THE PLOT THICKENS: WHS-2011 Question: Odd CPU-usage from the Git-Go!

[BonzaiDuck]

So far, this 2 or 3-year-old WHS rig has served well: Never lost a disk; always seemed responsive; never crashed; managed sleep-states for four other computers; always backed up those workstations with "success."

A week ago, I installed Acronis True Image 2014 -- first and foremost so that I could clone the oversized boot-system disk to an adequately-sized SSD. Then I tuned up my Volume Shadow Copy activity so that it continues to work properly. And then -- I started making backups on a hotswap disk -- a little trial and error -- ultimately with complete success.

There are three add-ins: Axonet Lights-Out (for the sleep-states and scheduled backups); Advanced Admin Console; and Stablebit drive pool.

Always concerned about anti-virus support, I'd switched over from Kaspersky "home-office" version for SOHO server and workstations, and installed ESET Nod32 (or whatever it is) based on recommendations from WHS sites like "We Got Served." Hadn't paid a lot of attention to it: it seems to work.

Late last night, I decided to run an overall system virus scan. It picked up some 16 "threats" which were all installation files for trusted software. For instance, "PrimoPDF" and "PDF Exchange Viewer." I've been using the shareware "free" version of Primo for years; the Viewer is licensed. I dismissed all the "threats."

I noticed my disk-drive LED flashing wildly -- which is occasionally going to happen with a server. Stablebit was doing its "balancing" and cleanup thing. I waited for that activity to end.

Finally, I checked Task Manager, accessing the WHS OS from its desktop -- directly. In the "Performance" window which shows a graphic presentation of CPU usage, one of the C2D cores was pegging 100%, the other looked more like normal activity. I switched to "Processes" to see what was hogging clock cycles.

Windows Trusted Installer service was running here and there up to 10%; True Image Monitor would occasionally jump to 7% or 10%. System Idle process -- barring these other activities -- seemed to be above 90%.

So I had to ask "What's going on with Task Manager's Performance tab?"

Checked to see when the daily backups would start; then -- with an hour to spare -- I rebooted the server. Now the Performance tab of TAsk Manager is showing "normal" activity. Even if the graph shows CPU usage close to 50% or even occasionally 60%, it is evenly divided between the cores.

What was going on with that?

This C2D rig has seemed to be totally adequate -- as I say. But even the normal CPU activity in Task Manager's Performance tab might make me think of finding an i5 quad core and no-frills motherboard to replace the older hardware. This "older hardware" is rocking 4GB of RAM.

Also, can anyone tell me what "True-Image Monitor" does as a service? Why does it need to be running?

 

[BonzaiDuck]

I think, at one time when I put this box together, I may have overlooked putting in the proper "[APIC],[APCI]" driver. This driver would provide information about the processor, the mobo, etc. in matters including such things as power usage.

For some time, the node in my Device Manager tree that would have been identified for this feature-set showed either "Unknown Device" and/or "Disabled." I finally got around to tying up loose ends with this system -- difficult because the drivers must be a match between the old NVidia chipset and the WHS [2008 R2] 2011 OS. It looks as though I actually had FOUND the APIC/APCI driver, but never installed it. then I maybe forgot that I had it, and chose to deal with it by toggling to "Disabled."

After re-enabling the "device," Windows UPdate went out and got the "AMD Cool and Quiet Utility" and installed it as a driver. Now I'm wondering if I should uninstall it and then install the driver I found among my downloads of months/years ago.

I'm wondering if this driver might have affected the operation of Task Manager's "Performance" tab. The "Processes" tab doesn't jive with "Performance," and the "Processes" tab looks realistic in its depiction of clock-cycles and CPU activity on this system.

 

[BonzaiDuck]

I think, at one time when I put this box together, I may have overlooked putting in the proper "[APIC],[APCI]" driver. This driver would provide information about the processor, the mobo, etc. in matters including such things as power usage.

