manufacturer for more information regarding your system.

Intel® Core™ i3 processor (45nm and 32nm)

Intel® Core™ i5 processor (45nm and 32nm)

Intel® Core™ i7 processor (45nm and 32nm)

Intel® Core™ M processor family (45nm and 32nm)

2nd generation Intel® Core™ processors

3rd generation Intel® Core™ processors

4th generation Intel® Core™ processors

5th generation Intel® Core™ processors

6th generation Intel® Core™ processors

7th generation Intel® Core™ processors

8th generation Intel® Core™ processors

 

https://www.intel.com/content/www/us/en/architecture-and-technology/facts-about-side-channel-analysis-and-intel-products.html

Edited by dburne
Link to post
Share on other sites
coconut

coconut 1490

Posted
Posted

Regarding games not being affected: IL-2 and some other flight simulators have a specific hardware usage pattern that is unlike other games, namely they are CPU-bound. That together with frequent switches to the OS context could be a problem. I would expect such context switches to happen more often in multiplayer, when reading/sending data. I'm not convinced the preliminary measurements done on games can be extrapolated to IL-2.

 

Fortunately, network I/O and rendering are likely performed in separate threads, and may also be infrequent enough (from a CPU's point of view, a couple milliseconds is an eternity) that the performance loss due to this isn't noticeable.

 

I've read the paper on meltdown (https://meltdownattack.com/meltdown.pdf). I won't pretend I understood everything, but I think I get the overall picture. The enabling factors for the exploit are 1) the difference in speed of the memory access security check and the speculative execution and 2) visible side-effects of speculative execution in the cache.

 

Fixing 1) would require a clever redesign of the check (if possible at all), and can't be considered a "flaw".

 

Problem 2) is part of a more general issue that some clever tricks to speed up execution expose the execution path of a process, which can be used to infer the value of sensitive secret data. Intel's defence there is basically "security or speed, you can't have both, and you picked speed when you bought our processor". I think they have a point. Moreover, it took some time to find the attack vector, so you can't really accuse intel of negligence. This attack vector, especially its real-life feasibility, is not an obvious one. Because of this, I don't think the class action in court is grounded.

 

I would think problem 2) can be fixed by adding another L1 cache for speculative execution. Its content would be computed to the regular L1 cache when an instruction is retired. I wonder if that's what they did on the more recent processors and this is why these aren't affected?

  • Upvote 1
Link to post
Share on other sites
KpaxBos

KpaxBos 47

Posted
Posted

I would think problem 2) can be fixed by adding another L1 cache for speculative execution. Its content would be computed to the regular L1 cache when an instruction is retired. I wonder if that's what they did on the more recent processors and this is why these aren't affected?

 

 

Hello,

 

Happy new year 2018,may the spectre of hacking not meltdown your enthousiasm to play :salute:

 

Which recent processors are not affected ? I did not find this information in the several news I read.

 

Be Happy

Link to post
Share on other sites
ZachariasX

ZachariasX 2653

Posted
Posted

I wonder, of all the people in terror about this bug, how many are running a system where they use an account with admin priviledges? It is the default setting.

 

As admin, the "thing that is now broken as we have heard" cannot protect anything, even if it would work perfectly fine (and not just working as intended according to Intels press release).

 

On the other hand: datacenters. Who of you honestly thinks that your data stays yours as soon as you blow it onto the drives of someone elses computer?

Link to post
Share on other sites
ZachariasX

ZachariasX 2653

Posted
Posted

Also AMD and Arm cpu do have “problems”

 

 

Like handing out TPM keys to anyone? C'mon. More important that the base is layd to not being allowed to boot anything other than MacOS or Winows.

 

And besides that, we do need complicated chips. If you cannot check all boxes on the list, decision makers are not gonna buy those. And worse, they would be too cheap. So you have VT on your chip even if it is just to play BoX. :)

 

You know, it could be worse. Like when security really matters. Like when you even bother to use RSA keys to get your login. And just imagine, just for one second, that RSA actually stores the seed for the random number generator in your dongle. How does that feel? Bear with me, the feeling will get better, turns out they not only sored it, but that somebody had a good time on their hard drives (German source, sorry) and stole those seeds. To have a further good time on Lockheed Martins hard drives. Something LM of course denied. But who wouldn't.

 

It makes you really wonder what it would take to put RSA out of business. I can't think of anything, honestly.

Link to post
Share on other sites
Mitthrawnuruodo

Mitthrawnuruodo 709

Posted
Posted

I am stopping Windows updates until the effects on performance are clearly known. For me, even the loss of a few percent would not be acceptable.

 

If necessary, I am willing to remain in the current state until I build a new system. Of course, I would not recommend avoiding updates to anyone that cares about the security of their computer.

  • Upvote 1
Link to post
Share on other sites
=362nd_FS=Hiromachi

=362nd_FS=Hiromachi 1161

Posted
Posted (edited)

The windows update is only the first step, then its the bios update that seems to hit hard system, at least based on the following tests made by users:

vbymibcgj9801.png

Source: https://www.reddit.com/r/intel/comments/7oeh84/performance_impact_of_windows_patch_and_bios/

 

TbjpoQ.png

Source: https://www.epicgames.com/fortnite/forums/news/announcements/132642-epic-services-stability-update  

 

 

I just finished running Rise of the Tomb Raider benchmarks, 1080p, very high preset, FXAA.

