TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

GVM: A GPU Virtual Machine for IOMMU-Capable Computers

102 pointsby ArcVRArthuralmost 3 years ago

7 comments

evol262almost 3 years ago
This summary page is absolutely terrible.<p>It appears to be an open-source implementation of nvidia-cli for managing mdevs, which is arbitrarily nice, but it&#x27;s not clear to end-users what this means.<p>Pre-Ampere GPUs (Ampere+ is MIG) were able to use the mediated device subsystem to partition cards into time slices. It&#x27;s similar to SR-IOV, except that you can specify the size of the partition with more granularity than &quot;give me a new virtual device&quot;.<p>Intel has GVT-g, which is reasonably widely supported on Gen12 and above (edit: too late&#x2F;early -- Gen11 and earlier, not Gen12 and later -- thanks to my123 for the correction). nVidia had vGPU&#x2F;mdev, and newer generations (Ampere and later) use MIG. It&#x27;s unclear whether this supports MIG at all. AMD uses MxGPU, and they&#x27;ve never really cared about&#x2F;pursued anything related to this, probably because their datacenter penetration is about 1%.<p>MxGPU is only supported on some FirePro cards. mdev was largely on GRID cards (mostly Tesla, some Quadros). MIG is on AXXX cards.<p>It&#x27;s unclear why anyone should use this over mdevctl, which already supports GVT-g, and it&#x27;s also unclear whether this is tied to the (very much &quot;don&#x27;t use in production&quot;) open source nvidia drivers.<p>For end-users, GVT-g, getting a cheap older GRID card, or using Looking Glass for your GPU are all more reasonable options.<p>This effort is great, but the readme is appallingly short on information even for someone who knows the problem domain.
评论 #31997817 未加载
评论 #31999124 未加载
评论 #32001361 未加载
评论 #31998880 未加载
评论 #31997784 未加载
评论 #31998448 未加载
gigatexalalmost 3 years ago
Can I use this to create a kvm vm with part of my GPU to run 3D accelerated things?
评论 #31999308 未加载
ConstantVigilalmost 3 years ago
Might I kindly ask where the plan for AMD support comes in, if at all?
评论 #31999279 未加载
metadatalmost 3 years ago
Can this run a qemu Windows x86 VM?
评论 #31999312 未加载
shmerlalmost 3 years ago
What limits it to Nvidia? And is it using SR-IOV?
评论 #31999350 未加载
评论 #31997767 未加载
rob_calmost 3 years ago
no AMD, why for Intel and doesn&#x27;t cover 90% of nvidia products most people will have (unles they own and run a datacenter)... which leads me to ask...<p>why even bother?
评论 #32003960 未加载
ncmncmalmost 3 years ago
This seems to be at the wrong level of abstraction.<p>I want a virtual Vulkan, one per CPU VM. I think I read somebody was working on that, or something like. That way it works on any GPU, not just NVidia or NVidia&#x2F;Intel.
评论 #31997822 未加载