Several years ago I stopped building desktops and instead bought desktop replacement laptops. This transitioned my behavior away from doing cold boots into instead leveraging Sleep and Hibernate. So when I later returned to building desktop machines I was looking for that same level of flexibility. My first new desktop PC since the Athlon XP/nForce2 days was a Nehalem (Bloomfield) system and while that upgrade from a performance and stability perspective has been immensely rewarding, the platform has some quirks with sleep.
The inability to use the PC the way I wanted chafed against me. The Tech Report (TR) poll on the subject implied that my use of sleep and hibernate is not the norm. The vast majority of x58 users most likely never saw sleep problem as they either shut down or never cut their PC off. I’ve patiently endured this quirk for two and a half years now and I’m more than ready to make the leap over to SB-E / x79.
I had originally planned to wait until Asus released the Rampage IV GENE in the hopes it might be an x79 mATX solution with more than four DIMM slots. Unfortunately my local Microcenter made me an offer I couldn’t refuse. This meant a retirement of my Silverstone Temjin TJ08-E, a case I’ve become so enamored with I’m holding onto it for the future. Additionally my VM work had lead me to max out the capacity of my six DIMM x58 board, so a four DIMM x79 solution was unacceptable.
Rant: Why is any board manufacturer making a four DIMM x79 board? The chipset is meant to be a workstation platform, it is meant to be the high end, and the platform is supposed to fail the value argument for most users. Yet, the vendors are going to ship board with four instead of the max eight slots? Whisky Tango Foxtrot.
I would have loved to stay with Silverstone cases, like for example the Raven RV02-EB, but since that case has its own gravitational field I opted to look a different direction. Microcenter carries the Obsidian, Graphite, and Carbide series and the Carbide 500R White had caught my attention. So 7pm EST on Black Friday I’m rolling into the local Microcenter and I’m picking up the 500R, the Asus P9X79 Deluxe, one Intel 3930K, and the free Corsair H80.
I had two goals in mind for this migration:
- Convert boot disk to GPT.
- Migrate my applications, profiles, and OS over from the x58 to x79.
- I had to build my Windows Preinstallation Environment (WindowsPE/WinPE) thumb drive with ImageX.
- I used Action Center to send and purge all crash information on both the Admin and Standard account.
- I used Disk Cleanup to purge junk information from both the System and the Standard profile.
- *I emptied the Temp directories on the System and under both the Admin and Standard profile.
- I purged all IE9 information (cache, history, cookies, forms, etc.) with built-in deletion tool.
*The image provided shows only the location of the Temp directories, not an expedited way to purge the temp directories.
With the junk data cleared it was time to remove any driver packages that used an installer and uninstaller. For example:
Reboots are often required after each driver uninstall to aid the system in purging active files.
With the installer based driver packages cleared it was time to sysprep the system. This process will turn my existing install into a clean gold/master image that could be deployed to other machines.
Fire up an Admin instance of command prompt:
cd x:\windows\system32\sysprep
sysprep /generalize /oobe /shutdown
The system will be purged of all machine specific information, rolled back to the Out of Box Experience state, and will initiate a shutdown of the current session upon finishing its cleaning tasks. With the sysprep complete and the machine powered down all that remains is to capture the image.
Powering the up the x58 and booting from the Windows PE thumb drive delivers me into the WinPE command line interface (cli). I used the diskpart command to enter into the diskpart interpreter. Using the diskpart commands of list disk, select disk, and detail disk I am able to determine the default drive letters assigned to each storage device in the system.
With this knowledge in hand I have everything needed to use the ImageX tool. This encompasses three major items:
- The drive letter of the bootable thumb drive, which houses imagex.exe.
- The drive letter of my source partition.
- The letter of my destination partition.
H:\imagex /capture E: F:\x58.wim “Nehalem” /verify
Notes:
- To learn more about how to build a WinPE boot disk with ImageX see this video and/or this TR discussion.
- The order of drive letters in WinPE may not match your expectations. The fact that hardware controllers play a bit fast and loose doesn’t help. The built in 100MB boot partition used by Windows 7 also creates some drive letter differences.
- The imaging process creates a compressed file based image of the system and is smart enough to drop certain files from the capture. For example there was little reason to capture my 25GB Pagefile or 19GB Hiberfil and make that part of the image. The OS will be able to rebuild those dynamically.
