Microsoft Kernel Patch CPU Before and After Benchmarks Thread


Recommended Posts

LimeMaster
1 minute ago, Yogurth said:

Yeah, both of my systems are Intel+Nvidia.

Might end up being an unlucky combo. I guess we'll just have to wait and see. :/

Link to post
Share on other sites
adrynalyne
Just now, LimeMaster said:

Might end up being an unlucky combo. I guess we'll just have to wait and see. :/

That would be unfortunate.

  • Like 1
Link to post
Share on other sites
DeusProto

RE: Microcode updates. For older motherboard users who likely are never going to get official EFI updates, VMWare has a tool that can update a CPU's microcode

 

https://labs.vmware.com/flings/vmware-cpu-microcode-update-driver

 

Good luck finding the relevant 'microcode.dat' file needed for Intel CPUs though. I haven't yet located where Sandybridge's latest one is.

The only one I was able to find was for Linux distributions, I guess specific to virtualization of Linux distributions, nothing pertinent to Windows.

Edited by DeusProto
  • Like 2
Link to post
Share on other sites
Mando
On 1/7/2018 at 3:37 AM, LimeMaster said:

I understand why Linus is angry then.

cos hes an attention seeking bellend? What bout his beloved ARM he proclaims, they are affected also......yup confirmed intel hating bellend! 

 

 

  • Like 3
Link to post
Share on other sites
Yogurth
2 hours ago, DeusProto said:

RE: Microcode updates. For older motherboard users who likely are never going to get official EFI updates, VMWare has a tool that can update a CPU's microcode

 

https://labs.vmware.com/flings/vmware-cpu-microcode-update-driver

 

Good luck finding the relevant 'microcode.dat' file needed for Intel CPUs though. I haven't yet located where Sandybridge's latest one is.

The only one I was able to find was for Linux distributions, I guess specific to virtualization of Linux distributions, nothing pertinent to Windows.

I am in the same situation as You are. Nothing from either ASUS or Intel for Sandy Bridge. In fact, I am almost certain Intel will not release anything for Sandy Bridge and older since, theoretically, these chips would be more affected by the slowdown than Broadwell and later. I still can't find anything on Haswell V1 either.

Link to post
Share on other sites
Godflesh

It's exhausting when you all think you are testing old CPU's when it's been said the problem goes back to CPU's from 1995.

Seriously all the sarcastic and flippant little remarks about it being overblown is silly nonsense.

I seen 1 guy bench an newer i7 then right away a guy made a crack about the problem being no big deal.

I wouldn't have a problem with that had people actually done testing first.. but they haven't.

Clearly most of you did not hear what was said in the first place.

No one anywhere on the internet said *ALL* CPU's would get a 20/30% CPU penalty. (The same amount of speed loss)

You are making that up in your head.

What was ACTUALLY said  is could vary by CPU and WHAT YOU ARE DOING.

Get it yet ?

It's tiring when people fail miserably to comprehend what is posted int he stories.

Then.. get extremely rude snotty arrogant and offensive about it.

 

My "old" Haswell 4770k i7 i used to have was more powerful than 20 of the PC's i am using right now.

So think about it.. is your old dual core from 2005 going to be more affected than a CPU from last year (that you call "old") ?

 

I don't like the It's not a problem at all guys or.. the hysteria people saying it;s the end of the world.

Can we get a fair HONEST and accurate balance that is based of of REAL testing ?

And don't forget people already made their opinions / judgments for all CPU's based on the first few "old" CPU;'s people benched.

 

I've been bench'ing for a decade or two and already have a pile of software installed.

I will do some testing but i was waiting for WU to offer the patch to me via Windows update.

I am not inclined to seek out and manually install updates.. on Windows 7.

I have seen 1 or 2 problems with WU over the last 20 years. LOL

Remember when it was a website only ? ;)

 

So.. system calls you say ?

Well i wonder what to use then.. my all time fav was LinX for Intel and an AMD patched version.

I've found it to be a real CPU crusher when hunting for instability with your OC. (better than others like Prime95)

I've used all the old popular ones of course.

