Question SSD OS - Tertiary Mechanical HDD access slows entire system

JackrumMadthing

Distinguished
Mar 31, 2014
32
4
18,545
Hallo Toms Hardware Netizens,

Lets open by stating ich bien kein noob! Decades of blowing up things and then figuring out how to put them together again from my BBC Micro days in the early 80s, to still building machines and being the general know it all.

Apparently, I don't know it all.



SYSTEM:
Intel i5-12400f, 24GB Corsair DDR4 3600mhz, MSI B660M mobo, GTX 1080 8gb, Win10 Enterprise N / 21H2 19044.2604

OS is on WD 500GB SSD, chopped into 70gb and 400 because they <Mod Edit> ... me on the box, 456gb capacity bastards (ya ya).

SATA is on AHCI, all are plugged into 6gb/s. SMART reports all happy.
  • WD 240GB SSD (S: Drive)
  • WD 500GB SSD (C: & R: Drive)
  • WD 1TB M2 (put OS on it for a day, no noticable difference, rather lower the wear and tear....so SSD boot) (M: Drive)
  • WD 2TB HDD 5400rpm (K: Drive)
  • Seagate 2TB HDD 7200rpm (D: & E & F: Drives)
Furthermore all relevant drivers are in happy states and firmware etc., necessary KB's and feature updates as well as runtimes etc all behaving. Scheduling is set for Background, and Virtual RAM is set to a fixed 20gb on the M2 (Star Citizen needs it amazingly).

Evil Services have been torn out:
Hibernate, (when did it change its name to SysMain???), Biometric, Backup, Auto Time Zone, AVCTP, Connected DEvices, User Experiences, Data Usage, Geolocation, Intel Dynamic Application Loader & Management Engine, Shadow Copy, Store Install (grrr), Print Spooler, Radio Management (LAN), Security Center, Sysmain (Superfetch...when did that change its name??), Touch, User Experience, Windows Backup, Biometric, Search (i use Everywhere...its awesome)

Windows Update has been manually told to shutup, and Indexing is off for all drives. System Restore has been shouted into submission.

I can run Star Citizen at 72fps in Cities and 90ish in space, 40ish high player activity zones. PC is happy!

Mechanical drives are regularly defragged, and chkdsk'd ... I run a tight ship.




BACKROUND:

So, I have near instantaneous boot from POST. Steam games are slick and in fact my rig outperforms friends with better hardware often (FPS and load times).

I use my "fast" mechanical for uninportant programs like CCleaner and Office, as well as my Documents and Photos because the load times are near instant still and I like to keep my apps off my OS not only for recovery but its neater. And I don't trust SSD's quite yet, probably because none have really failed yet. I'm so stressed out about that ANYWAY....

...my "slow" mechanical has primary games I don't play often / older games as they can load up in 2mins instead of 0.2s while I go make tea. And my porn. Its all midgets and circus clowns for the most part. And 1 confused goat.



PROBLEM:

The issue revolves around the 5400rpm mechanical drive (K:).

And here is where confusing aspect number 1 comes in: It can peak at 2MBps read/10MBps write....and its totally fine. Other drives have marginal 1-15kbps access. The response time though is recorded as being 100% active and 15000ms. Now I have no idea what's a good speed for a 5400rpm drive but that seems ludicrously insane considering the 7200rpm drive never goes over 50ms.

Then, because all Microsoft products secretly have it in for me, it will just decide at around 4MBps write (half its peak) to go completely berserk and very nearly freeze the system. CPU 2-5% and Memory is sitting at 15-20%, but applications will takes forever to respond or even freeze totally into "this app has stopped responding" messages.

Confusing aspect number 2: The access speed plummets in these situations to 999999.


ANALYSIS/THEORIES:

So while the drive is defragged, and has space (this can happen when full or 50%), and yes free space is defragged too, I'm wondering what the hell the Win10 Storage controller/AHCI Controller are doing.

Presumably, they are overloading the bus with the maxed requests to the 1 SATA drive, and the other channels for some reason are not acting independently for the other 4 drives.

While the M2 is on the ASMedia SATA controller, and the 4 SSD/HDD Drives are on the Chipset Controller, as I understand it. So are not in fact physically separated by hardware but fighting for the same procedure calls and responses....?

There is nothing listed for the controllers or anything else scary in the Event Logs, apart from this bloody DCOM timeout on MicrosoftWindows.Client that has appeared I'm pretty sure after I tore out IIS. #FuckIt

Finally, I have recently done a full OS re-install complete tear out (when testing M2 vs SSD), and the raw bog standard zero drivers Win10 had the same issue, and after drivers installed. I did not "tweak" it in any way, however I may have imported my Service settings via scripted install because of old habits...I honestly don't remember. So the Services might have been the same.


QUESTIONS/DESPERATE PLEA FOR HAAALP:

So this has been going on forever, and usually I just cancel the download/copy and do it when im not busy again later. But recently I've had a sudden burst of motivation and have even accidentally on purpose started exercising again having been a Springbok Athlete in my youth, my small human girl child referred to me as a Pork Chop recently and ran off giggling...sooo....