For some time, the node in my Device Manager tree that would have been identified for this feature-set showed either "Unknown Device" and/or "Disabled." I finally got around to tying up loose ends with this system -- difficult because the drivers must be a match between the old NVidia chipset and the WHS [2008 R2] 2011 OS. It looks as though I actually had FOUND the APIC/APCI driver, but never installed it. then I maybe forgot that I had it, and chose to deal with it by toggling to "Disabled."

After re-enabling the "device," Windows UPdate went out and got the "AMD Cool and Quiet Utility" and installed it as a driver. Now I'm wondering if I should uninstall it and then install the driver I found among my downloads of months/years ago.

I'm wondering if this driver might have affected the operation of Task Manager's "Performance" tab. The "Processes" tab doesn't jive with "Performance," and the "Processes" tab looks realistic in its depiction of clock-cycles and CPU activity on this system.

UPDATE: Somehow, it seems that the CPU activity is linked to "Delayed Procedure Calls and Interrupts." Further wisdom discovered on the web suggests the Windows SDK "WPT" or Windows Performance Tools to assess what process or service is hogging clock-cycles. "Processes" DOES JIVE with "Performance" now that I discovered this line-item .. .

The plot thickens .. . I'll either solve this by brute force myself, or someone here may have some insights that help with the brute force.

 

[ViRGE]

Also, can anyone tell me what "True-Image Monitor" does as a service? Why does it need to be running?

It's a component of Acronis True Image. It sounds like you may be running it in Nonstop Backup mode?

 

[BonzaiDuck]

It's a component of Acronis True Image. It sounds like you may be running it in Nonstop Backup mode?

Well, that may be. It's not as though I remember getting a chance to choose, even if there was some arcane opportunity. And despite missing something "arcane" or obscure, I've had a good impression of True Image 2014.

This morning I was thinking of removing Acronis and simply using the Windows Backup features. Then I thought I might troubleshoot what I thought was a "problem:" one core running at near-100%, the other one meandering along at between 3 and 5%. Before my morning errand run for the fam-damn-ily, I ran another web search. It turned up this:

http://blogs.technet.com/b/winserve...cpu-utilization-for-performance-analysis.aspx

This explains it, or seems to -- thus far. There's nothing wrong with my system except that EIST and C1E are enabled. the reason I enabled them: various sleep-states may rely on those features. I think I proved it with my flagship workstation already.

Why would I enable EIST and C1E on a server? I had the server set to sleep ONLY when the flagship workstation was either shut down or also asleep. The Axonet Lights-Out Add-in manages the sleep-states for the entire household. If another workstation comes out of sleep, it fires up the server if the server happens to be asleep. But the server will only go to sleep when the flagship is off-line.

Of course, I also wanted to save power. But if I thought I could find the right compromise, I could simply down-clock the server CPU.

I think the solution could be in more than one option. I could simply turn off the power-saving features in BIOS altogether. Or I could find the right combination that would enable some basic server sleep-state, even if only S1. Of course, the only reason to have the server go to sleep is in an assumption that nobody will be awake at a workstation at 2AM, but that's a bad assumption in this household. I could turn off power-saving features and downclock the processor, but I don't like that idea so much either. The "search-and-tweak" option seems to be the best, and the result -- if reliable -- won't be needed too often.

In the process of getting to this level of awareness and perception about what we do with a basic Win 2008 R2 OS running on an old workstation or desktop board with a single (C2D) processor, I installed the SDK and Windows Performance Tool(s) to the WHS-2011. Now, I find that "WHS is not designed to allow these programs to run," and I need to find a way to use those performance tools -- obviously designed for Win 2008 and R2 -- therefore -- WHS 2011.

I can poke around and figure it out myself, but somebody with easy keyboard fingers and more experience or insight could save me some trouble.

And while on that topic, I'm wondering what happens if you let the disk drives spin down in the "Power Management" settings. If they spin up reliably without a hitch, hopefully you'd only wait longer trying to get access through a workstation. At this point, I don't know for sure.

But when it comes to power consumption, four disks or more running 24/7 is definitely "food for thought."

