1812

Do you mean like this? bingo.

stupid question time: when swapping out a cache drive, could you not have just changed the share setting from prefer to yes, and then invoked the mover to send them back to the array, then after swapping out the drive, reversed the process? Or does that eat the files?

like this? Fresco Logic FL1100 USB 3.0 Host Controller (sold as Inatek KT4006) Usage: usb disk access/usb3 to ethernet adapter/keyboard-mouse/bluetooth dongles Rating:Good+, good bandwidth with on card, no aux. power cable needed for card, 1 out of 5 I ordered died within 2 weeks Instructions: for OS X-plug and play though Sierra 10.12.2, windows not tested but manufacture claims windows 7+ compatibility (come with disc of drivers for windows) 666172-001 MNPA19-XTR HP 10GB Ethernet Network Card (Mellanox ConnectX-2 EN) Usage: Moderate to heavy capable Rating: Excellent/Good if you need to cheaply move data faster than gigabit Instructions: no OS X support, win10 plug and play but works better with Mellanox drivers (downloadable from their website,) native unRaid support for server to the use card in 6.2.4 (possibly earlier,) I currently use a pair to backup one server to the other. Geforce GT 710 Usage: good cheap general desktop card, light gaming, watching movies Rating: Good Instructions: Plug and play in OS X through Sierra 10.12.2 as long as smbios set to 14,1 or 14,2 before assigning the card (otherwise black screen occurs,) did not try to give it nvidia drivers Geforce GT 730 Usage: good cheap desktop card, light gaming, watching movies Rating: Good Instructions: Plug and play in OS X through Sierra 10.12.2 as long as smbios set to 14,1 or 14,2 before assigning the card (otherwise black screen occurs,) did not try to give it nvidia drivers GTX 760 SC Usage: mild video editing in fcpx, 3 simultaneous screes utilized Rating: Good+ Instruction: Plug and play in OS X through Sierra 10.12.2 as long as smbios set to 14,1 or 14,2 before assigning the card (otherwise black screen occurs,) using Nvidia web drivers can be a challenge as incremental updates to OS X can cause the drivers to no longer work with the card.

New test for audio/video stability and core pinning in os x (for those who care) For this one, I used one of my other servers with dual e5520 processors. Thread pairings as follows: Proc 1 cpu 0 <===> cpu 8 cpu 1 <===> cpu 9 cpu 2 <===> cpu 10 cpu 3 <===> cpu 11 Proc2 cpu 4 <===> cpu 12 cpu 5 <===> cpu 13 cpu 6 <===> cpu 14 cpu 7 <===> cpu 15 Proc 2 isolated from unRaid. vm1 with a gt730 gpu, assigned cores 4-7 vm2 no gpu, apple screen share, assigned cores 12-15 emulator pin 0-3 for both vm's vm disk images located on same ssd This cpu assignment was selected because it is accepted to cause audio/video issues when placing two vm's onto the same physical core(ht pairs). Result when vm2 was running cinebench with its cores at 100%, vm1 had zero audio/video issues streaming youtube, netflix, or blu-ray movies on vlc from a network server. No matter activities I did, I could detect no slow down, no lag, Just for fun, I put both vm’s on the same thread/core: vm1 with a gt730 gpu, assigned cores 4-7 vm2 no gpu, apple screen share, assigned cores 4-7 youtube videos and local blu-ray streamed with little video quality issues but audio was at times dirty and/or non functional (to be expected) even when vm2 was only idling. Interestingly, when cinebench pushed the shared cores at 100% during blu-ray playback, the video quality did not suffer at all, but as before, audio was less than desirable when it actually worked. So in my case, it is still fine (if not better performance-wise) to put os x vm’s on not ht paired cores since it does not cause audio/video issues while retaining he benefits of higher performance when the other vm isn’t 100% active. Why is like this? I don't know if this is OS X specific in regard to not causing the audio/video issues that are said to occur in windows 10 with this core assignment , if it is the particular gpu that is less susceptible to audio noise in ht pinning, if it is due to the fact that I am using it on older enterprise equipment, or only a gpu on a single vm (though people have reported issues with gpu-less dockers causing problems on ht cores of vm’s.) ok, no more tests for a while...

