Jcloud

Pfsense running on a Jetway NUC.

What sort of work do you want these to do? Office clerical? compiling? Stats? I just want a feel for workload vs expectations is all, and to better answer question of performance and ram. But, from your server specs, I'd guess it should be very plausible.

I had same line of thoughts. With the Third gen K series core (3930K) had 36 lanes? Then at some point Intel dropped the number of lanes in their CPUs (to 24) to reduce cost and because there wasn't high demand for Crossfire and SLI. Again I forget where all the cut-offs, specifics, and changes were made so this makes for handy resource for researching Intel's stuff: Intel Ark Page

Edit (2/3/2018): As of version 6.4.1 unRAID contains the clean patch by the Linux Kernel developers, and therefore this should NOT be used anymore. This is for any ThreadRipper owners out there wanting to compile their own kernel for, "Ugly Patch." I've finally wrapped my head around CHBMB's script (again thanks for your help @CHBMB ) so here is a script which accurately redoes my steps that led to this -- I use this wording, because my first compile was more by hand as I was learning. I have tested this on my own setup and seems to be working correctly. For you Ryzen folks, this is NOT for your cpus! (just in case) USE AT YOUR OWN RISK!! Instructions: Make a work folder for this script to do its work. Recommend putting it in /mnt/cache and if you don't want it to show up as a user share make it a hidden file/folder, (I.E. leading period) /mnt/cache/.WorkFolder/ copy supplied .tgz to folder. Then in CLI: tar -xvf uglyPatcher.tgz IF you are running a VM right NOW, set the ALL auto-starts for VM to OFF and restart your server. Once or twice I had errors in compiling when I had VM's up, probably was me (because this was in my first attempts). However VM was lagged to worthless during compile as the script uses all your cores If one was so inclined the script could be changed to use X threads. After reboot go to work folder in console, and type : ./uglyPatcher Now given your internet connection it will probably take about 3-6 minutes. The file everyone here will want will be in $YourWorkFolder/$VERSION/bzimage-tr Now in the work folder we go to the newly compiled bzimage, copy, and test. (( Example as prompt )) root@HYDRA:/mnt/cache/.custom# cd 6-4-0/ root@HYDRA:/mnt/cache/.custom/6-4-0#mv /boot/bzimage /boot/bzimage-STOCK6_4_0 root@HYDRA:/mnt/cache/.custom/6-4-0#cp ./bzimage-tr /boot/bzimage root@HYDRA:/mnt/cache/.custom/6-4-0# Once file has been moved, reboot and test. Information worth noting: If anyone looks at what I did there are a couple of ugly-bits I did to make it work around some of my bits of ignorance. Most notably pci_tr.c file, I tried to use a .patch file initially, but compiler would reject it and I wasn't sure why. Did I need a CLI flag to accept the .patch file because it's not a sanctioned patch? No clue. My main concern, as a package like this, is that over time pci.c may change and since pci_tc.c is simply a modified version of pci.c over time it might make errors -- solution being just go to source and make a new one (clunky I admit). uglyPatcher.tgz

Queue the South Park ski instructor, "You're going to have a bad time." Those 1x to 16x are generally useful to just the cryto miners because they don't have that much data to pass to GPU, and therefore don't need all the bandwidth in a 16x bus. Personally I would look for a motherboard that has x16 slots which will fit all the cards, or if space is impossibility a 16x to 16x pass-thru to keep performance but able to move the physical bits would be a better idea. Problem with this is that Ryzen seems to have a number of issues with C-power states, check the forums you'll see lots of heads bashing against the desk on these currently (could always change, but for now I would recommend holding on buying Ryzen). Threadripper could be an option but again, this has been hit or miss depending on the board you get and the bios revision -- as for working TR boards, I can tell you from personal xp, ASUS PRIME X399-A is working on "Ugly patch." If you're set on 4-gaming virtual rigs I would stay away from the Intel core-K processors; mainly because you're going to run out of PCIe bus with four graphics cards -- okay this will be true for 6,7, and 8 series; I'd have to go back to Intel Ark page to remember where the Core CPUS have the non-nerfed number of PCIe channels. My 2-cents

For anyone lurking here. If you're on, 6.4.0 and want to give "Ugly Patch" a shot I've compiled it, see this post My system has only been running it for about eight hours now. Edited follow up report (1/21/2018): Been running my compiled "Ugly patch" since time of post. For me it's been 100% functional on 1 gpu VM; have yet to try two gpu vm's (waiting on my friend to upgrade his computer ((my hardware)) and 'deal out all the cards in my deck').