Now I'm also looking at the latest issue of Maximum PC. "Rest in Peace: WHS-2011" appears in the pages comparing various NAS boxes. The best of the lot costs $1,000. You'd want to get a lot of leverage for that $1,000 over five years or longer. In a business, it pays for itself. But we're not a business -- we're retired. I know a retired person who has a huge model train setup -- trains and engines in nick-nack selves over the walls, tracks on multi-level tables spanning the room, his command-post with all the controls at his fingertips. I might have some things to say about computerizing his hobby -- that would be truly interesting -- lie Denzel Washington or Walter Matthau "Taking of Pelham 123." My version of my friend's hobby: I AM my OWN mini-NSA network administrator. So I might want to wait on the NAS box.

Tell ya what. I'll think about it, before it comes up for staff review next year.

 

[ViRGE]

Well, that may be. It's not as though I remember getting a chance to choose, even if there was some arcane opportunity. And despite missing something "arcane" or obscure, I've had a good impression of True Image 2014.

This morning I was thinking of removing Acronis and simply using the Windows Backup features. Then I thought I might troubleshoot what I thought was a "problem:" one core running at near-100%, the other one meandering along at between 3 and 5%. Before my morning errand run for the fam-damn-ily, I ran another web search. It turned up this:

http://blogs.technet.com/b/winserve...cpu-utilization-for-performance-analysis.aspx

This explains it, or seems to -- thus far. There's nothing wrong with my system except that EIST and C1E are enabled.

I still suspect you have something wrong. 100% CPU time would normally kick up the processor to its full clockspeed, and unless you've ruled it out that may very well be happening here.

 

[BonzaiDuck]

I still suspect you have something wrong. 100% CPU time would normally kick up the processor to its full clockspeed, and unless you've ruled it out that may very well be happening here.

Ha-hahhh! Not laughin' at you, Virge!! Often -- I read something and it's a dead-end path. But sometimes -- I get it right! Take a closer look at the article I linked.

Finally got home after getting the car tweaked to pass smog. No backups underway; server is idle; same thing -- "100%" on one core. Not perfect, only paraphrase: "Using the Task Manager Performance tab to diagnose load and power-state problems . . . not a wise idea . . . " Then it goes on to explain how the power-saving features would lead you on a chicken-little wild goose chase.

Anyway, for some reason, and apparently known to the "server gurus" who wrote the article, it might not kick up the cpu-to-bus ratio and speed if the idle-state core is enough to allow the services to run. Or -- it may be that a server OS doesn't adequately handle the power-saving features. At that point, I can't say. But the article certainly DID mention something about running server OS versions on processors, motherboards and configurations meant for workstation user profiles.

Nothing wrong inherently with doing that -- they call it "Home Server" fur-chrissake. We've been doing it for decades. I had Win Server 2000 running on an Intel Tucson board with a Pentium or Pentium II. If it was meant for "something else," we were able to tweak it for what it wasn't.

But what I CAN say is this. I rebooted the server box, got into BIOS, disabled the EIST and C1E. Booted back into WHS-2011; logged on to the console.

Now these cores are both mostly ambling between 0% and 4%, both of them just ticking along. So I'll incline to the notion that there's nothing wrong with my server installation, and that I need to explore further the limitations without power-saving states and how it bears on sleep states.

Frankly, if it were possible to stably spin down the drives after so many hours of inactivity so that they spin back up on demand, the CPU and RAM are just some tiny light-bulbs running 24/7. Otherwise, I'm moving toward a configuration with fewer drives, so .. .

 

[BonzaiDuck]

Sorry about the double post, but I looked at it once and made my previous post.

Looked at it 20 minutes later and it's doing the same thing again.

So then, the issue still remains. Should it be doing this? Or should it not? And if it is not supposed to do it, then what is causing it to happen?

Very strange. But I also have to weigh it together with the information in that article.

Somebody must know, though.

