jbartlett

I saw a 3% increase in the CPU score with 3DMark Timespy on a Ryzen 2950 (all but 4 cores assigned to the VM) after disabling the protections but several other tests/benchmarks showed no change within a small margin of error on it and a 1950x and an Intel CPU (can't think of the model off hand).

Updated three servers, no issues. New BIND method working seemingly perfectly vs stubbing.

I wouldn't replace them right away. Run a benchmark on the drives with the speed variations every 1-3 weeks to see if those variations are steady or not. If they're steady, it may just be how that drive has been mapped out at the factory. There could be more factory mapped bad spots in those regions or some weirdness with drive geometry. If they're Seagates, I've noticed that many of the drives have a slow area at the start of the drive. It's consistent enough across drives that I switched to Western Digital.

2950 - I planned on building two of these, it'll go into the 2nd build.

I just picked up a 2990wx to switch into this board. Exciting times! Needed the extra juice.

With Downstream set, I was able to pass a video card and an onboard LAN to a VM once. The VM would start after shutting down but there'd be no response from it. Haven't tried combinations of one or the other yet with Downstream set though I suspect the issue is related to the above. I plan on investigation further but won't be able to until tomorrow.

OP updated with additional onboard passthrough information. Also learned that the 10GB passthrough issue only happens if a video card is also passed through. Dropping the video card after the 10GB isn't responding and starting up the VM, the 10GB works - and doesn't if I put the video card back on. The two Quadro's don't need their BIOS dumped and included to work but I wonder if that's somehow affecting things by not doing it.

Oh man, I wish!

Slot 4 can be passed through to a VM with PCIe ACS override set to Multifunction. In progress of testing what else I can pass through in this state.

My memory might have been faulty.

Okay - whittled it down a bit. Seems like sometimes the 10GB controller survives a reboot/reset, sometimes it doesn't, when passed through to a VM.

Twice now, the 10GB onboard LAN wouldn't work and would disable/enable itself every 10 seconds or so. It happened using it as unraid's port and when bound to a VM. Rebooting the box fixes. Not sure yet what causes it but when bound to a VM, it started when I rebooted the VM after a day of heavy usage. I've rebooted the VM several times since then with no issues with the 10GB port.

Testing the intended use of this motherboard, I've been ingesting four 4K Brio's via an USB3 add-on card passed through to a Win10VM and outputting a 4K stream to YouTube as well as outputting those 4K Brio's via NDI. Stable as a rock after 12 hours. I also had 4 additional USB2 web cams (Logitech 922/930, 2 each) passed into it via the motherboard's USB3 ports who's controller was passed through to the VM. The USB2 cams were able to be passed through as a USB passthrough (no controller) though that was a shits-n-grins test.

Updated OP: The 10GB LAN port can be passed through

Yes, a GeForce GT 1030 single-slot card using it's fan to also bring air over the USB3 card which I have in the 2nd slot.