and the results are in.... Details: HP DL380 G6 dual processor server Xeon x5670, 6 core HT processor, 12 logical cores per processor 2nd Processor used in testing isolated from unRaid/dockers/etc vm’s on separate unassigned SSD drives Cinebench R15 (higher scores are better) no emulator pin set 8gb ram on each vm clean non updated OS X sierra, (matching disk images) no GPU, vm’s accessed via apple screen share unless stated otherwise, cpu cores were run at 100% for benchmark, run 3x back to back no topology defined (I don't think os x cares, based on some previous tests I ran which showed almost no difference) Thread pairings Proc 1 cpu 0 <===> cpu 12 cpu 1 <===> cpu 13 cpu 2 <===> cpu 14 cpu 3 <===> cpu 15 cpu 4 <===> cpu 16 cpu 5 <===> cpu 17 Proc 2 cpu 6 <===> cpu 18 cpu 7 <===> cpu 19 cpu 8 <===> cpu 20 cpu 9 <===> cpu 21 cpu 10 <===> cpu 22 cpu 11 <===> cpu 23 ----------------------------- section 1 all testing in this section on following cores: vm1 6-11, vm2 18-23 vm1 6 cores non ht paired vcpus, vm 2 off (vm1 baseline score for non ht pairs) 521 527 526 vm1 while vm2 idle on ht cores 517 522 520 vm1 100% with vm2 20-50% use on ht cores 464 441 463 vm 2 while vm 1 idle on ht cores 497 514 515 vm1 & vm2 max 100% usage (simultaneous benchmark from here on down) vm1 314 vm2 309 vm1 315 vm2 311 vm1 317 vm2 315 ------------------------- section 2 all core configurations listed on each the test vm1 on 6-8, 18-20 (vm1 baseline score for HT paired cores) vm2 on 9-11, 21-23 HT core Pairs for each vm vm1 347 vm2 346 vm1 345 vm2 345 vm1 349 vm2 346 vm 1 on 6, 18, 7, 19, 8, 20 vm 2 on 9, 21, 10, 22, 11, 23 mixed vcpu ordering of ht paired cores vm1 346 vm2 345 vm1 346 vm2 344 vm1 348 vm2 346 vm 1 & 2 sharing 2 cores (8,20) utilizing only 10 total cores, but 6 on each vm vm1 on 6-8, 18-20 vm2 on 9,10,8, 21, 22, 20 (funny pairing to keep vcpu0 off the shared cores) vm1 286 vm2 286 vm1 286 vm2 283 vm1 288 vm2 285 and just for fun vm1 & vm2 on same cores (6-11) 100% utilization vm 1 254 vm2 255 vm1 254 vm 2 257 vm1 250 vm 2 255 -------------------------- What does it mean? Using HT paired cores scores a max of 349, vs non ht cores max of 522 with vm2 idle, meaning non ht cores have 50% more power in this setting. Even with 50% usage on vm2, vm1 only drops to 441, which is sill about 27% more power than ht paired cores (which makes sense actually.) Even if vm 2 is off, the max score of vm1 does not change if using ht pairs, so there is no possible performance gain regardless of what any other vm is doing or not doing, unlike using non ht paired assignments. Now, if both vm's are using 100% CPU resources, then ht pairs are the clear winner by about 10% over non HT core assignment which at best hit a 317 (compared to 349.) No video/audio testing was done in this situation because I only have 1 video card in this particular server. For many people none of this may apply. But for someone like me who needs cpu power and not graphics/audio on all vm's, then this has some merit and interest. I may drop in a card from on of my other servers and test for stability in the future. this actually explains quite a bit for me on why I thought my machines were wonky, it's more how I'm using them vs operating differently.

