Intel processor bug leads to Windows and Linux kernel updates and possible performance hits

[GoodBytes]

58 minutes ago, DarkSmith2 said:

win10 after version 1511 is crippling your performance so hard for gaming anyway. Just look how much services they decided to implement that hard that you cant even disable them anymore. This is what iam running tweaked:

 

You are not going to come even close with version 1709 .. gave me up to 40% more FPS in some titles, pubg went up from 70FPS to 110FPS, after game patches im sitting now on ~140fps with a 2600k and a GTX970, CSGO went up 50FPS aswell.

 

Also noticed horrible stutters before while gaming that are completly gone.

Nha. Your problem is something else. All Microsoft did is break out the services processes.

 

Basically, svchost.exe is a Service that programs can register to and interact with to have basic service capabilities. In other words, say you have an app, and you need to have a running service to do some specific thing, but that is not very special in the 'service world'. You can either make your own Service from scratch and face with many challenges in security issues that you can cause on people computers and system stability issues due to potential bugs, or simply use a pre-made service body that you can register to and interact with with pre-made functionalities that you need and runs in a sort of container, like this you don't need to deal with security and worry about bugs.

 

Windows itself uses this service. 'cause why not? 1 problem to deal with not 20.

But in order to reduce every MB of RAM back in the old days, Microsoft coded the many OS components that uses svchost.exe to be groups of thread in 1 process. The savings is minimal, but back in the days, every KB counted as RAM was really expensive due to technology limitations of the time. The problem with this, is that if 1 service in the group fails, the whole group failed. If there was a memory leak, you don't among the many services have the problem. So now that 4GB of RAM is common on PCs, Microsoft removed grouping thing it does. so each service that uses svchost.exe are separate process. Now if your system is low in RAM, Windows 10 will use the old behavior.

 

Now for references,

I have 168 process and 2271 Threads, with 120415 Handles (fluctuating). I have a potato of PC: Core i7 930 non-OC, GeForce GTX 680, 6GB of RAM, OCZ Vertex 4 256GB SSD. I don't have performance issues, and my games that barely run, don't have a performance impact form the switch.. else I would notice visually.

[DarkSmith2]

52 minutes ago, GoodBytes said:

Nha. Your problem is something else

No its not.

 

I could also write a wall of text with absolutely unnescessary/unrelated facts.

Such as "by default settings windows is holding back 10-20% of your CPU ressources for unnecessary Background processes, its reserving them."

 

Or question yourself.

 

Why do extreme overclockers cut windows down to its bare minimum while also having multiple installs ready for benchmarking to reach the highest scores if it doesnt affect anything?

 

Why did they bound cortana so hard into the search.exe and also make the search.exe necessary to run your system stable at all with latest releases? They dont care if it eats up your performance, they want to spy on you to make money, so being able to manually kill it as a user is undesirable for them. 

 

Why do they implement a gaming mode that literally cripples your gaming performance? You were able to remove the whole thing and edit the priority of your GPU/CPU by yourself without running extra processes for it. And now they even made that impossible to kill completely.

 

As stated, i made my own benchmarks, made my own gaming performance testing, went through different win10 versions with the maximum tweaking i could achieve with the knowledge i had on every version i tried, and the results are pretty clear and more then measurable. And its not that i havent reinstalled 1709 4times just to make sure it isnt just a bad install...

 

 

[erroneum]

This is probably something worth a video since I can imagine people will be wondering why their performance tanks after whichever update it's delivered in.

[Noirgheos]

1 minute ago, erroneum said:

This is probably something worth a video since I can imagine people will be wondering why their performance tanks after whichever update it's delivered in.

Current benchmarks don't show any decrease in gaming and rendering, but that's on Linux.

[DarkSmith2]

1 minute ago, Noirgheos said:

Current benchmarks don't show any decrease in gaming and rendering, but that's on Linux.

which titles did they bench? Hopefully not the ultra GPU bound games

[erroneum]

2 minutes ago, Noirgheos said:

Current benchmarks don't show any decrease in gaming and rendering, but that's on Linux.

