Windows Systems Performance Impacts from Spectre and Meltdown

Joined
Aug 17, 2016
Messages
704
My parents each have Q6600 and E8400 CPUs respectively. Yes, ten year old old CPUs but just fine for their basic needs, though updated and newer software does tax them more than in the beginning. These patches, if they are even released for these CPUs, will make these CPUs useless with today's software, maybe even the old stuff. Annoying. I need to understand the risks of not patching. If it is low I may just avoid patching.
My daughter is using my old Q6600 with a GTX 950 for Sims and Overwatch I hope it doesn't make it any worse. Going to be a year before I can build her a new one.
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189
Anything you use to access the internet, or on the edge of your network infrastructure should be aggressively tested and patched. Desktops at home, Patched. Workstations in the office. Tested/patched ASAP. Web servers, PATCHED, anything users access to see the internet. PATCHED. Shared resources.. PATCHED. (meaning VM hosts for VDI's, or Citrix hosts.)
 

RedWagnum

Gawd
Joined
Mar 30, 2007
Messages
588
>>Come on folks. It's a bug, a fault, design flaw. Just like the Pentium math bug from years ago....

Not really.

This is not a compute problem like the math bug, it's not getting wrong results during computation.
It's not even really a flaw.

The CPUs work just fine. If humans were all honest and trust worthy, none of this would be an issue.

It's definitely a security problem, but it's not a bug.

Not picking on you or trying to start anything. It's being commonly called a bug and a flaw.
It just helps to frame it properly for what it is so we can all understand the issue better.

.
You say po-ta-to, I say po-tah-to. :D Any way you look at it, they are both problems at the silicon level.

I agree with you about human nature. If Intel had known 10 years ago how despicable some humans could be it probably would have designed a more secure CPU. Problem is that Intel doesn't know how bad things will be 10 years from now or 1 year from now for that matter. The next generation CPU may not have this issue but a few years down the road some rotten apple may find another. Don't know what the answer is. Short of changing human nature, I don't really think there is one. Someone will always want what someone else has and go to great lengths to get it.

Well if it isn't a bug, it's bad design. Or at least flawed design.

I get what you mean -- things are working exactly as intended -- but failure to foresee that one's intentions could be exploited in a way that fundamental compromises a device's basic purpose is definitely a design issue.
Yeah, man. Give me back my TRS-80 Model 4 and 9600 baud modem.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,166
You think a TRS-80 could handle the throughput of a 9600 baud modem? Come on man that's clearly 386 territory with a math co-processor. (making it a 386 DX)

You are confusing 486sx/dx with 386sx/dx ;)




To clarify for the young generation
386 SX/DX different is the bandwidth of the external bus

486 SX/DX difference is the presence of a 387 math processor
Thereby upgrading them with a 487 math processor making them the same
 

Megaslug

Limp Gawd
Joined
Jul 30, 2015
Messages
173
You think a TRS-80 could handle the throughput of a 9600 baud modem? Come on man that's clearly 386 territory with a math co-processor. (making it a 386 DX)
Easily, I did system to system transfers with null modem cables at 19.2K with my Model 4. Radio Shack even had a classroom network system that hooked 16 systems to a teacher master using 9600 bps serial connections.
 

MMitch

Gawd
Joined
Nov 29, 2016
Messages
807
LOL I was talking actual size of the image on my screen not the size of the image in KB. But I remember line by line pron as well as the next man.
You're blind ? I guess even blind people have pr0n :wideyed: ... line by line...
Ok ok not a good one :D
 

focbde

Gawd
Joined
Jan 31, 2008
Messages
545
wanted to say it explains why my old i3 laptop runs like shit now... but to be fair it always has. Always on the fence of whether or to put an SSD into a $400 budget laptop to speed it up.
Just my opinion of course, but I'd say jump off the fence into SSD land. The speed bump is significant no matter what the laptop, and always worth it - plus, you can always use it in your next budget laptop if it also comes with a HDD. :)
 

kju1

2[H]4U
Joined
Mar 27, 2002
Messages
3,058
Are you naive enough to swallow the story they say? Sorry I don't own a tin foil hat just the shield of cynicism..... and a sword of sarcasm...
Maybe I just dont see conspiracy theories everywhere...
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,166
how convenient for win7/8 to be victims of the biggest performance hit, you sure you don't want win 10 ?
or does it means win10 was already slowed down with something similiar
I mean sombody sitting on the floor also falls less than a guy sitting on a chair... :D

Asbestos suit on
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189

RedWagnum

Gawd
Joined
Mar 30, 2007
Messages
588
I have absolutely no idea what you mean. But hey, that's OK.
:D In reference to:
-- things are working exactly as intended --
My old Model 4 always worked as intended...