 
Unpatched:
 
Mountain Peak: 131.48 FPS (min: 81.19 max: 197.02)
Syria: 101.99 FPS (min: 62.73, max: 122.24)
Geothermal Valley: 98.93 FPS (min:76.48, max: 117.00)
Overall score: 111.31 FPS
Windows patch only:
 
Mountain Peak: 135.34 FPS (min: 38.21 max: 212.84)
Syria: 102.54 FPS (min: 44.22, max: 144.03)
Geothermal Valley: 96.36 FPS (min:41.35, max: 148.46)
Overall score: 111.93 FPS
Windows patch and BIOS update:
 
Mountain Peak: 134.01 FPS (min: 59.91 max: 216.16)
Syria: 101.68 FPS (min: 38.95, max: 143.44)
Geothermal Valley: 97.55 FPS (min:46.18, max: 143.97)
Overall score: 111.62 FPS
Average framerates don't seem affected.

https://www.reddit.com/r/pcmasterrace/comments/7obokl/performance_impact_of_windows_patch_and_bios/ds88idv/

 

Considering Il-2 relies heavily on CPU I'm interested in seeing some tests. Such drop in minimums could particularly affect VR players.

Edited by =LD=Hiromachi
Link to post
Share on other sites
ZachariasX

ZachariasX 2653

Posted
Posted

Maybe with subsequent work, things will get better. I seriously doubt that something like that can be made right in one fix. What really is needed, is that Intel, AMD, ARM, etc., update their CPU microcode. Rest is makeshift.

 

But as said, most people don‘t use the safety factor that is in question here. There should be an option to turn it off along with the patch. If you are sitting on your box on an account with administrator priviledges, you can load unsigned kernel drivers anyway and the funtion in question just produces heat for your mancave.

 

If indeed it will result in a significant performance loss, then this will be the Pentium bug from Hell for Intel. I smell class action.

Link to post
Share on other sites
Mitthrawnuruodo

Mitthrawnuruodo 709

Posted
Posted

In my opinion, the uncertainty faced by users in this debacle clearly demonstrates the great flaw of 'Windows as a service'. 

 

It is scary to know that performance is at the mercy of updates that are beyond your control. It leaves very little choice; you either accept the decrease or skip updates altogether. 

 

I know that only a tiny fraction of all users will actually notice the change. However, it has the potential to be significant for those that often analyze frame rates. A well-documented option to opt out would be comforting. 

 

Fortunately, at least the 'BIOS' update stuff remains under our control. I hope it remains this way. 

Link to post
Share on other sites
ZachariasX

ZachariasX 2653

Posted
Posted

What I find more strange, is that we don‘t hear from cloud providers. There, we essentialy get 20% less for our money (we better have a system hardened against such side channel attacks). Anyone seen rebates?

 

First news about that were discreetly coming up in December (I have to further check that) then the Intel CEO dumps all the shares he legally can dump, then suddenly we have this software patch that basically removed the kernel memory mapping from the user MMU mappings, basicalli trippling the latency for this.

 

See here:

http://lists.dragonflybsd.org/pipermail/users/2018-January/313758.html

 

Anything besides updating the CPU microcode is just makeshift.

Link to post
Share on other sites
ZachariasX

ZachariasX 2653

Posted
Posted

Just tested with Passmark, no performance issues since yesterdays Windows patch. One might wonder what that one actually did.

 

Besides,

 

I am stopping Windows updates until the effects on performance are clearly known

Just install Antivirus software, such as from Avira, then you don't get the patch. How bad can it really get? You thought it could get that bad?

Link to post
Share on other sites
coconut

coconut 1490

Posted
Posted

 

 

Considering Il-2 relies heavily on CPU I'm interested in seeing some tests. Such drop in minimums could particularly affect VR players.

 

Meh, in the test you quoted, maximum FPS goes up, which is unexpected. We need to see the distribution of frame times to get a better idea, because just looking at numbers it looks like the test runs were borked.

 

In any case, reports with min-max FPS aren't very reliable, as a freak transient value may skew results significantly. I mean, I care a lot about the min FPS, but not if I have one single frame that for some reason took half a second to compute. What's more interesting is to see how many frames had times above 11ms (90fps), 12ms... and so on until 22ms(45fps)

Link to post
Share on other sites
BeastyBaiter

BeastyBaiter 412

Posted
Posted

For those concerned about BoX performance, here are my results for the balapan test track:

 

Pre-patch:

Min: 60

Max: 91

AVG:87

 

Post-patch:

Min: 68

Max: 91

AVG: 87

 

This is with an i7-8700k at 4.7Ghz (no avx offset) with an Oculus Rift. So yeah, no real change.

Link to post
Share on other sites
dburne

dburne 2583

Posted
Posted

For those concerned about BoX performance, here are my results for the balapan test track:

 

Pre-patch:

Min: 60

Max: 91

AVG:87

 

Post-patch:

Min: 68

Max: 91

AVG: 87

 

This is with an i7-8700k at 4.7Ghz (no avx offset) with an Oculus Rift. So yeah, no real change.

 

Good to hear, thanks for sharing.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share
Followers 1
Go to topic listing
×
×
  • Create New...