- In this case the transfer was from a 2TB WD Black to an eSATA connected 1TB WD Black. The 44GB image took about an hour to create and I potentially could have shaved some more time off of that had I been willing to leverage my SSD as the temp file source for ImageX.
The tear down of the TJ08-E didn’t take very long, but the build of the actual x79 system on the other hand I suspect may have consumed nearly eight hours of my time. Why so long? Perfectionism and because neither the Carbide 500R or the H80 come with real manuals.
The lack of documentation left me little choice except to search and read for forum discussions, YouTube videos, and website reviews in an effort to figure out how certain things work. I still feel like an idiot regarding the top of this Carbide case, which is removable to aid in installation of a radiator like the Corsair H100 or to mount 120/140mm fans. I couldn’t find any discussions on how to remove the top metal panel, so I just pushed the top panel up and out from below. During reassembly I discovered all I had to do was push down on the front part of the mesh and the spring mechanism would release it. Grr.
People reason in different ways. Making an assumption that your invention is clever enough to be understood by everyone is bad idea. This is why we as species invented the manual and chant its praises with RTFM.
Getting the bracket off the side of the “tool less” 3.5” drive bays to mount a 2.5” drive was a displeasing experience. With the lack of a manual with images to confirm my assumption based on the design of the side railing I was left to presume I was right and then leverage entirely too much force to snap it free of the cage. My assumption was correct, but my general view of things is if I have apply force I’m doing it wrong. I know there are many devices and mechanisms in life that expect and require a forceful transition, but it helps when there is information making that clear.
The installation of the H80 also failed to clarify a few key points. Both the Molex and the 3-pin header for the block must be plugged in. I initially believed it was one or the other, but after having difficulty with this configuration I realized that the 3-pin header only had one line and no ground: it couldn’t be the source of power. It was a clear detail I shouldn’t have missed, but I discovered I wasn’t alone in this failing. My Google searches for a more concise answer found others similarly confused by this arrangement. Equally perplexing is that both of the 120mm fans were supposed to be directly connected to the water block. I had initially expected that they would need to be connected to their own headers on the motherboard. I had assumed the series of pins on the block for the fans were for the Corsair Link add-on that allows more dynamic control and monitoring. There also wasn’t a clear indication that the object in the center of the water block was a button to control the behavior of the pump.
Mounting the fans to the radiator was easy enough to understand at least after I figured out that it was in fact intended to mount to the back of the case. While the process of assembly is not difficult it is precarious and it is large enough to obstruct memory channels A and B making RAM installation difficult. A situation that during assembly caused me to utter a phrase that tempted fate.
The LGA2011 bracket also introduced new challenges. Much like Scott/Damage I also found the dual-lever retention mechanism rather unique and at least from my perspective unintuitive. Luckily for an idiot like me Asus has a fantastic manual. :P Meanwhile Corsair and the H80 were once again falling down on the job.
The pictures provided with the H80 implied that an X clamp was needed on the back of the motherboard. From there the mounting screws would thread into that clamp and represent the base for block. That’s great information, except it’s not true for LGA2011. The mounting point for LGA2011 are included on the socket already. As you scroll through the images I’ve provided with this post you’ll see that the LGA2011 socket already has an X clamp shape and that there are threading points on it. Despite this Asus decided to confuse the situation, by still having the classic mounting holes exist on their motherboard. To prevent their use a material exists between the layers of the PCB to close them off. This material almost sounds like tape as it crinkles when touched with a plastic spudger and is soft enough to leave impressions: such as the tiny point of your spudger. On the back of the board one of the four traditional mounting points through the board has been obstructed with a sticker to stop the installation of an X clamp. You can of course remove that sticker, but then the layer inside the PCB gets in your way. Maybe you can even tear through that the layer if you were bound and determined to use those mounting holes. Feel free to post a follow up to this point and let me know how that works out for ya.
The H80 includes LGA2011 mounting parts, but again due to lack of a manual provides no information on how to actually mount the block to the CPU. The tiny little pictures only getting started guide is out of date and still only shows mounting through those now sealed holes. I sorted this problem out on my own after realizing the mounting points were sealed, but not before removing the obstruction sticker I referenced previously.
