billington.mark Posted February 16, 2019 Share Posted February 16, 2019 (edited) Yep, looks like its fixed the driver crippling memory scaling (in windows anyway). Im seeing a 5-10% increase in GPU benchmarks after updating to RC4. Was hoping for more, but it looks like my bottleneck is my aging CPU now! (2x E5-2670). Ive been meaning to put my hand in my pocket and upgrade to a threadripper build for a while now.... Im very interested to see what performance gains you guys are getting after this patch... Thankyou @limetech Edited February 16, 2019 by billington.mark Quote Link to comment
Nooke Posted February 16, 2019 Share Posted February 16, 2019 I'm on i440FX and dunno if it's worth it to switch to Q35. I had serious issues before with Q35 and i440FX is working kinda good right now. Switching template to Q35 broke my system the last time and I had to start by zero. So would like to know if there is much of a difference before starting from scratch again Anyone could do some comparison with both machine types on RC4 and share? Specs in signature. Quote Link to comment
bastl Posted February 16, 2019 Share Posted February 16, 2019 @Nooke I'am currently testing a lot of different settings. Main VM runs still on i440fx and i have a second template as Q35 configured with the same devices passthrough as my main VM. So far it looks promising. Still trying to find the best settings for numa settings pinning etc. Will report back later. Quote Link to comment
Nooke Posted February 16, 2019 Share Posted February 16, 2019 3 minutes ago, bastl said: @Nooke I'am currently testing a lot of different settings. Main VM runs still on i440fx and i have a second template as Q35 configured with the same devices passthrough as my main VM. So far it looks promising. Still trying to find the best settings for numa settings pinning etc. Will report back later. looking forward to it thanks mate! Quote Link to comment
bastl Posted February 17, 2019 Share Posted February 17, 2019 The changes aren't that big in synthetic benchmarks like Timespy or Heaven. In FarCry 5 i saw more improvement. I guess the fact that a real game constantly streaming textures and stuff is what I see here. A synthetic benchmark loads all the stuff right at the beginning into memory. Games like Doom and FarCry at least feel smoother now. Below an overview what I've tested. Test 1 was my original i440fx VM with some manual tweaks like numatune, emulatorpin and iothread set. For test 2 I created a fresh Q35 VM with the same corecounts, RAM, NVME, SSD, GPU as in test 1 and applied all the tweaks from the i440fx VM + the Qemu arguments at the end of the XML <qemu:commandline> <qemu:arg value='-global'/> <qemu:arg value='pcie-root-port.speed=8'/> <qemu:arg value='-global'/> <qemu:arg value='pcie-root-port.width=16'/> </qemu:commandline> Test 3 is a fresh Q35 VM with no manual tweaks. Only GPU and SSD/NVME passthrough, same cores and RAM as before without the Qemu arguments at the end. Test 4 is the same as test 3. I only added the Qemu part. And finally test 5 is basically test 2 with a couple tweaks. In test 5 I changed the memory mode from 'preferred' to strict <memory mode='strict' nodeset='1'/> added some changes in the hyperv section <hyperv> ... <vpindex state='on'/> <synic state='on'/> <stimer state='on'/> <reset state='on'/> </hyperv> and I changed some parts of the EPYC fix old: <cpu mode='custom' match='exact' check='partial'> <model fallback='forbid'>EPYC-IBPB</model> <topology sockets='1' cores='7' threads='2'/> <feature policy='require' name='topoext'/> <feature policy='disable' name='monitor'/> <feature policy='require' name='x2apic'/> <feature policy='require' name='hypervisor'/> <feature policy='disable' name='svm'/> </cpu> new: <cpu mode='custom' match='exact' check='full'> <model fallback='forbid'>EPYC</model> <topology sockets='1' cores='7' threads='2'/> <cache level='3' mode='emulate'/> <feature policy='require' name='topoext'/> <feature policy='disable' name='monitor'/> <feature policy='require' name='hypervisor'/> <feature policy='disable' name='svm'/> <feature policy='disable' name='x2apic'/> </cpu> CPUZ scores also looks pretty good now. WIN10_NVME_UEFII_Q35_RC4.xml Quote Link to comment
billington.mark Posted February 17, 2019 Share Posted February 17, 2019 1 hour ago, bastl said: The changes aren't that big in synthetic benchmarks like ... Thank you for this. This is a great baseline to compare my Xeon build to. Quote Link to comment
Nooke Posted February 17, 2019 Share Posted February 17, 2019 I made some benchmarks aswell. Not much difference between 440fx and Q35. https://docs.google.com/spreadsheets/d/1hhP34z_fSDPy2Uu2umJ1xfq5Zd4_F5IRchqAkUyiXJw/edit?usp=sharing Quote Link to comment
jbartlett Posted February 19, 2019 Share Posted February 19, 2019 Rabbit Trail: I encountered this audio breaking up issue on my Win 10 VM on Threadripper with broadcasting to YouTube via xSplit. I resolved it by not passing through a PCI audio card (on the video card) and passing through a USB2 sound card dongle. What I used: https://www.amazon.com/gp/product/B06XNJLHXS/ Quote Link to comment