The cinebench scores I posted earlier show the opposite in os x. Could you try some benchmarking of paired HT cores and non-paired vcpus in OS X? Maybe this will shed some light if my machines are just wonky of if os x is different in pairing cpu cores vs. windows. Yes that is to be expected if the situation is like this. my cpu pairing is like this cpu 0 <===> cpu 14 cpu 1 <===> cpu 15 cpu 2 <===> cpu 16 cpu 3 <===> cpu 17 cpu 4 <===> cpu 18 cpu 5 <===> cpu 19 cpu 6 <===> cpu 20 cpu 7 <===> cpu 21 cpu 8 <===> cpu 22 cpu 9 <===> cpu 23 cpu 10 <===> cpu 24 cpu 11 <===> cpu 25 cpu 12 <===> cpu 26 cpu 13 <===> cpu 27 so if i pinned 8 vcpus 0,1,2,3,4,5,6,7 then they would be over 8 separate cores so would get good performance as those cores would not be doing anything else, But, Then if i set up another vm with 8 vcpus 14,15,16,17,18,19,20,21 this would be sharing the same 8 cores and when both machines run at once then performance would be bad. So over those 8 cores 2 vms should be, vm1 . 0,1,2,3,14,15,16,17 vm 2 4,5,6,7,18,19,20,21 That way there is no overlap. Hope that makes sense yes and no. If you have a bunch of cores, then you can run them non HT and get good performance. If you don't, you get reduced performance. Just for kicks, i'm going to setup a couple 4 core vm's tonight on each other's pair and run some simultaneous tests... I do expect degraded benchmarks, but I'm curious to know if it's better to have 1 vm on it's own ht pair, or let it share with another vm, which may not be using the ht pair at the same time as the vm, therefore lessening the performance hit vs using ht pairs... (sorry if that's confusing, still on a bunch of medicine... results in a few hours though...)

The cinebench scores I posted earlier show the opposite in os x. Could you try some benchmarking of paired HT cores and non-paired vcpus in OS X? Maybe this will shed some light if my machines are just wonky of if os x is different in pairing cpu cores vs. windows.