With the mounting points for the H80 installed and the pre-applied grease removed from the block it was time to apply my own thermal compound. I used IC Diamond, which recommends applying the paste in the amount and shape of a pea. Given the size of this CPU I was a bit concerned I might not have used enough paste, but successful temperature feedback implies otherwise.
Notes:
- Expect to use a bit of force to get the block over the threads.
One of the biggest things I took away from this mounting process is it would have been a mistake to pay for the H80 with the Carbide 500R. The H100 is a superior solution to mount and would represent less impact and obstruction to reaching and working with the motherboard in the case. I meanwhile am left to complain against free, which is hard.
The final task was dealing with the cable routing, which is always an involved process. Looking at the pictures I feel as if it could be better, but given all the things I have to connect I’m not sure how! Long term I want to change the fans over to some Silverstone Air Penetrators and buy some modular cables for my power supply that are individually sleeved.
You’ll notice that the hard drives are a bit spaced. One of the advantages to a larger case like the RV02-EB or the Obsidian 800D is you gain access to a backplane for drive mounting, which makes the cabling far easier. In a more traditional setup – like in the mid-tower Carbide – you have to route those power cables the old fashion way. The Silverstone Strider Plus 1000W has single modular cables with three SATA connectors on them, but the spacing per connectors leaves a bit to be desired. When connected to sequential drives I perceived that the SATA power connector was under undue stress when bending the loop of the cable left or right. Using an every other drive bay pattern created a better flow for the cable and connectors, which in turn aided in tying down the cables to allow the side panel to close.
With the system installed, cabled, and routed it was time to plug it in and watch an electrical short ignite a fire and send the thing up in smoke.
Luckily the only hardware difficulty I had to deal with was that two of the RAM modules behind the radiator weren’t installed properly. Remember that tempting fate bit earlier? Yeah.
What made this particular bit of troubleshooting especially difficult is that I’m an idiot. Instead of getting a no memory POST code the UEFI firmware instead was able to boot up on what RAM it could address. Inside the UEFI interface it was able to tell me exactly which DIMM(s) were not cooperating. In this case the DIMMs in channel B1 and B2 were visible (UEFI could read the SPD), but had tossed some sort of error due to not being firmly installed.
This should have only been a moderately difficult fix. Unfortunately I made the mistake of getting the channels on the board mixed up. I thought channel A and B were on the east side of the board. So I spent entirely too much time troubleshooting, swapping DIMMs, even upgrading the UEFI firmware only to eventually discover I’m messing with the wrong memory slots.
So after figuring out that the UEFI firmware update hadn’t in fact bricked the board - it needed a hardware CMOS reset - I set to work on removing the radiator obstructing my access to the DIMMs. Once I had arrived at the real A and B channels of the board I was able to reseat the DIMMs and get the board up and running with the full 24GB of RAM. I also made one other subtle tweak, but mostly for purpose of consistency. All six modules of RAM are Corsair Vengeance 1866, but two of the six are different versions of the same DIMM (v3.19 vs. v5.11). I decided to group the v3.19 in channel A1, C1 and move the four v5.11 DIMMS into B1, B2, D1, D2.
Solving the hardware problem moved me to my next hurdle. As I said at the beginning of this odyssey, one of my goals for this install were to move the boot disk over to GPT. As it turns out making Windows boot from a GPT disk is not simply a matter of having UEFI instead of a BIOS. Oh, no, it required that BCDBoot be leveraged inside of what Microsoft calls UEFI mode. That means that my WinPE thumb drive, which had been built only to boot in BIOS mode, could not be used to create the new BCD boot structure on the disk.
This was especially troublesome as the Windows Automated Installation Kit (WAIK) tools I needed to fix the thumb drive were not on my laptop, but instead inside Windows Imaging Format (WIM) captured from the x58. It was a chicken and egg sort of problem.
Eventually I realized I decided to approach this situation in a distinctly less subtle approach. I booted from the installation DVD in UEFI mode, had it partition and format the boot disk and then install a bare installation of the OS. I then booted from my BIOS mode thumb drive into WinPE and used diskpart to format just the data portion of the drive and then used ImageX to extract my image to the disk. Since I had not touched the boot sector, I should then be able to reset and having a work installation leveraging winload.efi instead of the winload.exe.