bastl Posted February 19, 2019 Share Posted February 19, 2019 @jbartlett Did you had the MSI fix applied for the audio device of that card inside this VM? Quote Link to comment
jbartlett Posted February 19, 2019 Share Posted February 19, 2019 30 minutes ago, bastl said: @jbartlett Did you had the MSI fix applied for the audio device of that card inside this VM? No. I was not aware of this thread at the time. Quote Link to comment
Marshalleq Posted February 22, 2019 Share Posted February 22, 2019 (edited) I've found RC4 helps with Lag etc making it much more viable to game with - However at first attempt I am still showing up PCI Express x1. Does anyone know if I need to modify my vm xml config as well as run RC4, or should it just work out of the box? I'm guessing I need to add this to make it work? <qemu:commandline> <qemu:arg value='-global'/> <qemu:arg value='pcie-root-port.speed=8'/> <qemu:arg value='-global'/> <qemu:arg value='pcie-root-port.width=16'/> </qemu:commandline> Edited February 22, 2019 by Marshalleq Quote Link to comment
Marshalleq Posted February 22, 2019 Share Posted February 22, 2019 For anyone else, that seems to have worked. So not quite out of the box, but at least simple to make happen within the xml. Quote Link to comment
bastl Posted February 23, 2019 Share Posted February 23, 2019 It only works for Q35 Vms. Quote Link to comment
Symon Posted February 23, 2019 Share Posted February 23, 2019 Can I change my i440FX VM to a Q35 or do I have to create a new one ? Quote Link to comment
bastl Posted February 24, 2019 Share Posted February 24, 2019 @Symon Both should work. Depending on the unraid version and the stuff you passing through and how your current VM is setup, upgrading can end with an error. The savest way is to creat a new VM with the Q35 machine type, same cores RAM etc as the original one, different name. You can point it to the existing i440fx vdisk and carry over your maybe manual tweaked parts in the xml to the new one if you need. Before you screw something up, always make a backup first. Quote Link to comment
Symon Posted February 24, 2019 Share Posted February 24, 2019 17 minutes ago, bastl said: @Symon Both should work. Depending on the unraid version and the stuff you passing through and how your current VM is setup, upgrading can end with an error. The savest way is to creat a new VM with the Q35 machine type, same cores RAM etc as the original one, different name. You can point it to the existing i440fx vdisk and carry over your maybe manual tweaked parts in the xml to the new one if you need. Before you screw something up, always make a backup first. Thanks for your help, I will create a new one to be safe Do you know if this creates problems with the Window 10 License as the hardware ID's change? Quote Link to comment
JonathanM Posted February 24, 2019 Share Posted February 24, 2019 13 hours ago, Symon said: Do you know if this creates problems with the Window 10 License as the hardware ID's change? Yes, it can. If you copy over the UUID from the activated VM to the new one, it may help. I don't know how many changes it takes before MS gets cranky, but the less you change between the VM's the better. As he said, keep RAM, core assignments, passthroughs, etc as consistent as possible. Quote Link to comment
Giggity_Grant Posted October 10, 2019 Share Posted October 10, 2019 @bastl how were you able to get Unraid to utilize your 3rd slot GPU instead of the 1st slot GPU (1080ti)? I haven't been able to find a way to force Unraid to utilize my 3rd slot GPU (GT 710) and passthrough my 1st slot GPU (2080 Super), so I just have the passthrough GPU in slot 3 (which is assigned to Node 0) and my unraid GPU in slot 1 (assigned to Node 1). Quote Link to comment
bastl Posted October 11, 2019 Share Posted October 11, 2019 @Giggity_Grant You have to stub/bind the GPU you wanna passthrough so unraid doesn't load it at all. If I add the pci ids of the card in the 1st slot Unraid wont pick it up and will use the card in the 3rd slot. Not sure if I changed something in the BIOS as well but I think I only had an option for the link speeds of the slots not for the order or how IOMMU handles it. Quote Link to comment
Giggity_Grant Posted October 14, 2019 Share Posted October 14, 2019 @bastl Thanks for your reply! I originally tried disabling CSM, which seemed to work - but then I ran into the same issues that other folks in this forum mentioned. After seeing your comment, I re-enabled CSM, and stubbed the pci-ids of the 1st slot card. It's back up and running at least. I'm still running through the thread and trying a few of the mentioned solutions. Is there anything in particular that I would have to add in order to get unRaid to pick up a basic GPU in the 3rd slot, or is this something that happens automatically (since the 1st slot GPU is stubbed)? Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.