Added onboard USB3 controllers that could be routed to a VM

ASUS ROG Zenith Extreme Alpha X399 HEDT Gaming Motherboard AMD Threadripper 2 (TR4) EATX DDR4 M.2 10G LAN USB 3.1 Gen2 https://www.amazon.com/gp/product/B07M6SD5GP/ https://www.newegg.com/Product/Product.aspx?Item=N82E16813119186 A video card in PCIe slot 1 will pass through to a VM if you dump the card's BIOS and pass it in on the VM configuration, however, a nVidia drivers will detect KVM and will not initialize (Code 43). Not passing through the BIOS results in no video for the VM in slot 1. The PCIe 4 (x4) slot can't be passed through (it's grouped with a SATA controller) without PCIe ACS override. Slots 2, 3, & 5 were able to be passed through to a VM with PCIe ACS override set to "Disabled". nVidia drivers do not detect KVM and report a Code 43 if Slot 1 has a graphics card dedicated to unraid and the nVidia card is in a different slot passed through to a VM. To enable full virtualization for unraid in the BIOS: Advanced > CPU Configuration > SVM Mode: Enabled Advanced > AMD PBS > Enumerate all IOMMU in IVRS: Enabled NUMA Node info for the 2990WX Node 0: PCIe 3 & 5, CPUs 0-7/32-39 Node 1: CPUs 16-23/48-55 Node 2: PCIe 1 & 2, CPUs 8-15/40-47 Node 3: CPUs 24-31/56-63 Notes: The onboard video can take a bit to display on power up, be patient. After a BIOS update, the system will automatically reboot and your screen will be black/off for awhile. Be patient. 6-7 watts at power off 86-88 watts at OS/idle, one spinner powered down, 3 SSD's No watt measurement under load taken yet Primary slot-1 video card does not init if not attached to a monitor. No option in the BIOS to change which slot is the primary video. If you have the BIOS set to remember the last power on state in the event of power loss and it keeps powering back up after shutting down, change the BIOS setting to disabled, save, and then change it back to remember. The PCIe clamps tend to re-engage after unlocking when you start to pull the graphics card out. If VM's are extreamly slow to start after flashing a BIOS, load the defaults and reset the Enable virtualization. Onboard Controllers passthrough A [1022:145f] 48:00.3 USB controller: Advanced Micro Devices, Inc. [AMD] Zeppelin USB 3.0 Host controller B [1022:145f] 0d:00.3 USB controller: Advanced Micro Devices, Inc. [AMD] Zeppelin USB 3.0 Host controller C [1b21:2142] 0d:00.0 USB controller: ASMedia Technology Inc. ASM2142 USB 3.1 Host Controller *1 D [1b21:2142] 0f:00.0 USB controller: ASMedia Technology Inc. ASM2142 USB 3.1 Host Controller *1 E [1022:43ba] 01:00.0 USB controller: Advanced Micro Devices, Inc. [AMD] X399 Series Chipset USB 3.1 xHCI Controller (rev 02) *1 F [1d6a:07b1] 09:00.0 Ethernet controller: Aquantia Corp. AQC107 NBase-T/IEEE 802.3bz Ethernet Controller [AQtion] (rev 02) *2 G [8086:1539] 05:00.0 Ethernet controller: Intel Corporation I211 Gigabit Network Connection (rev 03) *1 *1 Requires PCIe ACS override=Multifunction *2 Does not always survive a reboot with a Graphics Card also passed through It looks like the wifi & SATA controllers can be passed through if the PCIe ACS override is set to Multifunction but I did not test and don't plan on testing. I did not find a USB controller for the two USB2 ports located near the lower USB3 port on the motherboard itself. If passing through the 10GB LAN to Windows, install the LAN driver in the VM prior to doing so. The attached IOMMU groups shows the details with three video cards and one USB3 card. PCIe ACS override=Both.txt PCIe ACS override=Disabled.txt PCIe ACS override=Downstream.txt PCIe ACS override=Multifunction.txt

Wow, that is an impressive dip! I know I have logic in place to prevent some wonky benchmarks from being submitted to the online database. I need to find time to review this to see if it would catch this test. Though now that I think of it, I should add a "Wonky" flag to my database to exclude them from the averages so you would still be able to purge the benchmark and restore it later (future DiskSpeed update). If you're willing to consider this a junk drive (I would), I would reimburse you for shipping if you mail it to me so I can use it for app logic testing. PM me if you are interested.

Just an update to say I'm not dead. Just busy renovating my house and it eats up most of the time I have available after work.

Good idea! I'll add it to my To Do list.

When it's testing a portion of the drive, it does a balls-to-the-wall straight read using the dd utility which is instructed to log a progress report every second. The read speeds should be pretty consistent from one second to the next. If it detects a large difference in the read speed from one spot to the next, that is typically a sign that something is reading from the drive so it tries again at that same location. It should also be increasing the gap by 5MB each time it retries a spot. Sometimes though, there could be a bad spot there or something else funky going on causing it so I added the option to disable that re-check. It should just keep going and not retry. From what you're saying, it sounds like it's not honoring the disable option and the gap allowed should be a lot more than 45MB after 22 retries. I'll look into this.

It means that you may have saturated your bandwidth. Speeds are supposed to drop as it moves from one end of the drive to the other. If it doesn't drop, then the bus couldn't handle the full read capacity of the drive.

Pushed an update to fix the the Drive Edit to upload a new image. I couldn't consistently get the full page overlay to work like it did before, issues with iframes accessing the parent creating iframes, so I reconfigured it to work inside the Drive Information frame.