For gaming and having another VM on the side, it might be tight depending on the games you want to play. Also, I don't understand why you wouldn't just play the game in the WindowsVM directly? Single monitor setup but if you wanted to pass graphics cards to each VM (assuming I'm reading this correctly)? And assuming the monitor has multiple inputs (okay now I'm dangerous territory:double-assumption), you could use the input switch on monitor between the two? If I was to do it, I'd shoot for: WindowsVM: 2 cores + their HT's ==> 4 vCPU's 8GB of RAM provisioned LinuxVM: 1 core + it's HT ==> 2 vCPU 2-4GB of ram provisioned unR Host: (The rest.) Nope, number of cores your VM has is parameter defined at it's execution (hence the cool templates in webgui . Your way will be good for WindowsVM, but I'm not sure how pleased you might be with your LinuxVM performance. Give it either/both a shot though. I suspect you'll be longing for more RAM. Also, you might need to dump the ROM for your graphics card, although looking for the link it appears there is another way to do it (at the time I setup my vm I had to dump my graphics card rom). I guess, if I have a point, treat it like a programming project: Hello world ==> just a simple vm add a feature ==> more provisioning. n+1 ==> gfx card working or second VM at same time. n++ ==> Other previous undone option tweak it and play That's my, two cents.

Yup, yup. So far, I've been lucking out -- "Ugly patch" been working well. I also keep an eye out on my mobo support page for firmware updates - although ASUS documentation of these firmware updates on their site is nonexistent. I also told my friend he couldn't get the upgrade until, "I thought my system would run without crashing." (he's not a computer-guy).

I did this with my 3930K, GTX1070, and R9 290 for about a year. I'd pin four cores (2 cores and their HTs), 16GB of ram, keyboard, mouse, and a PCI address which resolved to a USB3 hub on the controller (so each VM had USB3 for gamepads/USB devices); audio passed through graphics cards to HDMI. That left me with 32GB for host and 4 cores for Host/Docker. I also had a, "Pickle Rick" VM which was basically 32GB of ram allocated and 6 cores to my GTX1070, for when I gamed alone. I found it to be a fun technical exercise. For performance reasons, host/docker, gpu_vm0, and gpu_vm1 all had their own cores pinned. I haven't done it on my Threadripper yet, mainly because I was planning on giving old MB and the R9 to a friend, so I can get back my 1st-gen core2dou, and he can get an upgrade finally. lol

I own one. The Antec 900 is a solid case, it would be a decent buy. The main CON I found with it, there is no room to run cables behind the motherboard or motherboard tray. So be sure to run good lines with your cabling, and also it can get cramp in cabling because you can't go behind, example the 12-volt CPU power cable, best route for that is to have it go over your motherboard, but under your graphics card -- if the header is located in the back of the case (or top left corner, looking inside) otherwise if the header is on the right you go around the board, but depending on your power supply this route could get a tad short; and you take up more room you'll need for enclosures.

Woh, wave of nostalgia seeing those sorts of beasts in Lawyer offices or cdr cloning days of old. Queue, the South Park detectives, "Niicce!"

Actually I just bought this case. General impressions: it's exactly what I wanted ATX, 9x 5.25" bays that you mentioned, the metal used is good gauge, it's not flimsy. CONS: This S.O.B. IS BIG! Really big! By BIG I mean the height, you look at it and you find yourself asking, "Am I over-compensating for something? Sigmund?" I personally don't like the top bit, The plastic is like a 90's computer front-panel cover. Functionally I like it, and as you said it's inexpensive. Aesthetically I perfer it with out the plastic top and front cover, but that's me. Also note below the 9x 5.25 external bays there are two more INTERNAL 5.25" bays below it with some 2.5" to 3.5" shims it's a good place for cache/SSD drives. My grand scheme is to couple the case with three of these Icy Dock 3x(5.25" bays) 5x(3.5" drive tray less enclosure) I have one, need to buy/grow into the next two. Forgot to mention, if you want to go with the trayless Dock I suggested, you'll need to do some case modification (dremel with a number of disc cutters) to remove the bottom holding tabs for each of top two slots (out of the three the bays take up).