Good that you got it running. Want it faster? Where is the vm mounted? I see it's on a cache drive, but is that an SSD or spinning disk? Spinning disk is much more sluggish. Even if it's an SSD cache, some believe that it is better for disk i/o to have the vm on an SSD mounted via unassigned devices. That's how I do it, but for other reasons. Also, cpu pinning.... What i've found is that the common idea of using HT cores with os x vm's don't give the best cpu performance. For example, I ran the following cinebench test this morning on 12 of my 24 cores (on 2 processors) Non paired threads spanning 2 processors <vcpupin vcpu='0' cpuset='12'/> <vcpupin vcpu='1' cpuset='13'/> <vcpupin vcpu='2' cpuset='14'/> <vcpupin vcpu='3' cpuset='15'/> <vcpupin vcpu='4' cpuset='16'/> <vcpupin vcpu='5' cpuset='17'/> <vcpupin vcpu='6' cpuset='18'/> <vcpupin vcpu='7' cpuset='19'/> <vcpupin vcpu='8' cpuset='20'/> <vcpupin vcpu='9' cpuset='21'/> <vcpupin vcpu='10' cpuset='22'/> <vcpupin vcpu='11' cpuset='23'/> Cinebench scores 1081 1104 1121 Paired HT threads on a single processor <vcpupin vcpu='0' cpuset=‘6’/> <vcpupin vcpu='1' cpuset=‘7’/> <vcpupin vcpu='2' cpuset=‘8’/> <vcpupin vcpu='3' cpuset=‘9’/> <vcpupin vcpu='4' cpuset=’10’/> <vcpupin vcpu='5' cpuset='11’/> <vcpupin vcpu='6' cpuset='18’/> <vcpupin vcpu='7' cpuset='19'/> <vcpupin vcpu='8' cpuset='20'/> <vcpupin vcpu='9' cpuset='21'/> <vcpupin vcpu='10' cpuset='22'/> <vcpupin vcpu='11' cpuset='23'/> Cinebench scores 759 757 749 topology: in previous tests I've done with os x I didn't see any improvement or degradation specifying the topology to the vm. Quite a bit of difference. unRaid shows in the dashboard that when both tests are run, cpu usage is 100% on those cores, but the results are about a 45% increase in cpu power benchmark score when not using HT cores for vm assignments. I'm not saying the way I do is right for your setup, but you could at least experiment and run your own tests. ALSO isolate your vm cores from unRaid if you haven't already, and then set emulator pin to one of the cores you left for unRaid. This makes sure that only the vm uses the cores assigned to it. splashtop/teamviewer may still use cpu to render even if you have a video card. I don't remember... someone else will have to chime in on that. but OS X screen share will use the gpu but sadly no sound... There may be some other gains from adjusting clover settings. I am about to start a new topic on that to compare notes with others and no clutter this one up so much. Thanks for the tips I'm definitely going to do some playing around with it later tonight. I use my vm's for building and developing c++ applications. So most of my build environments are on a spinning 2tb array disk with 2*250gb and 2*256gb cache drives in btrfs raid. My main OS(win10) which i'm typing from now, has gpu passthrough and is stored directly on the raided cache drive. But I've never isolated any cores for the VM's or used an unmounted drive as my belief was that the VM would perform better with paired threads and a cache drive or on the raided cache drive itself. The only thing I have done is opted to leave cores 0/6 free for unraid. 1. Is it possible my other VM's could benefit from core isolation as well? Also, you said that most recommend to run the VM on an isolated drive which is in an unmounted state. 2. Wouldn't the VM benefit more from having a SSD array drive with raided SSD cache IF that SSD array drive was only being used for VM's? Or like I'm currently doing for performance on my main machine which is storing on the raided cache drive? maybe my machines are just weird with OS X, i don't know. but what I do know is that the scores are pretty solid. I use mine as a transcoding cluster for edited/produced video, and can confirm that a vm using non ht paired cores to transcode does it faster than one with ht cores. I haven't done the test with win10 (have a vm but don't use it often) but my guess is that windows, optimized for virtual environments, will perform better with ht pairs like the common usage dictates. 1. all vm's can benefit with not having to share cores with unRaid services and dockers. if you've used isolcpus in your syslinux.cfg, and isolated everything but 0,6, then 0,6 are the only cores unpaid gets. The rest is clear for whatever vm's you run. drives aren't unmounted, they are mounted using the unassigned devices plugin. essentially if you put the disk image there, it doesn't have to contend with any other action/file transfers/etc that the cache drive may end up doing. 2. ssd cache will be faster than spinning cache, which is way faster than spinning array. If the ssd was in the array, it will still be limited by the speed of the parity drive read/writes. With a spinning parity drive, that's 30-50ishMB/s write and probably less than normal ssd reads? Using turbo write you can hit parity protected write speeds of 80-125MB/. I don't know of anyone who has a wholy ssd only array (someone does I bet around here) but they'd be the ones to ask about ssd array performance for specific numbers. Many people just leave the vm's on the cache drive, and it works fine. you all get the benefit of having a redundant copy if 1 of the paired drives fail (depending on raid configuration of the cache pool.) otherwise, you have to back it up yourself or use one of the backup program/scripts (which I don't use, i just move them manually.) But since I have a few heavy plex/docker app users in the house, then it's better for me to move mine off the cache drive and onto an unassigned disk where it gets the full bandwidth available. (if this is rambling, sorry, i'm under a quite bit of medication for seasonal allergies and things...)