I could always simply back up my software downloads and licenses for WHS-11 and then reinstall the OS. That's about the simplest thing in the world. Less of a PITA, but still a PITA. Problem with that: I'd have to inventory our "users" and their folders, and re-establish all the security allows and disallows.

 

[BonzaiDuck]

Given the focus in the first link I posted pertaining more to Win 2008 R2, this "problem" supposedly has a history going back to Beta WHS "Vail."

http://social.microsoft.com/Forums/...all-the-time-is-this-normal?forum=whssoftware

Somebody here at the forums might know something more than what I've found out so far.

It doesn't seem to impair performance of the server in its functions on this LAN, and it doesn't seem to impair any work or interaction with the WHS desktop directly. Further, the hardware monitoring of the two cores (E6600) shows equal temperatures characteristic of idle. Lemme check that again . . .

Yeah! Core 0: 30C; Core 1: 29C.

You would think Core 0 would show significantly higher temperatures for CPU usage on that core bouncing around between 90 and 100%.

Before I "do" anything more (like adding SDK WPT to SW installs, or uninstalling Acronis, or reverting back to v.1.x of StableBit) -- I'll keep looking for either an explanation of this, a solid opinion that "something is wrong," and with that -- a way to fix it without rebuilding the entire server, "users," "machines (for backup)", and security restrictions.

The fact that this cropped up with the 32-bit Vail version Beta -- suspicions that it linked to Drive Extender -- suggests to me that the phenomenon is known. The article referencing Win 2008 R2 only suggests it might not be a problem, or it might be a more general problem with OS's like Win 2008 R2.

I think I'll start by reinstalling all the hardware drivers from scratch. I might start then removing some software. But I want to move slowly on this: it might not be a "real" problem.

 

[BonzaiDuck]

Found this sevenforums post this morning:

http://www.sevenforums.com/performa...-sleep-causing-i5-2500k-core-0-pin-100-a.html

 

[BonzaiDuck]

Anyone interested -- if you use WHS-2011, Win 2008 R2 or possibly even more recent -- here's something to take away from my "big crisis of CPU % usage."

This box was a slight misfortune born of short-sighted convenience. Of two LGA-775 systems, I might have picked the one with my Gigabyte P45 (Intel) motherboard. It was easier for availability and less trouble with backups to leave my WHS v.1 VAIL system running and online while I installed WHS-2011 on a *&@!!% M__F__g!! ASUS [NVidia] 680i motherboard.

There was the problem of finding the right drivers for a variant of Win 2008 R2. Rule of thumb: Find the Win 2008 drivers; no 2008 drivers? -- use drivers for Win 7-64. After that, you might get lucky, but more likely be S-O-L. So I was somehow able to avoid the "SOL" ending here.

That being said, I had these *&@!!% M__F__g!! nFarce drivers for the onboard gigabit NIC ports -- throwing up red-bangs in the event logs, even though they appeared to work properly. Disabled this; killed that; replaced with an Intel Pro CT Gigabit Desktop PCI-E card. The "advice" suggested the Intel drivers to the "Pro" card were bundled into the OS from the git-go.

And -- they were. But who ever fully trusted M$ WHQL drivers downloaded with Win Update? I tell my family: "You see any hardware drivers in your pending updates -- uncheck the boxes!"

Under this regime of the 680i board, old hardware, limited drivers -- we may lower the bar, drop our standards, accept with feverish gratitude -- any drivers that Windows offers up (or throws up -- regurgitates.)

SO -- THE PROBLEM WITH THE CPU USAGE: IT WAS THE FREAKING INTEL NIC DRIVER!! The Win Update version was dated 2009; the latest downloaded direct from Intel -- 2013 or 2014. Whatever.

I also disabled "JUmbo packets" for the card until I can assure that the other client machines are configured for same.

So now I get to uninstall the SDK, WPT and other crap I thought I needed to diagnose this, although they aren't really pushing the SSD boot drive toward a "drive full" crash or difficulty.