 

Anyway once i know what to test with (That utilizes system calls) what ever that means.. and i am offered the update i will test and post.

I know what a system call is generally speaking.

For example i have coded .dll emulator stubs before that patched the export table and emulated internal built in functions on programs.

I have coed a lot of programs and cracked / reverse engineered a lot of "targets" ranging from Windows apps to Linux to Android to Java etc.

People are repeating this "systems call" comment all over vaguely.

it matters then to choose a benchmark that will ACTUALLY test the issue.

Vs people testing with a test program that will never show a difference.

 

Lastly don't mistake my comments for anything other than the truth.

I like AMD and Intel etc.. i have no desire to cheerlead or jump chips or minimize or hype up the problem.

I simply desire the honest facts.. nothing more.

I started reading this to see "old" benchmarks and i didn't really see any after a few pages.. but a whole lot of sarcastic comments.

I have logged bench stats from many app's going back to 2010 and earlier.

I test everything and log everything and have it all stored. (655 files in 60 folders) and maybe more "Becnhmark stats/pics" elsewhere i forgot about.

  • Like 2
  • Thanks 1
Link to post
Share on other sites
Mando

Nice post godflesh :) 

 

@warwagonmight be worth asking to mods to clean up the thread to leave just benchmark outputs. Im guilty as the rest going off on a tangent.

Link to post
Share on other sites
muratoner

i'm really curious as to what's gonna happen with the old motherboards, i mean really old, something that was manufactured in the late 90's or early 2000. I'm pretty sure  these boards with core2duo/quad or AMD X2/X4/Phenom machines won't get any bios updates anymore. Does this mean they'll always be vulnerable to these exploits?

Link to post
Share on other sites
Jampe

My tests with performancetest 64bit by passmark shows significant changes. Results dropped from 1600 points to 1200. Expecially 2D graphics suffered big way, over 30 % performance loss

Link to post
Share on other sites
LimeMaster
3 hours ago, Mando said:

Nice post godflesh :) 

 

@warwagonmight be worth asking to mods to clean up the thread to leave just benchmark outputs. Im guilty as the rest going off on a tangent.

Don't request that. Neowin needs to feel more lively to get encourage members to participate.

  • Like 1
Link to post
Share on other sites
adrynalyne
49 minutes ago, Jampe said:

My tests with performancetest 64bit by passmark shows significant changes. Results dropped from 1600 points to 1200. Expecially 2D graphics suffered big way, over 30 % performance loss

Why would 2d or 3D graphics suffer when gpu code isn’t x86 as far as I know? The fix is only for x86. 

Link to post
Share on other sites
+warwagon
1 hour ago, muratoner said:

i'm really curious as to what's gonna happen with the old motherboards, i mean really old, something that was manufactured in the late 90's or early 2000. I'm pretty sure  these boards with core2duo/quad or AMD X2/X4/Phenom machines won't get any bios updates anymore. Does this mean they'll always be vulnerable to these exploits?

Well how about the motherboard the average user is using? Do you really think they will ever download a bios update?

  • Like 1
Link to post
Share on other sites
adrynalyne
5 hours ago, Godflesh said:

It's exhausting when you all think you are testing old CPU's when it's been said the problem goes back to CPU's from 1995.

Seriously all the sarcastic and flippant little remarks about it being overblown is silly nonsense.

I seen 1 guy bench an newer i7 then right away a guy made a crack about the problem being no big deal.

I wouldn't have a problem with that had people actually done testing first.. but they haven't.

Clearly most of you did not hear what was said in the first place.

No one anywhere on the internet said *ALL* CPU's would get a 20/30% CPU penalty. (The same amount of speed loss)

You are making that up in your head.

What was ACTUALLY said  is could vary by CPU and WHAT YOU ARE DOING.

Get it yet ?

It's tiring when people fail miserably to comprehend what is posted int he stories.

Then.. get extremely rude snotty arrogant and offensive about it.