Good that you got it running. Want it faster? Where is the vm mounted? I see it's on a cache drive, but is that an SSD or spinning disk? Spinning disk is much more sluggish. Even if it's an SSD cache, some believe that it is better for disk i/o to have the vm on an SSD mounted via unassigned devices. That's how I do it, but for other reasons. Also, cpu pinning.... What i've found is that the common idea of using HT cores with os x vm's don't give the best cpu performance. For example, I ran the following cinebench test this morning on 12 of my 24 cores (on 2 processors) Non paired threads spanning 2 processors <vcpupin vcpu='0' cpuset='12'/> <vcpupin vcpu='1' cpuset='13'/> <vcpupin vcpu='2' cpuset='14'/> <vcpupin vcpu='3' cpuset='15'/> <vcpupin vcpu='4' cpuset='16'/> <vcpupin vcpu='5' cpuset='17'/> <vcpupin vcpu='6' cpuset='18'/> <vcpupin vcpu='7' cpuset='19'/> <vcpupin vcpu='8' cpuset='20'/> <vcpupin vcpu='9' cpuset='21'/> <vcpupin vcpu='10' cpuset='22'/> <vcpupin vcpu='11' cpuset='23'/> Cinebench scores 1081 1104 1121 Paired HT threads on a single processor <vcpupin vcpu='0' cpuset=‘6’/> <vcpupin vcpu='1' cpuset=‘7’/> <vcpupin vcpu='2' cpuset=‘8’/> <vcpupin vcpu='3' cpuset=‘9’/> <vcpupin vcpu='4' cpuset=’10’/> <vcpupin vcpu='5' cpuset='11’/> <vcpupin vcpu='6' cpuset='18’/> <vcpupin vcpu='7' cpuset='19'/> <vcpupin vcpu='8' cpuset='20'/> <vcpupin vcpu='9' cpuset='21'/> <vcpupin vcpu='10' cpuset='22'/> <vcpupin vcpu='11' cpuset='23'/> Cinebench scores 759 757 749 topology: in previous tests I've done with os x I didn't see any improvement or degradation specifying the topology to the vm. Quite a bit of difference. unRaid shows in the dashboard that when both tests are run, cpu usage is 100% on those cores, but the results are about a 45% increase in cpu power benchmark score when not using HT cores for vm assignments. I'm not saying the way I do is right for your setup, but you could at least experiment and run your own tests. ALSO isolate your vm cores from unRaid if you haven't already, and then set emulator pin to one of the cores you left for unRaid. This makes sure that only the vm uses the cores assigned to it. splashtop/teamviewer may still use cpu to render even if you have a video card. I don't remember... someone else will have to chime in on that. but OS X screen share will use the gpu but sadly no sound... There may be some other gains from adjusting clover settings. I am about to start a new topic on that to compare notes with others and no clutter this one up so much.

I have a couple 710 cards and a 730. I just feed them straight to OS X with no issues. I'll try to post rom's soon.

Evga GTX760sc evgagtx760sc.dump.zip

Thanks for the pointers. I made a bunch of changes, removed xvga from the xml, upgraded to Sierra, changed smbios setting in clover to match an older mac, etc. And IT WORKS!!! Finally. It recognized the card and without any boot arguments or installing web drivers, it worked. Phew, spent so many hours Now I need to get audio working ;-) EDIT: Got hdmi audio working thanks to another one of your posts http://lime-technology.com/forum/index.php?topic=51915.msg524900;topicseen#msg524900 :-) Thanks so much I have spent tons of time getting mine working, playing with settings, breaking things, redoing things....but that was also mainly because I was also learning how unRaid works AND how KVM works at the same time. But it was completely worth it and I've loved the learning experience. I couldn't have done it without many other's contributions and help on here, including gridrunner! now----> make a backup copy of your working disk image! Trust me, it saves time vs reinstalling again later. I keep a "base" image to use when creating a new vm or for when I break mine messing around.

I use 2 gt710's and a gt730 (710 variant) with almost no issues... https://lime-technology.com/forum/index.php?topic=54786.msg523314#msg523314 I use host dev on them all. no nvidia drivers. no boot args. but that is no your problem. (sharing is caring though) what I use (from a running vm, so it has a few extra lines auto populated): <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x07' slot='0x00' function='0x0'/> </source> <alias name='hostdev0'/> <address type='pci' domain='0x0000' bus='0x02' slot='0x01' function='0x0'/> </hostdev> <hostdev mode='subsystem' type='pci' managed='yes'> <driver name='vfio'/> <source> <address domain='0x0000' bus='0x07' slot='0x00' function='0x1'/> </source> <alias name='hostdev1'/> <address type='pci' domain='0x0000' bus='0x02' slot='0x04' function='0x0'/> </hostdev> try: removing "xvga=yes" from your xml. I don't see it in mine, for this card or my gtx760 Look in the boot log for the gt710 card, search by device id and by it's slot assignment and see if it is showing any errors there, like a bios bug or similar. also, try it in a different slot. sometimes that helps some folks.