I WOULD SO HAVE HATED TO GO THROUGH THE DRIVER INSTALLS FOR THE MORE DIFFICULT AND OLD NVIDIA HARDWARE!! But -- We're home free!! Now I can add some bigger drives to this sucker, and squeeze a few more years out of it as a "HOME SERVER." NAS-BE-DAMNED. IF I get one, I may just use it for a backup box for the server-box. Whatevah whatevah.

I could end this with a quote from Russell Crowe in "Gladiator:" "Are you entertained?!! Are you entertained!?"

At least I didn't have to "slay" my current WHS-2011 installation.

 

[BonzaiDuck]

I thought I waited long enough to show both cores operating at between 1% and 5%.

Posted on another thread here -- with my workstation -- and switched back to the server desktop 20 minutes later.

It's back the way it was!! between 80 and 95% "usage" on core 0, 4 to 5% on core 1.

Nothing seems particularly stressed. Now I could try replacing the "other" drivers. Maybe . . .

 

[RampantAndroid]

In past, I've used Win8 drivers on the latest server builds and been OK; as for CPU usage, have you tried process explorer? What is using CPU time with the new driver?

 

[BonzaiDuck]

In past, I've used Win8 drivers on the latest server builds and been OK; as for CPU usage, have you tried process explorer? What is using CPU time with the new driver?

I have to fold your suggestion into my "plan of attack." What I'm experiencing here is a far cry from similar complaints -- about server or client systems and OS's. One guy complained about a burnt-out CPU!!

In Task Manager, the overall CPU usage bounces between 45 and 50%. Core 0 goes between maybe 80 and 90% after an initial period of normal 4 to 5% performance maybe 10, 20 minutes following boot-time.

I got in touch with the Cove-Cube support tech. He agrees: "The CPU usage may reflect a problem, and then again maybe not." But all this started after I installed the new StableBit 2.x. Problem is -- I'd also installed True Image 2014 and an update to another WHS add-in. I can continue to uninstall and reinstall drivers -- I can't much make things worse, and they're not seeming all that bad at the moment.

As part of my "crash program" I've purged the sources of red-bangs and yellow-bangs in the event logs -- only a few benign yellows remaining on the server, and a couple benign "reds" on my rock-solid workstation.

He's urging me to try the Beta to Stablebit, and his remarks insinuate a possibility that it might fix it if Stablebit was a source of the problem.

What I'd noticed: Even if I turned off Stablebit services, the high-usage core continued on its established path. But I also noticed that the graph for Core 0 had a "pulsating" dynamic to it, and I could see either the minor pattern in the "process" tab for the Stablebit components pulsing along at the same frequency.

Between someone's posts somewhere else and indications I had firsthand, it seems that the usage is linked to "delayed procedure calls and interrupts."

I'm going to disable all the old hardware on the system that WHS doesn't need, and purge the drivers for any of those items. Then I'm going to methodically reinstall the drivers for all the needed hardware one at a time.

But with the "true-blue" System and Application logs inspiring confidence, I may try the Beta version first. Gotta think about it. . . . .

I can say this, though. I may have been putting off "office" work too long for this IT stuff, but things are getting better across the board. And I don't want to screw the pooch with some radical action taken in panic. So far, I don't see any need to panic about any of it. The server still works -- otherwise -- and with that degree of CPU usage on Core 0, its temperature doesn't much clear 30C. The disk drives all seem healthy; the software doesn't malfunction.

We're going to solve this!!

Offhand -- does anyone readily know how I can get the SDK WPT components to work without compromising security? I get message boxes telling me that the OS doesn't allow use of those programs. But I had carefully chosen to download the MS versions for Win 2008 R2.

 

[VirtualLarry]

Have you enabled "Show Kernel times" in Task Manager's Performance tab?

 

[BonzaiDuck]

Have you enabled "Show Kernel times" in Task Manager's Performance tab?

Thanks, Bro. I'm going to do that tonight or soon enough today so's not to forget about it.