 

My "old" Haswell 4770k i7 i used to have was more powerful than 20 of the PC's i am using right now.

So think about it.. is your old dual core from 2005 going to be more affected than a CPU from last year (that you call "old") ?

 

I don't like the It's not a problem at all guys or.. the hysteria people saying it;s the end of the world.

Can we get a fair HONEST and accurate balance that is based of of REAL testing ?

And don't forget people already made their opinions / judgments for all CPU's based on the first few "old" CPU;'s people benched.

 

I've been bench'ing for a decade or two and already have a pile of software installed.

I will do some testing but i was waiting for WU to offer the patch to me via Windows update.

I am not inclined to seek out and manually install updates.. on Windows 7.

I have seen 1 or 2 problems with WU over the last 20 years. LOL

Remember when it was a website only ? ;)

 

So.. system calls you say ?

Well i wonder what to use then.. my all time fav was LinX for Intel and an AMD patched version.

I've found it to be a real CPU crusher when hunting for instability with your OC. (better than others like Prime95)

I've used all the old popular ones of course.

 

Anyway once i know what to test with (That utilizes system calls) what ever that means.. and i am offered the update i will test and post.

I know what a system call is generally speaking.

For example i have coded .dll emulator stubs before that patched the export table and emulated internal built in functions on programs.

I have coed a lot of programs and cracked / reverse engineered a lot of "targets" ranging from Windows apps to Linux to Android to Java etc.

People are repeating this "systems call" comment all over vaguely.

it matters then to choose a benchmark that will ACTUALLY test the issue.

Vs people testing with a test program that will never show a difference.

 

Lastly don't mistake my comments for anything other than the truth.

I like AMD and Intel etc.. i have no desire to cheerlead or jump chips or minimize or hype up the problem.

I simply desire the honest facts.. nothing more.

I started reading this to see "old" benchmarks and i didn't really see any after a few pages.. but a whole lot of sarcastic comments.

I have logged bench stats from many app's going back to 2010 and earlier.

I test everything and log everything and have it all stored. (655 files in 60 folders) and maybe more "Becnhmark stats/pics" elsewhere i forgot about.

If you were referring to me as he person who “brushed” off the newer i7, that isn’t true. I said more testing needed to be done. Too many false positives, unrelated changes, and fake results out there to be trusting single reports. 

Link to post
Share on other sites
+warwagon
52 minutes ago, LimeMaster said:

Don't request that. Neowin needs to feel more lively to get encourage members to participate.

 

4 hours ago, Mando said:

Nice post godflesh :) 

 

@warwagonmight be worth asking to mods to clean up the thread to leave just benchmark outputs. Im guilty as the rest going off on a tangent.

All I care about is comments and thread views. As long as my thread gets bumped to the top, stays active and gets lots of comments and thread views i'm happy.

 

Right now I have a thread that is on page 5, 113 replies and 3,553 Views, i'm thrilled!

  • Like 2
  • Thanks 1
Link to post
Share on other sites
muratoner
15 minutes ago, warwagon said:

Well how about the motherboard the average user is using? Do you really think they will ever download a bios update?

Good point. i'm afraid we might hear about web sites stealing passwords/credit card info's in masses.

Link to post
Share on other sites
Mando

right its now been reported some NVidia cards/tablets are  susceptible.

 

Alice in wonderland effect :p 

 

https://www.nvidia.com/en-us/product-security/

 

Latest drivers 390.65 

 

NVIDIA is providing an initial security update to mitigate aspects of Google Project Zero’s January 3, 2018 publication of novel information disclosure attacks that combine CPU speculative execution with known side channels.

The vulnerability has three known variants:

Variant 1 (CVE-2017-5753): Mitigations are provided with the security update included in this bulletin. NVIDIA expects to work together with its ecosystem partners on future updates to further strengthen mitigations.

Variant 2 (CVE-2017-5715): NVIDIA’s initial analysis indicates that the NVIDIA GPU Display Driver is potentially affected by this variant. NVIDIA expects to work together with its ecosystem partners on future updates for this variant.