has worked on 3 nvidia cards for me, after I was pointed in the right direction This worked almost perfectly for me to get HDMI sound from my Radeon RX460! Thanks!! Only issue I have is that there seems to be a very minor crackle/pop every few seconds from the sound. It's easiest to hear if you listen to something without any variation. I find listening to a tone makes any crackles/pops immediately obvious: Ex: https://www.youtube.com/embed/TxHctJZflh8 Any thoughts on a cure? Later in the thread there's talk of isolating cpus and such. Would this cure a minor crackle/pop in the sound from HDMI? It wouldn't hurt and will make it perform a little bit better. How many cores does your system have and how are you assigning them?

downgrade to 6.2.4 And that solves what exactly? 6.3 uses 6.2 ovmf as far as I know. it solves making you vm working again and eliminates all the other things in the release client that you don't know about that could be causing the issue.

downgrade to 6.2.4

So far so good! No video issues with it at all. Boots every time, and has been running without problems for a week straight. Only thing I haven't tested is audio. My monitor does not support HDMI, so I'm not sure how the audio functions on it. That said, I have purchased an hdmi matrix that splits the hdmi audio out and will hopefully prove to be a solution to macOS audio. I've tried external usb cards (dongles), and they work great on my WInodws 10 VM, but I cannot get them functioning in Sierra... It doesn't benchmark very well, from a gaming perspective, but if you're just looking for Office and design use, I think it's a great choice... Awesome! thanks for the info! yea, I don't care about gaming. Just looking to sell my iMac and use this as my desktop machine for general stuff! I run 2 zotac gt 710 and 1 msi gt 730 cards. All have worked out of the box for video. All run audio with the hdmi kext added. They are very inexpensive good choices for basic computing and very light gaming (like roblox, but with some image tearing here and there.) I haven't watched the video yet but is it mentioned in the video or somewhere here how to add the hdmi kext, and which one to get? I used to build hackintosh's but its been a few years. Thanks! +1! Not sure which kext. Please expand :-) nvidia (and maybe others?) hdmi audio you need: https://dl.dropboxusercontent.com/u/248599/HDMIAudio-1.0.dmg note: as with all downloads, be vigilant. this link come from https://www.tonymacx86.com/threads/clover-hdmi-audio-for-nvidia-amd-cards-easy-way.145399/ and is probably fine. then: has worked on 3 nvidia cards for me, after I was pointed in the right direction

So far so good! No video issues with it at all. Boots every time, and has been running without problems for a week straight. Only thing I haven't tested is audio. My monitor does not support HDMI, so I'm not sure how the audio functions on it. That said, I have purchased an hdmi matrix that splits the hdmi audio out and will hopefully prove to be a solution to macOS audio. I've tried external usb cards (dongles), and they work great on my WInodws 10 VM, but I cannot get them functioning in Sierra... It doesn't benchmark very well, from a gaming perspective, but if you're just looking for Office and design use, I think it's a great choice... Awesome! thanks for the info! yea, I don't care about gaming. Just looking to sell my iMac and use this as my desktop machine for general stuff! I run 2 zotac gt 710 and 1 msi gt 730 cards. All have worked out of the box for video. All run audio with the hdmi kext added. They are very inexpensive good choices for basic computing and very light gaming (like roblox, but with some image tearing here and there.)

Though I have different hardware, I figured I'd give this a shot. Since I am new to unRAID, where do I find the sysconfig file? the sysconfig file is on your flash drive, in the syslinux folder. modified with the text added, it would look like this. default /syslinux/menu.c32 menu title Lime Technology, Inc. prompt 0 timeout 50 label unRAID OS menu default kernel /bzimage append vfio_iommu_type1.allow_unsafe_interrupts=1 initrd=/bzroot label unRAID OS GUI Mode kernel /bzimage append initrd=/bzroot,/bzroot-gui label unRAID OS Safe Mode (no plugins, no GUI) kernel /bzimage append initrd=/bzroot unraidsafemode label Memtest86+ kernel /memtest are you getting the "unable to get iommu group" error or something else? allowing unsafe interrupts can sometimes fix that, as is the problem on my hardware. I only started using unRaid about 6 months ago and learn something new almost every other day...