[didn't put it off . . ] The "red" -- which I assume is "kernel times" or clock-cycle consumption -- is just a shade below the total. I opened resource monitor and it reminded me how I came to the "DPC's and interrupts" perception. That's where it shows up.

Incidentally -- I can post this here or in a new thread.

Does anyone have a recommendation for a hot-swap disk mounting utility that works with that family (Win 7, 2008 R2, WHS 2011) of OS's? I had scoured the web for something that replaces "Swap-Man" -- which came from StarTech with IDE bays and caddies. Latter worked in Vail, but not in v. 2011. I found a utility authored by some "Yamaguchi Sarazawa" -- obviously a Japanese programmer -- don't really remember the name. It was "unsigned" but seemed to be reasonably popular and current. Only problem -- it's configured to keep asking me to start it at boot up. While I'm just cancelling it now, I'm going to suppress it.

And I might want to replace it. Maybe there's even a feature in WHS, but I'd never found it. With enough effort, I might have. But then -- maybe there isn't such a feature. Whether some plug-in program or 2008 R2-compatible utility, or a built-in feature, maybe somebody in the forums knows of something . . .

As to "extra programs," I'm going to uninstall Crystal-Disk-Info and buy Cove-Cube's "Scanner" program, which does the same thing and likely better. For the kind of advice I got from their tech-rep, the $25 is chump-change. A lot better than I had from some other tech-support communications . . .

 

[BonzaiDuck]

I started taking stabs at "hardware unneeded," BIOS disables and drivers.

Right away -- went for first thing you might think of a server you wanted rock-solid for basic tasks, and likely the driver I had most doubts about as simple and peripheral as it might seem .. .

I've been logged on directly to the Windows Desktop now for 40 minutes. I KVM'd over to my workstation and opened some folders on the server in Win Explorer. Then I went back to the server and opened Dashboard. Then I ran "re-measure" under StableBit and watched the TM performance tab. After Stablebit finished, it was back to showing below 10% for both cores . . .

I'll reveal the Mystery Offending Hardware & Driver after I've watched it for another hour or two . . . But let me lay the groundwork.

Once I was driving through Tennessee at 5AM on the way to No. Va. to deal with property and other matters. I stopped at a Waffle House, like the ones shown in the Patrick Swayze movie "Black Dog." The counter was crowded with truck drivers, farmers, businessmen -- travelers.

A 50-something waitress -- rotund and jolly -- stood in front of me with a checkered-pink apron and a big ladle in one hand. She had wire-rim glasses, a big smile and rosy cheeks.

Looking right at me, she says "Do ya'll know the difference between a Northun fairy-tale and a Southun fairy tale?"

"No -- what?"

"In the Northun fairy tale, it begins 'Once upon a time. . . ' The Southun fairy tale begins with 'D'you ba-leeve dis shee-it?!'"

I'll be back soon . . .

 

[BonzaiDuck]

http://www.analog.com/en/content/soundmax_audio_signal_processing/fca.html

And the hardware:

http://www.techexcess.net/ProductInfo.aspx?productid=SUPREME-FX

The Path of Progress:

BIOS->Onboard Devices->HD Audio->"Disable"

Uninstall "SoundMax" in Programs & Features.

Task Manager has shown < 10% on either core now for an hour and 45 minutes, after going through the paces I mentioned .. .

"D'you ba-leeve dis Shee . . . ?!"

 

[Raduque]

I believe it. I never even installed drivers for the audio on my WHSv1. Servers don't need audio.

 

[BonzaiDuck]

I believe it. I never even installed drivers for the audio on my WHSv1. Servers don't need audio.

Ah, Raduque. I'm like the boy who cried wolf! Every time I thought I had it solved, it wasn't solved. Wasn't the audio -- even with more questionable drivers. Wasn't the drive controller. wasn't my SSD. Wasn't the add-ins like Stablebit or Axonet Lights-Out.

The server and the client workstations must synch in time within a margin of +/- six minutes. Otherwise, this stuff gets backed up at the server -- thus Delayed Procedure Calls and Interrupts hogging 47% of clock cycles on one core.