Variant 3 (CVE-2017-5754): At this time, NVIDIA has no reason to believe that the NVIDIA GPU Display Driver is vulnerable to this variant.

 

there we have it guys. 

 

Affected Products

 

  • GeForce, Quadro, NVSWindows, Linux, FreeBSD, Solaris
  • Tesla Windows, Linux

 

Get updating peoples! 

Link to post
Share on other sites
Mando
1 hour ago, warwagon said:

Well how about the motherboard the average user is using? Do you really think they will ever download a bios update?

no but they should be using a half decent AV which should now protect against the issue. If your chosen vendor doesnt yet, bin them and use one that does :) 

Link to post
Share on other sites
Yogurth
1 hour ago, Mando said:

right its now been reported some NVidia cards/tablets are  susceptible.

 

Alice in wonderland effect :p 

 

https://www.nvidia.com/en-us/product-security/

 

Latest drivers 390.65 

 

NVIDIA is providing an initial security update to mitigate aspects of Google Project Zero’s January 3, 2018 publication of novel information disclosure attacks that combine CPU speculative execution with known side channels.

The vulnerability has three known variants:

Variant 1 (CVE-2017-5753): Mitigations are provided with the security update included in this bulletin. NVIDIA expects to work together with its ecosystem partners on future updates to further strengthen mitigations.

Variant 2 (CVE-2017-5715): NVIDIA’s initial analysis indicates that the NVIDIA GPU Display Driver is potentially affected by this variant. NVIDIA expects to work together with its ecosystem partners on future updates for this variant.

Variant 3 (CVE-2017-5754): At this time, NVIDIA has no reason to believe that the NVIDIA GPU Display Driver is vulnerable to this variant.

 

there we have it guys. 

 

Affected Products

 

  • GeForce, Quadro, NVSWindows, Linux, FreeBSD, Solaris
  • Tesla Windows, Linux

 

Get updating peoples! 

Now that Nvidia finally admitted it, a benchmark would be nice on a fully slowed system... secured that is :)

  • Like 1
Link to post
Share on other sites
Mando

Now that the NVidia fix is out in 390.65 benchies below against post Win patch. Lower scores for NVidia patch in Bold.

 

Cinebench 4D BEFORE Win patch

CPU Benchmark = 966CB

OpenGL Bench = 140 FPS

AFTER

CPU Benchmark = 975CB ???

OpenGl Bench = 149.95 FPS ???

NVidia Driver update

CPU Benchmark = 962CB

OpenGL Bench =147.64FPS
 

3D Mark timespy extreme BEFORE

3998

AFTER

3997 ????

NVidia Driver Update

3971

 

Novabench 

2713

After

2703

NVidia Driver update

2677
 

RealBench BEFORE

Image Editing 212,486/ Time 25.0745

Encoding 93,501/time 56.9829

OpenCL 113,529/ Ksamples per sec 20877

Heavy Multitasking 110,647

Time 68.9761

System Score 132,540

RealBench AFTER

Image Editing 208.830/Time 25.5135

Encoding 92,841/time 57.3881

OpenCL 113,529/ Ksamples/sec 20922

Heavy Multitasking 103,867/Time 73.4779

System Score 129,766

 

Post NVidia 

Image Editing 108,788/ Time 48.9757

Encoding 84,939/time 62.7273

OpenCL 113,529/ Ksamples per sec 20948

Heavy Multitasking 99,848

Time 76.4357

System Score 101,776 

 

A drop of 28k overall in Realbench is a large amount.

 

So a bigger hit overall post NVidia patch, not good IMO :( hopefully post Microcode/Biosd update it recoups some of the lost perf. Once Microcode/Bios is out ill benchmark a 3rd time.

 

Must be noted, this is not extensive testing, did a single run where possible.

 

14 minutes ago, Yogurth said:

Now that Nvidia finally admitted it, a benchmark would be nice on a fully slowed system... secured that is :)

 

  • Like 1
Link to post
Share on other sites
Yogurth
34 minutes ago, Mando said:

Now that the NVidia fix is out in 390.65 benchies below against post Win patch. Lower scores for NVidia patch in Bold.