Success!
Notes:
- imagex /apply d:\x58.wim “Nehalem” F: /verify – an example of the command I used to apply the gold image
- To learn how to create a UEFI bootable WinPE see here.
- For a general overview of how UEFI boot works see here.
- Those advanced enough to be using Windows System Image Manager (SIM) see here for some UEFI help.
- It seems that UEFI mode requires the 100MB boot partition to be FAT32. In BIOS mode NTFS is acceptable.
With the deployed image now booting Windows transitions into its installation phase. Hardware detection begins, services are installed and initialized, machine SIDs are generated, and any other machine specific info that was wiped by sysprep is created. Machine specific settings that are variable in nature (such as locale, date and time, etc.) are presented after the pre-boot initialization has completed (usually involved two resets) and with your arrival at the Out of Box Experience (OoBE).
Notes:
- See this video to watch this process unfold.
Now it’s time to deal with the drivers. My to do list of drivers to download included the NVIDIA GTX 590, the USB 3.0 controller from ASMedia, a management sub-function of the Marvell 6G SATA controller, the Intel GigE NIC, X-Rite i1 Display Pro, and a stupendous number of x79 chipset functions.
As a bit of serendipity I had used Windows Update to update the drivers for the Realtek GigE NIC on the x58. Windows Update adds those driver to the driver store and the latest version of the WHQL driver supported the Realtek NIC on the x79. During the hardware detection process the Realtek NIC had been installed.
With the Realtek setup I was able to go out onto the Incredible Edible Internet and download the Intel INF drivers for the x79 and get that huge list of chipset related devices sorted and installed. With the chipset drivers installed I moved on to the two other driver installs I knew would require the use of an installer: the NVIDIA drivers and the X-Rite Eye Match package for color calibration.
At this point I stopped and took stock of the driver software that remained. I don’t really trust software installers and I don’t trust drivers either. So I was looking to determine if the installers were necessary or merely a convenience. After finding that none of them needed their installer I turned to another Windows command to assist me: PNPUtil.exe.
PNPUtil allows you to add, list, or remove drivers from the driver store. With an Admin instance of command prompt in hand I was able to remove some old lingering drivers in the store and to install the drivers for the remaining unidentified devices.
Notes:
- According to the ReadMe for the ASMedia USB 3.0 drivers, the installer when tasked to uninstall wouldn’t actually remove the drivers. So the installer is effectively worthless. It’s stuff like this that reinforces my peculiar habits. :P This was the only driver that required me to use the group function of PNPUtil to install multiple INFs instead of just one INF.
- Using the Intel NIC driver setup (for the second NIC) ends up offering the ProSet tools for advanced functionality like VLAN support and Teaming. More than I need, albeit it does at least have a real uninstaller. With the help of the Windows Search Bar I found the exact INF I needed and did a driver only install into the store with PNPUtil.
- The Marvell controller (mv91cons) only required the single INF installed for the management device, not the multiple bundled in the driver distribution.
bcdedit /set useplatformclock true
Unlike with the x58 boards, Asus offered the option of disabling the High Precision Event Timer (HPET) on their x79 series. I haven’t looked at the specification update for the x79, but presumably there shouldn’t be any flaws with HPET like those that lingered in the ICH10 on the x58. Not using the listed command will result in a Windows install sitting on a version of the Time Stamp Counter. The command will transition you to native HPET, this change can also be observed through a change in frequency for the hardware timer to 14.31818MHz.
One last reset to initiate that change and boom, I’m back up and running after entirely too much effort. God I do love new computer hardware, but I so loathe to put it together.
So of course after all that effort did it fix my woes? Yes. Sleep works perfectly. I also gained a learning experience as I don’t often build PCs these days. From a usability perspective the extra two cores will help in my VM work and so really the last major bottleneck I have is with the disk subsystem. For now I’m very happy and this more than meets my needs. Hell, with luck this platform might even be a drop in upgrade for IB-E.
For those of you who stuck around until the end of this rendition of War and Peace: thank you! As I was typing this the word counter at the bottom stopped about halfway through and simply said: TLDR. If a Corsair rep might read this I got one two words for you to take back to your bosses: Need manuals.