That's good to hear; I only use Linux.

[Noirgheos]

2 minutes ago, DarkSmith2 said:

which titles did they bench? Hopefully not the ultra GPU bound games

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

[DarkSmith2]

4 minutes ago, Noirgheos said:

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

looks good

[Noirgheos]

1 minute ago, DarkSmith2 said:

looks good

Yeah, so for the everyday gamer or person that renders things, there won't even be a difference, and it may even improve by like a frame with certain games. The synthetic I/O tests do have me worried though. I think at this point I'll either grab Ryzen+ or stave off upgrading until Intel fixes it through hardware.

[Phentos]

32 minutes ago, Noirgheos said:

Yeah, so for the everyday gamer or person that renders things, there won't even be a difference, and it may even improve by like a frame with certain games. The synthetic I/O tests do have me worried though. I think at this point I'll either grab Ryzen+ or stave off upgrading until Intel fixes it through hardware.

DirectX =/= OpenGL/Vulkan. Wait and see before making assumptions

[Razor512]

I wonder, how much influence does the microcode have over the security features of the CPU?

 

Many drivers make heavy use of various kernel level functions, and there may be some games that may end up getting impacted by this. This can also be an issue with video editing, for example, if storage IOs are suddenly going to have a massive overhead, what happens when working with 4K or 8K raw video where an NVMe SSD will be working at its max read speed when scrubbing through a timeline.

 

What happens when you want to record raw video with fraps?

what happens when you want to record raw video from a capture card?

 

 

[Zeitec]

10 hours ago, TheGlenlivet said:

Any idea what CPUs are affected by this yet? The article doesn't seem to list any other saying Intel or AMD.

From what I've read the fault goes back to the Pentium 3. So yeah basically all of them. 

[Zeitec]

9 hours ago, mrkaru said:

I wonder if there's anything anybody could do about getting refunds or something. a potential 30% decrease in performance is a pretty substantial loss. Especially since it's all down to manufacturing errors and the consumer is in no way to blame. 

Class action law suit seems to be a likely outcome if this is as bad as it seems. 

[Bit_Guardian]

3 hours ago, erroneum said:

That's good to hear; I only use Linux.

Join the real Open Source Master Race. Free BSD For Lyfe!!!

[thorhammerz]

3 minutes ago, Zeitec said:

Class action law suit seems to be a likely outcome if this is as bad as it seems. 

So.... incoming $30 check in the mail, 3 years later, if a claim is filed within a 3 month window ?

[Zeitec]

5 minutes ago, thorhammerz said:

So.... incoming $30 check in the mail, 3 years later, if a claim is filed within a 3 month window ?

If the performance cuts are as big as it seems then it could be a lot more than 30 bucks. 970 owners got that much for a much smaller performance hit. Also this applies to everyone with an intel CPU since like 2007. This could make intel really hurt. 

[thorhammerz]

5 minutes ago, Zeitec said:

If the performance cuts are as big as it seems then it could be a lot more than 30 bucks. 970 owners got that much for a much smaller performance hit. Also this applies to everyone with an intel CPU since like 2007. This could make intel really hurt. 

Thus, you can bet (on that note, might consider placing shorts on BOTH the major chipmaker stocks now  ... jk, only AMD stock is volatile enough for that) Intel will drag out the legal process for as long as possible - I do wonder how much of the settlement will end up being eaten by the lawyers as a result .

[Zeitec]

1 minute ago, thorhammerz said:

Thus, you can bet (on that note, might consider placing shorts on BOTH major chipmaker stocks now ) Intel will drag out the legal process for as long as possible - I do wonder how much of the settlement will end up being eaten by the lawyers as a result .

If there are literally millions of CPUs affected from the past decade the lawyers could literally take tens of millions of dollars and the users could still get a sizeable amount. Also they may just settle to save face. This is a really really bad look for them, and if it's as bad as it seems (I'll repeat that a lot) then Epyc  and Zen+ will be looking really nice. 