You think a TRS-80 could handle the throughput of a 9600 baud modem? Come on man that's clearly 386 territory with a math co-processor. (making it a 386 DX)
That old 3.78 MHz Z80 based Model 4 would easily run two 9600 baud modems -- helped run a two line (at the time) BBS for a few years before we moved on to x86 based systems.
 

Batboy88

Limp Gawd
Joined
Dec 25, 2017
Messages
323
Slowed down some on the nvme might be cause it's kinda half way full though. Nothing real Major though, still pretty speedy.
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189
Yea for desktop performance with close to current in generation of CPU we are good. But it's still BS. VM hosts.... well I haven't seen any tests on current host level hardware. Just some guy running ESXi on his desktop hardware claiming a 50% I/O impact.
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189

4saken

[H]F Junkie
Joined
Sep 14, 2004
Messages
11,067
upgraded, win10 7700k, didnt see a blip on any bench/game i have run. Buddy with an 8700 saw similar results, nothing changed. so overblown.
 

heatlesssun

Extremely [H]
Joined
Nov 5, 2005
Messages
44,155
upgraded, win10 7700k, didnt see a blip on any bench/game i have run. Buddy with an 8700 saw similar results, nothing changed. so overblown.
Typical desktop scenarios weren't expected to see much impact. I/O intensive scenarios were the ones expected to see the most impact and that sees to be the case, a big problem for things like SQL databases.
 

4saken

[H]F Junkie
Joined
Sep 14, 2004
Messages
11,067
Typical desktop scenarios weren't expected to see much impact. I/O intensive scenarios were the ones expected to see the most impact and that sees to be the case, a big problem for things like SQL databases.
right. not looking forward to seeing when our virtual infra is updated. Just answering all of the chicken littles on trying to claim intel was screwed for home use.
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189
When I was in the hospital a couple of weeks ago, I was amazed that some of the mobile equipment in use was still running XP.
In a lot of cases it's because the vendors for the software are in what we call... a sweet spot. They've created the software and it works. Sure a reboot might be needed from time to time but other than that... software is functional. Now if a customer wants the software to run on a newer OS... sure they can do that.... but it will cost the customer. Even in the industry I am in every new feature we want they want $$$ for it even though they are going to just re sell it to our competitors.
 

JMccovery

2[H]4U
Joined
Feb 9, 2006
Messages
2,211
In a lot of cases it's because the vendors for the software are in what we call... a sweet spot. They've created the software and it works. Sure a reboot might be needed from time to time but other than that... software is functional. Now if a customer wants the software to run on a newer OS... sure they can do that.... but it will cost the customer. Even in the industry I am in every new feature we want they want $$$ for it even though they are going to just re sell it to our competitors.
It was a bit worrying when the unit BSOD'd while the nurse was trying to check my vitals.
 

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189
We just completed some testing.

Dell R930 server.
Windows 2012 R2 all current patches.
SQL 2014Sp2 with patch for Specter.
Updated Dell BIOS to 4.5.
I/O is a VNX 5400 mostly operating in SSD and the 1TB of flash cache is avaialble. Connected via 2 dedicated 8GB FC cards.

Load testing with production like load. Pre patch we were looking at 4-5% CPU and our Operations per second in the 2k range.
Post patching testing the same... and the numbers were the same.

This is a bare metal R930
2 EC8867 CPU's.
384GB of ram.


SQL takes 377GB and the DB itself is around 1.7TB uncompressed. We use no active compression in SQL.


Just thought I would share for anyone else our there. The VAST MAJORITY of our memory and operations are in the user memory space. So due to that we didn't expect a large impact because of how our DBA's have designed our DB.

Next we will be deploying to production after a week of observation and review.


I want to note in our initial round of testing we had not applied the Registry switches to enable the FREAKING patches. We have now done so and are collecting data for a few hours to prove out. Inital impression no update.

No discernible change with the patches in place.
 
Last edited:

Grimlaking

2[H]4U
Joined
May 9, 2006
Messages
3,189
Another fun update. Checked Dell's site to download some BIOS's for further testing on other Dell platforms.. Dell has pulled all of their enterprise BIOS Updates. No statement as to why. I suspect because of crashing.
 
Top