The newest thread on windows that jeremy did recently should be pretty mature. I tried to help a bit with it.
Keep in mind that there are some CPU with integrated graphics. You could always use the integrated graphics for display. My setup does not have integrated graphics.
I think NVME helps alot. I created this thread in the hope people would add to it with their setups. While not apples to apples, the ubuntu (on NVME) setup performed much faster than windows(SSD nand?) with all other hardware being equal. It also shows what a 1060 can do with 16GB of system RAM. Quite a bit slowerā¦
Anyone care to chime in with suggestions? I suppose the first thing would be switching the 1070 TI for 1080 TI but with the current state of GPU prices that will be hard.
The mobo/cpu combo seems a bit pricey too but I couldnāt find a solid motherboard that has 3+ x16 PCIe lanes.
Thatās an expensive mobo! NVME ssd would be a great addition. I would probably go for an AMD cpu and would not care much about the GPU lanes. I havenāt done any multi gpu training just yet but if I were assembling a box myself I would go for a single 1080TI (assuming it was within my budget) and maybe add a weaker GPU sometime down the road for experimenting while my main GPU is working on something. And the secondary one would probably be something very small / cheap.
You wonāt see 3x16 with your specād 28lane CPU, or even with the 40lane cpu. even if you could physically fit the cards in, and they ran full 16, there will still be a bottleneck between the cpu and the pci controller. you will probably be running 2x8 with that setup anyway.
Well noticed. I suppose focusing on a single higher-end GPU will be much less of a hassle, and given another 8 free PCIe lanes I could still add another one later (seems that the consensus is that 16 vs 8 lanes doesnāt make much of a difference).
Wondering if getting mobos with 2-3 SLIās would make any difference?
I would go for a single GPU and increase the available RAM. For a new desktop 16gb can be a restriction for some preprocessing / data manipulation tasks. 32 gb would be more āfuture proofā.
Hereās my DL build. Iām delighted with having a dedicated GPU with 11GB. I use the system as a cloud computer using TeamViewer and ngrok. If youāre in USA, France or UK, consider using shadow.tech instead of buying your own system. Their service is, as of now, glitchy but is arguably a better value than buying a $1500+ system.
Ok thanks for the feedback and resources. What about the os for the box and laptop? Do I want to dual boot windows and Ubuntu on the box? And then does the os for the laptop matter?
It sounds like the laptop doesnāt have to be anything special and I can just access the box remotely through ssh or Jupyter and run everything through the box. Is that right?
You dont need windows at all unless you need it for a non dl reason. Once you install ubuntu on both laptop and tower, you can use vnc to talk between them, i use tigervnc as it was easiest for me to get working on my setup, tightvnc is popular accros this forum. On the tower you run vncserver to which runs a service on port 5900:0 from memory, then you run a gui on the laptop to display a screen for the tower.
Or as you mention you can run jupyter without vnc too.
I have found rsync super handy for syncing data between two machines.
You could run a windows vnc client to view your āserverā linux desktop, so youve got 3 options there (linux/win/dual boot).
Im running linux on all my machines, pretty much only thing i miss is excel plotting (libre office ok but not as good).
Yes used, i wrote a post here (which i linked to somewhere else hidden away in this forum, please forgive the cross-post) with a bit more info on budget builds at the end of the post.
Laptop for remote access:
not sure if it needs any sort of requirements
os:
lets say I dual boot both the desktop and laptop with windows and ubuntu (could a mac with ubuntu work?)
and I can use ssh, jupyter notebook, or a vnc client to access the desktop remotely with the laptop and run all of the deep learning through the desktop
I wouldnāt put together a xeon based build for a first dl box unless youāve built some machines before, - unless you find a workstation that is ridiculously cheap. Takes a bit (or a lot) more research and less people using them, plus xeon motherboards can be non ATX form factor and may not fit in a standard case.
Iām no expert but the MB you mention above looks decent - PCI Express lanes are listed by your CPU and will be a max of 24 for Z270 chipset, The motherboard you mention then can utilize these lanes in following config.: 1 GPU at x16 PCIe 3 mode or 2 GPUās at x8 PCIe 3 mode.
Something worth doing is use pcpartpicker and look at DL builds at a reasonable price and copy the builds that have commonalities.
Use pcpartpicker to give you an indication of power consumption of the build then add 100W for safety, Iād err on the side of caution and have some spare power capacity to draw up - I think I read somewhere PSUās are more efficient at <80% load but not 100% sure on this. I just saw above you found a 650W PSU for a great price.
ECC ram may be trickier to find a MB that allows this, if you are doing research/commercial products and need <<1% RAM failure rate and really want reproducability of your experiments/stability then go for it - most home Deep Learning builds Iāve seen (me included) just use standard consumer RAM which is easier to come by and I think cheaper.
i spoke with someone from nvidia today and told them what i was trying to do. they recommended that motherboard for the 1070 and said to have minimum 500w for the PSU. pcparts says the setup is a little over 300w. nvidia also gave me this link: https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-1070/specifications
it looks like it has a lot of specifics about their gpuās and tech.
I was thinking about this while making a cup of tea and have revised, my post, mid range PSUās are relatively cheap, may as well just get one big enough so as not to worry about it. Have fun with the build.
Just a thought, but why do you need the laptop as well, will the desktop be located in a difficult/sub optimal place to work at?