[Stefan Payne]

10 hours ago, leadeater said:

Sounds more like all of them to me, newer ones are just less impacted but still affected.

Yes.

And that is why Monopolys and especially monoculture Hardware is a very very bad thing.

When did Intel really do a completely new CPU??
Core Series?? Or was it maybe eventually Nehalem??

 

There must be a reason why they made the fix enabled _ON ALL CPUS BY DEFAULT_ and not just some.
But manufacturers can opt out with a good reason....

 

That alone leads to believe that this issue is bigger than everything we had so far.

 

PS: I was the first to post it in some other thread.

And here's my source:
https://www.forum-3dcenter.org/vbulletin/showthread.php?t=585993

 

So it looks like there is a good reason to throw away older and newer Intel Servers...

And to stop recommending Intel CPUs until the fix is out and we know how it will impact it.

 

And it seems something similar is happening on ARM CPUs...

[leadeater]

4 hours ago, Stefan Payne said:

So it looks like there is a good reason to throw away older and newer Intel Servers...

There is no reason to do this after the patch and most servers do not use all their resources, most average 30%-50% utilization. Performance impact is still largely unknown and very dependent on workload. The largest impact is likely going to be on software switch fabric servers and is rather a big deal considering the rapid move over to SDN.

 

4 hours ago, Stefan Payne said:

And to stop recommending Intel CPUs until the fix is out and we know how it will impact it.

Not possible, because there needs to be an alternative option by your hardware vendor and currently AMD EPYC servers are largely non-existent. You can't buy an AMD HPE DL380 Gen 10 because it doesn't exist.

 

You can do it for desktop computers but servers you still by in large only have one choice, Intel (for x86).

 

4 hours ago, Stefan Payne said:

And it seems something similar is happening on ARM CPUs...

Sounds like more of a precaution, there is likely a known way to exploit systems if vulnerable that is not hardware specific to carry it out so they are patching everything then likely will go back and reassess what is and isn't needed.

[GoodBytes]

10 hours ago, DarkSmith2 said:

No its not.

 

I could also write a wall of text with absolutely unnescessary/unrelated facts.

Such as "by default settings windows is holding back 10-20% of your CPU ressources for unnecessary Background processes, its reserving them."

Side note: I forgot to mention, about the service separation,it also increases security. A virus that affects 1 service, can't jump in to the next service memory from the same group, as it is a different process and not thread.

 

On your reply:

What are you talking about?

 

Look at the end bit where I left my PC idle. It is all 0%.

 

 

 

Quote

Why do extreme overclockers cut windows down to its bare minimum while also having multiple installs ready for benchmarking to reach the highest scores if it doesnt affect anything?

They do this to get every single point possible. When you run something, you share resources with other process even if they consume 0 CPU cycle. If you have a Quad core, your CPU can only do 4 things at the same time, so Windows (like any other OS including Android and iOS powered phone and tablet), have a process scheduler. You can read up on all of this, but each process has a given time for the CPU, the process can, if it wants, give up its remaining time or not, but it can't say, "I need more time". The process is paused, and the Process Scheduler goes on the next process. This is done at a super rapid rate, and gives you the illusion that you can do more than the number of cores you have. The less process you have, the shorter the list, making the OS return to their benchmark quicker, and get that extra point or two, especially that they ensure that some other process doesn't kick-in (network connectivity check, some driver requirement by some hardware, search index kicks in, virus check on a created file by a process, etc.) If the top score is: 1825 on some benchmark, getting 1826 is beating that high score. So they want everything on their side.

 

Quote

Why did they bound cortana so hard into the search.exe and also make the search.exe necessary to run your system stable at all with latest releases? They dont care if it eats up your performance, they want to spy on you to make money, so being able to manually kill it as a user is undesirable for them. 