Set all the times from my cellphone as I go from desktop to desktop in the house.

I went through a freaking hardware panic, bricked two WHS workable OS's on disk, pulled the sound card, uninstalled add-ins -- finally reinstalled the entire WHS OS -- with all my data saved by the StableBit add-in. And still -- it happened. Then a little more web-research and reading. All that trouble over a seven minute difference between system times.

And this can't be "Help! Wolf!" Last boot time was around noon, and both cores are ticking along now at 0% to 1% four hours later.

UPDATE -- two hours later. Well, shucks. "Wolf!!" Resetting all the clocks made it behave normally for four hours -- so far a record as I try to find out why it does this. But once again, six hours after reboot, it's loading up Core 0 again.

This system is stable. Never really crashed. Mobo was barely ever used. I don't think the processor was ever abused. I have a spare Q6600 I can replace it with. the only hardware left in the system for which something might be wacky is the motherboard. There's just no indication of anything there specific.

 

[BonzaiDuck]

I found a utility most useful for analyzing this sort of thing: "LatencyMon" or Latency Monitor. Things become apparent rather quickly.

OEM WHS boxes may not even come bundled with a graphics card: "You should use the Dashboard from a client" all the "user-guides" say. But you NEED to access the desktop for any server OS. You either get a graphics card and KVM switch, or use Remote Desktop -- or both.

What else pops up when you install the OS? "Optimize for programs or background processes." You naturally choose the latter, because it's a "server."

Going through the available sources, I find the statement "Windows SERVER defaults to 'optimize for programs'." What's that?!! "De-faults?!" Then there's more. The wisdom begins to explore high server access in numbers of logged-on accounts -- BizTalk Servers, DBMS database servers.

But your HOME SERVER doesn't have 1,000 BizTalk accounts. It might in a blue moon have some disk access from the fam-damn-ily in the "Home" (it's called "Home Server"). Even the Octo-Mom might (in a blue moon) show less than 15 accounts logged in and accessing data.

You're not using a dual-processor quad core Xeon, either. You might be using anything from an "Atom" to your cast-off Conroe or Kentsfield -- like me.

So it seems both "counterintuitive" and -- well -- "intuitive." You're not running a server-farm; you're not expecting deadline log-ons as if you're hosting the Affordable Care web-site. And Win Server (the article was speaking of Win 2008 R2) -- DE-FAULTS to "optimize for programs."

 

[pw257008]

this is a very good thread. i've never run WHS but this all has me intrigued.

that's all.

 

[BonzaiDuck]

this is a very good thread. i've never run WHS but this all has me intrigued.

that's all.

Well, it's an interesting scenario.

The technology for this server hardware is approximately six years old, and it is orphaned. Orphaned. If there is anything that keeps it alive, it may be motherboards for the AMD CPUs with nForce chipsets. But this board preceded the great NVidia-Intel feud.

Meanwhile, OS's either have drivers for a certain generation and make of hardware, or they don't. Windows still supplies some generic WHQL drivers for the NVidia chipsets for Intel CPUs. But an analysis of forum posts on trouble-shooting web-sites reveals something about it.

It occurs for Windows 7 systems as well as Win Home Server 2011. There are also indications for Windows 2008 R2, but less in the way of an intersection with NVidia hardware. The latency problem points to two buggy drivers -- both by Microsoft: NDIS -- sometimes hand in hand with the TCP/IP service; and iastor -- a Windows storage driver. The issue also crosses paths of NVidia's casual accommodation to AHCI during the years the 680i and possibly 780i chipsets were released. Instead, the nForce SATA solution has an "almost AHCI" approach called "IDE-SATA." You either get it that way, or set up the BIOS for RAID. But NVidia RAID on this chipset doesn't make it easy to configure individual hard drives as "RAID" unless it is chosen as "JBOD" in the RAID controller BIOS.

There are also cases affecting laptops as well as desktops, and even Windows 8. Through all these variations, two words keep popping up: NVidia, and NForce.