...on the back of inspired activity, I ask my Brethren In Computing...what the hell man?!?
  1. Is it possible to allocate maximum usage of the controller per drive? So limit the k: drive to 50% for example
  2. Is there a way to improve whatever still exists of windows drive caches? Understandably pointless for SSDs, but mechs...?
  3. Have one of the services I've disabled decided to now be special and are slowing the drives requests because they are checking for them, and not being there....?
  4. Is this some other glorious thing I'd never guess, and have not been able to discover as GOOGLE HAS FAILED MEEE! Even the recently impressive ChatGPT quietly slinked off into the shadows...


CLOSING:

Thanks in advance for any actionable aid or insights.
...and no I do not need Windows Update!
 
Last edited:

DSzymborski

Curmudgeon Pursuivant
Moderator
SMART isn't perfect; I'm still inclined to think that there's a brewing hardware issue with the HDD if it's causing these issues while just having data/games on it. I'd be inclined to first swap out the cable, and failing any change in behavior, put it in an external enclosure and see if it the boot behavior changes any.

At that point, failing an improvement, I'd simply replace the drive. HDDs are really cheap and your time is almost certainly more valuable than the drive is, as a slow 2 TB drive.
 

JackrumMadthing

Distinguished
Mar 31, 2014
32
4
18,545
SMART isn't perfect; I'm still inclined to think that there's a brewing hardware issue with the HDD if it's causing these issues while just having data/games on it. I'd be inclined to first swap out the cable, and failing any change in behavior, put it in an external enclosure and see if it the boot behavior changes any.

At that point, failing an improvement, I'd simply replace the drive. HDDs are really cheap and your time is almost certainly more valuable than the drive is, as a slow 2 TB drive.
I swopped the functional drives cables out already ya, no difference.
I forgot to mention that as well as putting it in a near identical rig, my testing server, only difference is no M2 and 1650GT GPU and 64gb 3200ghz RAM....

...benchmarks it flies, I can even leave it doing zip compression tests, essentially burnout. Behaves fine. Its only when I'm doing ANYTHING else, like browsing, gaming, reading docs....anything....that this happens.

Hence my thought process regarding some sort of IO or R/W conflict. Possibly some arcane BIOS setting, but the only ones applicable are the SATA settings and the PCI thingies, timing etc, and those are stock....sooo

...and ya 100% can just unplug the thing, but I dont wanna! I want to knooooow! Teach me ObiWan....you are my only...oh wait you dead. NOOOOOOOOOOOOOO
 

JackrumMadthing

Distinguished
Mar 31, 2014
32
4
18,545
If disconnecting that particular drive solves the issue, then it would seem to be a physical problem with that drive.
..I think through my walls of text you missed the bit where I was saying the problem does not occur in an identical sister PC (same motherboard etc) with that drive plugged in.

So its either BIOS or Software...
I mean it could still be Hardware and it just didnt roll the dice badly, but....my instincts are its Win10 being evil.
 
The issue revolves around the 5400rpm mechanical drive (K:).
And here is where confusing aspect number 1 comes in: It can peak at 2MBps read/10MBps write....and its totally fine. Other drives have marginal 1-15kbps access. The response time though is recorded as being 100% active and 15000ms. Now I have no idea what's a good speed for a 5400rpm drive but that seems ludicrously insane considering the 7200rpm drive never goes over 50ms.
What is model name of your HDD ?
Is it SMR drive?

Can you show smart report for your drives?
Use HDtune health. Upload screenshots to imgur.com and post links
 

JackrumMadthing

Distinguished
Mar 31, 2014
32
4
18,545
What is model name of your HDD ?
Is it SMR drive?

Can you show smart report for your drives?
Use HDtune health. Upload screenshots to imgur.com and post links
WD = Western Digital, all benchmarks, performance tests, SMART records etc are as they should be. There is no hardware fault, and the drive functions without misbehaviour in a second identical system (apart form more memory and a differetn nVidia GPU).

Well I think I've found it.
Its seems to be the dumb built in Windows self optimisation thing, which is an internally scheduled Defrag hidden in sub menus that I've never seen before.

It does not see to make itself known as its a system process that runs under one of the Service Hosts. So in Resource Monitor you wont see it as a System Process going nuts with disc access, for....reasons. As Defrag doesnt peak R/W it peaks access apparently.

My K: Drive reports in the Windows Defrag to be 0% Fragmented, however in the Defrag tool I use it says 0.1 and fixes it to 0 and then windows has a fit.

Wonderfully, my other drives are 17 days untouched and up to 30% fragmented, but it doesnt care about them.

So, it seems when heavy disc access is occurring on the K: drive, for some reason known only to the Powers That Be, this kicks in a defrag. So the system is pausing to allow the r/w in order to not corrupt normal operations.

Anyway, I've torn that babymode collywoble out, and will monitor and report fully here if that was indeed the full issue,

puffpuffpufff