I would just copy the DOMAINS folder over to your new server. Then use the VM templates; make a new VM setup how you want it; click advanced slider; change the hard drive disk image over from, "automatic" to "manual" and point it at your old hard drive image. Windows won't have any problems as long as you use the same machine-type/bios emulation. Some work in making the VM's from templates again, but at least that way you're not dealing with hardware specific addresses from old box to new box.

http://www.r-studio.com/ Won't be able to recover the partition out-right, but will find files that haven't been zero'ed out.

Steam, a user share and a folder on my cache drive. I'm too old, lazy, and have better things to do with my time then to find; verify clean; and cobble Windows executive to working order to save on something I can get for an On-sale price on Steam platform. Also, I don't have any news-net accounts, which most scrapers seem to utilize, so little point for me to trying one of these Dockers. Steam is currently the only DRM platform I tolerate for personal use.

Well, that's fair, I must admit I haven't touched my DNLA docker (Emby) for a while due to not needing/using it. Good luck.

Guilty of this. I bought myself a Threadripper 1950X, Prime X399-A, heatsink, and 64GB of RAM to replace my six-year old 3930K; I did well in BTC trades this year; my short-term profits paid for new shinny bits.

I've seen some people in the forums setup their transcode/stream-buffer folder to /tmp/PlexFOO that way it utilizes the ramdrive and is cleared when you reset. Coupled with making a ring-buffer scheme, that might help you out. Assuming you haven't tired that setup yet. Edit: Found the link I was thinking about, Plex guide to moving transcoding to ram.

I'll admit my solution is more of a work-around. There might be a config fix, for what you're looking for, but my linux-kung-fu is weak. Good luck, hope you find your fix.

That sounds like you need to dump your GTX970 video bios to a file and then added it to your VM's config file. Check out: this thread on, "How to pass through an NVIDIA GPU as primary or only gpu..." I recommend trying to dump your own video bios, rather then using provided links in the thread (for best functionality/compatibility).

To me that sounds like the base of your issue. You need/want to make your cheep GPU the primary video card on the hardware/bios/POST level. Have your system's video post video to your cheap card -- this will give you the unRAID console access you are looking for. Then in web-ui for your VM set it to use your nice video card. If your nice gpu is the primary and you start a VM, KVM grabs the output and it's gone (for console); this has been my experience with it, but I'm not a Linux guru either. Since you can't move cards around, you could try plugging your monitor into the cheap card at boot, and have nothing plugged into your good card -- this might force your system to recognize the your cheap video card as the primary standard-out for computer. If there is a linux configuration way of doing this, I don't know how to do that.

Great, glad I could give back to community for a change. Please edit your original post title to include "(Solved)"

Give this a shot, and see if it helps. Right click on your start button. Click on "Command Prompt Admin" or "Windows Power Shell (Admin)" type: diskpart In the DISKPART> prompt, type: list disk You'll get something like: DISKPART> list disk Disk ### Status Size Free Dyn Gpt -------- ------------- ------- ------- --- --- Disk 0 Online 30 GB 1024 KB * Disk 1 Online 1 TB FOO KB * DISKPART> In my example, I'll reset disk1, type: select disk 1 then Be certain you have the correct drive selected prior to clean. clean This will re-initialize the disk, then using MS tools it can be rebuilt, "the Microsoft way." If you select the disk that has Windows on it, it won't let you run clean, but any other drive no issue. Next, right-click on Windows button and click on, "Disk management." A pop-up should come up asking if you want to initialize the 1TB drive. Defaults should be OK. (GPT, default values in the wizard, give it what drive letter you want). That should make it good to go. If you suspect the drive at all, HD Tune is helpful utility, mainly its "Health" (SMART status) and "Error Scan" (block surface scan); for calling the hardware good/bad.

Current server is 64GB; next week's TR upgrade will also be 64GB, until price of ram goes back down which point I'll go to 128G -- Totally overkill for unRAID itself, but I like and play with VM's.

Thank you, I wasn't aware -- bookmarking your link for myself. I tried to search before posting, to prevent redundant post/info, I guess I should have tried harder.