Huh? It mean yes it is integrated, but how does it affect performance? Beside web searches, and I guess voice analysis (assuming you have this on, its off by default), and some reminder related stuff (off by default, and has no reminders on a clean install of OS as none are setup), it doesn't affect anything. I think you really mean Search Indexer? But that isn't new. It was there since Vista. You can manually run it from the Indexer panel, you can let it run by leaving your computer idle... but I think more that you have Windows Maintenance kicking-in while you do stuff. You can change the time by doing Start > type: Maintenance > Pick: Security & Maintenance. A panel will open, on it, expand; Maintenance, than pick to change the time. (you can also trigger the maintenance now with the blue link next to change schedule time). Remember that if your computer goes to sleep, the maintenance process can't continue.

 

Quote

Why do they implement a gaming mode that literally cripples your gaming performance? You were able to remove the whole thing and edit the priority of your GPU/CPU by yourself without running extra processes for it. And now they even made that impossible to kill completely.

Game Mode help boost minimum frame rate and frame time on Win32 games, and reduce layers on UWP Games bringing them to the same performance as Win32 ones. Game Mode can reduce a game performance if the game is multi-processed, and Game Mode does not support that game. A possible multi-processed layout of a game, is that you have the game draw executable which displays the frame buffer and takes inputs, than you have another process that process the game itself including AI, and then you have another process that does the GPU cull, getting environment details, textures, etc. If Game Mode sets the Draw process to high priority, the others drops in priority as expected, and they have trouble running, making your game not run smooth.

 

You can post the game that has issues in the Feedback hub, or/and disable game mode for that game.

If you have a powerful system, Game Mode should give you minimal to no performance benefits for Win32 games. If you have a potato computer, it will help a bit getting a bit better gaming experience. Anything helps, helps. UWP games gets the performance boost it much needs.

 

Quote

As stated, i made my own benchmarks, made my own gaming performance testing, went through different win10 versions with the maximum tweaking i could achieve with the knowledge i had on every version i tried, and the results are pretty clear and more then measurable. And its not that i havent reinstalled 1709 4times just to make sure it isnt just a bad install...

Ok.. and no one else has problems. Ok.

[su /bin/bash]

I care about me is maximum productivity, security does not any matter. I think too much productivity is wasted in the name of security. The problem of loss of performance exchanged for security is so great that it slows the progress of all mankind.

[Stefan Payne]

30 minutes ago, leadeater said:

There is no reason to do this after the patch and most servers do not use all their resources, most average 30%-50% utilization. Performance impact is still largely unknown and very dependent on workload.

Phoronix made benchmarks.

IF we are talking about Database Servers like the one for this forum.

Well, you want to replace the server ASAP because the hit is THAT great...

 

30 minutes ago, leadeater said:

Not possible, because there needs to be an alternative option by your hardware vendor and currently AMD EPYC servers are largely non-existent. You can't buy an AMD HPE DL380 Gen 10 because it doesn't exist.

 

You can do it for desktop computers but servers you still by in large only have one choice, Intel (for x86).

I was talking about Desktop systems, here in this Forum, not about Servers in this case.

Because whoever gets a server won't ask here in this forum...

 

30 minutes ago, leadeater said:

Sounds like more of a precaution, there is likely a known way to exploit systems if vulnerable that is not hardware specific to carry it out so they are patching everything then likely will go back and reassess what is and isn't needed.

I don't know if ARM said something about it or not but lets wait and see.

 

Its entirely possible that the fallout might be greater than we all think but time will tell...

[ariz]

So this flaw has been going on for years and years and no one at Intel knew about this until now??

[Stefan Payne]

6 minutes ago, su /bin/bash said:

I care about me is maximum productivity, security does not any matter. I think too much productivity is wasted in the name of security. The problem of loss of performance exchanged for security is so great that it slows the progress of all mankind.

Why don't you say:

 

Hey, my computer is open for attack and part of a BOTNET. Because that's what you are saying. You don't seem to understand what Security is about and for.