 

Cinebench 4D BEFORE Win patch

CPU Benchmark = 966CB

OpenGL Bench = 140 FPS

AFTER

CPU Benchmark = 975CB ???

OpenGl Bench = 149.95 FPS ???

NVidia Driver update

CPU Benchmark = 962CB

OpenGL Bench =147.64FPS
 

3D Mark timespy extreme BEFORE

3998

AFTER

3997 ????

NVidia Driver Update

3971

 

Novabench 

2713

After

2703

NVidia Driver update

2677
 

RealBench BEFORE

Image Editing 212,486/ Time 25.0745

Encoding 93,501/time 56.9829

OpenCL 113,529/ Ksamples per sec 20877

Heavy Multitasking 110,647

Time 68.9761

System Score 132,540

RealBench AFTER

Image Editing 208.830/Time 25.5135

Encoding 92,841/time 57.3881

OpenCL 113,529/ Ksamples/sec 20922

Heavy Multitasking 103,867/Time 73.4779

System Score 129,766

 

Post NVidia 

Image Editing 108,788/ Time 48.9757

Encoding 84,939/time 62.7273

OpenCL 113,529/ Ksamples per sec 20948

Heavy Multitasking 99,848

Time 76.4357

System Score 101,776 

 

A drop of 28k overall in Realbench is a large amount.

 

So a bigger hit overall post NVidia patch, not good IMO :( hopefully post Microcode/Biosd update it recoups some of the lost perf. Once Microcode/Bios is out ill benchmark a 3rd time.

 

Must be noted, this is not extensive testing, did a single run where possible.

 

 

Seems like we are going to end up with Phoronix results on Linux :( If further testing confirms Your small test industry is going back to Haswell performance. 

 

 

 

I hear AMD is having a party at CES, entire Vegas is invited :)

  • Haha 1
Link to post
Share on other sites
Mando
3 minutes ago, Yogurth said:

Seems like we are going to end up with Phoronix results on Linux :( If further testing confirms Your small test industry is going back to Haswell performance. 

 

 

 

I hear AMD is having a party at CES, entire Vegas is invited :)

tbh that party may be shortlived, once we know more about Spectre exploits. 

 

Who knows tbh, until its all public its speculation.

Link to post
Share on other sites
adrynalyne
1 minute ago, Mando said:

tbh that party may be shortlived, once we know more about Spectre exploits. 

 

Who knows tbh, until its all public its speculation.

You are punny. 

  • Haha 1
Link to post
Share on other sites
Mando
18 minutes ago, adrynalyne said:

You are punny. 

i shoulda said purely speculative eh :p teehee 

Link to post
Share on other sites
Jampe
12 hours ago, adrynalyne said:

Why would 2d or 3D graphics suffer when gpu code isn’t x86 as far as I know? The fix is only for x86. 

as far as I understand this problem affects everything that goes between user mode app and OS itself.OS is isolated of the user data app by cpu feature.

 If you measure cpu speed only it will not give a good idea how much have changed. You have to measure overhead between your application and OS (drivers included). That means pretty much everything is slower and depending on the task at hand, some more some less. Tasks that require a lot of data passed from user mode to OS will be slower. Am I right?

Link to post
Share on other sites
adrynalyne
2 minutes ago, Jampe said:

as far as I understand this problem affects everything that goes between user mode app and OS itself.OS is isolated of the user data app by cpu feature.

 If you measure cpu speed only it will not give a good idea how much have changed. You have to measure overhead between your application and OS (drivers included). That means pretty much everything is slower and depending on the task at hand, some more some less. Tasks that require a lot of data passed from user mode to OS will be slower. Am I right?

Maybe? This isn't my forte. As it was explained to me, it matters if it is x86 code instructions being called with sys calls. The GPU isn't x86 as far as I know.

Link to post
Share on other sites
This topic is now closed to further replies.
  • Recently Browsing   0 members

    No registered users viewing this page.