I was setting up my last server tonight and was having issues getting a vm running via passing through a gpu, getting iommu error as follows: internal error: process exited while connecting to monitor: 2016-12-18T02:12:40.093534Z qemu-system-x86_64: -device vfio-pci,host=07:00.0,id=hostdev0,bus=pci.2,addr=0x1: vfio: failed to set iommu for container: Operation not permitted 2016-12-18T02:12:40.093567Z qemu-system-x86_64: -device vfio-pci,host=07:00.0,id=hostdev0,bus=pci.2,addr=0x1: vfio: failed to setup container for group 18 2016-12-18T02:12:40.093575Z qemu-system-x86_64: -device vfio-pci,host=07:00.0,id=hostdev0,bus=pci.2,addr=0x1: vfio: failed to get group 18 2016-12-18T02:12:40.093588Z qemu-system-x86_64: -device vfio-pci,host=07:00.0,id=hostdev0,bus=pci.2,addr=0x1: Device initialization failed The offending device is a gt 710. the trick i've learned on dl 380 servers to passthrough a gpu is by adding vfio_iommu_type1.allow_unsafe_interrupts=1 to the sysconfig file. So, it's there, and correct for this server (actually copied the file from one of the others that works perfectly. This server, is one of 3 that are nearly identical in every way: memory, cpu, gpu (including location.) so I started digging through the logs and noticed a few differences between the working server vs this one with the error in regards to the gpu. Instead of posting the entire log, i'll list the differences: working server Dec 17 17:50:12 Brahms4 kernel: DMAR: Hardware identity mapping for device 0000:03:00.1 Dec 17 17:50:12 Brahms4 kernel: DMAR: Hardware identity mapping for device 0000:07:00.0 Dec 17 17:50:12 Brahms4 kernel: DMAR: Hardware identity mapping for device 0000:07:00.1 Dec 17 17:50:12 Brahms4 kernel: DMAR: Hardware identity mapping for device 0000:3e:00.0 error server Dec 17 17:59:12 Brahms2 kernel: DMAR: Hardware identity mapping for device 0000:03:00.1 Dec 17 17:59:12 Brahms2 kernel: DMAR: Hardware identity mapping for device 0000:07:00.0 Dec 17 17:59:12 Brahms2 kernel: DMAR: Hardware identity mapping for device 0000:3e:00.0 so device 0000:07:00.1, the audio portion of the gpu is missing. Now, before you say "ok, the gpu is bad" I actually swapped it out from another server with a known working one. So the gpu is good. continuing on.... iommu assignments from the log: working server Dec 17 17:50:12 Brahms4 kernel: iommu: Adding device 0000:04:00.0 to group 18 Dec 17 17:50:12 Brahms4 kernel: iommu: Adding device 0000:07:00.0 to group 19 Dec 17 17:50:12 Brahms4 kernel: iommu: Adding device 0000:07:00.1 to group 19 Dec 17 17:50:12 Brahms4 kernel: iommu: Adding device 0000:3e:00.0 to group 20 error server Dec 17 17:59:12 Brahms2 kernel: iommu: Adding device 0000:03:00.1 to group 14 Dec 17 17:59:12 Brahms2 kernel: iommu: Adding device 0000:07:00.0 to group 18 Dec 17 17:59:12 Brahms2 kernel: iommu: Adding device 0000:07:00.1 to group 18 Dec 17 17:59:12 Brahms2 kernel: iommu: Adding device 0000:3e:00.0 to group 19 so, different group assignment, but both required parts of the gpu are there. but before the error server iommu groupings is the following: error server Dec 17 17:59:12 Brahms2 kernel: DMAR: Ignoring identity map for HW passthrough device 0000:07:00.0 [0xcf63e000 - 0xcf63ffff] Dec 17 17:59:12 Brahms2 kernel: DMAR: Setting identity map for device 0000:07:00.1 [0xcf63e000 - 0xcf63ffff] a little bit later the following shows up: error server Dec 17 18:06:59 Brahms2 kernel: br0: port 2(vnet0) entered forwarding state Dec 17 18:06:59 Brahms2 kernel: vfio-pci 0000:07:00.1: Device is ineligible for IOMMU domain attach due to platform RMRR requirement. Contact your platform vendor. Dec 17 18:06:59 Brahms2 kernel: br0: port 2(vnet0) entered disabled state I compared the bios dates on the error server and the other working ones, and oddly enough it has a roughly 1 year NEWER date. Perhaps functionality broke because of that. Either way, I'm not going to pay hp for their most up to date firmware on older hardware like this. Looking around this forum, this issue only appears a few times but with no real resolutions for the gpu assignment (one for hdd availability but no gpu conclusion stated.) In one of the previous threads on this topic, jonp made a suggestion about using acs override only for the offending device: This suggestion seemed to not work for the original poster, but I thought i'd give it a shot. So i changed the device id's appropriately, and rebooted. This split the gpu into 2 different iommu groups. I then fired up the vm and NOPE. Same error. I went back to the log and it still showed "0000:07:00.1: Device is ineligible for IOMMU domain attach due to platform RMRR requirement." So I tried removing that part of the gpu (07:00.1) from the xml (since it was now in a different iommu group) and the vm booted up just fine. Since i'm not using hdmi audio passthrough, it doesn't really matter that the audio portion of the gpu is not there. This was a bit of a long read, but I thought it was better to document it incase someone else comes across this issue. Actually, this post started off as a request for help, and as I got half way trough, I thought I should at least search the forums first regarding this problem and voila! This turned from a question to a solved problem.