I replaced the Conroe C2D with the Q6600 C2Q. It took five hours, but the loading of Core0 surfaced again at that time after boot-up. Even so, the system now seems twice as responsive. Task Manager shows "25%" overall CPU usage, but that's all in Core0.

In some other thread I had mentioned this disk controller which features an alternative to Intel ISRT: a Marvell chipset with "Hyper-Duo" or "Hybrid Duo":

http://www.newegg.com/Product/Produc...82E16815158365

The controller is compatible with Win 7, 8, 2008 R2 (and WHS-2011) for hosting bootable system disks. There is no software driver: It's all done I the card's BIOS, recognized by Windows. Generally, in a PCI-E v.1.x slot, it should give the SATA-II maximum throughput spec. And since the HDDs can't reach that speed anyway, there isn't much of a loss. In fact, the OS disk could be paired with a 60GB (even more!) SSD. The remaining ports can host port-multipliers, so one card might handle seven disks.

What is needed is something that meets a balancing act. I'm not eager to pay for "Server Essentials" or something similar, and I like the way WHS works for my Win 7 systems. Maximum PC had posted an article touting Windows 8 as a "server OS" replacement for WHS, but I'm not eager to do that now. I should be able to leverage WHS 2011 for a while.

If I can shut down the SATA controller on this motherboard completely, it removes the nForce drivers from the equation. That may offer a chance that NDIS will behave properly.

While this hardware was released in ~2006, the board has effectively only three years of wear and tear. The C2Q I obtained looks like brand-new OEM-wrap.

If I were to buy, say, a Haswell system, I might find troubles with the WHS OS running on Haswell. But WHS 2011 was released about the same time that socket-1155 and Ivy Bridge was being sold. I can either make the NVidia-based system work for me with WHS, or I can spend maybe $230+ for Z77 mobo + IB i3 CPU, and there shouldn't be any issues with the WHS that way.

So first step is to try out this Startech SATA card, for which I had other plans as well. Move the 4-drive 4TB drive pool to a 2-drive/4-TB drivepool with possibility of expansion.

 

[BonzaiDuck]

So. After all this troubleshooting over an old NVidia chipset and ASUS motherboard, what was the mystery about Core0 loading up all the time with the machine at near-idle?

I had a "theory" after seeing other forum posts around the web from people with similar problems fitting a pattern. There seemed to be "Delayed Procedure Calls" and "Interrupts" responsible for the loading, and diagnostics showed a common group of offending drivers: NDIS -- the networking driver; MSAHCI -- the Microsoft built-in SATA driver; and drivers like nVstor -- the NVidia SATA driver.

The "theory" suggested to me that certain things might "happen" if I moved my server hard disks to a different controller and perhaps even shut down or disabled the onboard nFarce controller.

Well -- I put in a new controller card: a little PCI-E SATA/RAID controller made by StarTech featuring a Marvell controller chip. The card meets the SATA-III spec, although bandwidth is limited if the PCI-E 4x slot that it needed only meets the PCI_E version 1.x standard. Several similar cards -- all with the Marvell chip -- are offered by Rosewill, ASUS and other companies, some as inexpensive as $40. And one sage advisor explained how easy it would be to move the boot-system drive from the nFarce to the Marvell without so much as a blink.

It appears that just putting the new controller in an appropriate PCI_E slot forces Windows (and WHS '11) to recognize the new controller chip, reporting that the native MS driver is "not successfully installed." But once the StarTech's driver disc has been inserted and once the driver has been updated, no disk migration seems urgently necessary.

Suddenly, there is no more loading for core0. I say this now, after the system has been running for a full 12 hours. It was certainly unnecessary to replace the Core-2-Duo Conroe E6600 processor as I did with a Q6600, obtained from a reseller for something less than $90.

However, I now have a server that is much more snappy and responsive. And "all is well." And I haven't even moved the boot disk to the new controller.

 

[pw257008]

:thumbsup:

this is the kind of resolution i like to hear about