Leaving this here for anyone doing what I'm doing and making an apple compressor transcoding cluster. I spent all day on this error. Finally figured it out. Running Compressor 4.3 in Sierra apparently now requires you to pass through a video card to the vm. Otherwise, it gives you the following error "Failed 3x [computer name] crash service down" and will not transcode. This generic error message can mean many things, but was resolved when I fed the vm a geforce gt 710. Looks like i'm going to go buy a few more tomorrow.

Just to confirm, you're talking about the GTX 760, not the 760ti, right? yes, Gtx 760, 2GB, EVGA.

For the sake of not having too many nested quotes, the mouse being stuck is no longer an issue with the geforce 710. I did not do anything in clover or add drivers. All I did was plug in a monitor. It appears the issue arises when trying to access it only as a vm with splashtop, vnc, or other viewer but not apple screen share (which has no audio.) Seems as if I will need one of those monitor emulators they sell on amazon as a solution for now.

In clover there shall be a boot argument that shall enable use of this driver, check nvda_drv=1 EDIT For me I was loosing HDMI sound with nvidia drivers, even if I reinstalled HDMIAudio-1.1.dmg. Back to OSX Default drivers :-) //Peter I'm trying to assign a GT 710 to sierra. I've tried all combinations of no driver and boot arg: nada_drv=1, with drivers and boot arg, and can't get the mouse out of the left hand corner. What am I missing? machine type is 14,2 and recognizes the card in system information. My understanding is the nvda_drv=1 boot arg is no longer applicable to Sierra. See this thread for more info on the new solution. Also, I'm not sure if this will fix the mouse in the corner issue. Frankly, I'm still struggling to get passthrough video working in Sierra so unfortunately can't be of more help. I'm hoping the upcoming holiday may give me some time to dedicate to the problem! my gtx 760 works fine in sierra using web drivers. I sort of fumbled around and found what worked with it, including mouse. This little 710 is becoming a pita. I'll check out the tonymac link.

In clover there shall be a boot argument that shall enable use of this driver, check nvda_drv=1 EDIT For me I was loosing HDMI sound with nvidia drivers, even if I reinstalled HDMIAudio-1.1.dmg. Back to OSX Default drivers :-) //Peter I'm trying to assign a GT 710 to sierra. I've tried all combinations of no driver and boot arg: nada_drv=1, with drivers and boot arg, and can't get the mouse out of the left hand corner. What am I missing? machine type is 14,2 and recognizes the card in system information.