Topic: VectorDash - Rent GPUs to AI researchers. $7.68 per 1080ti per day (Read 2962 times)

Hi guys,
I have about 25 GPUs configured for ML, as long as not using all of them now, I'd like to rent out to anyone doing ML, please contact me, I'm ready to give it away much cheaper than any Cloud like AWS.

Hi everyone! Just stumbled upon this topic. Could anyone tell me please how to start renting HW on this website? I filled in the form from the website... What's next? Are 2 Gb RAM per GPU strictly necessary? I've got like 5x 1080 Ti but only 8 Gb ram. Please advise. Thank you!

Most likely /var/vectordash/roms/ was saved as 0a:00.0.rom 0A:00.0.rom
So a simple
The same happened to me and I managed to fix it this way. Also it is possible that only 3/4 cards will be available for renting because some weird reason (please reply here if this it worked for you).

Did you manage to run this rig ? What is your experience ?

Hy,
I started to install the software package. I am in contact with their support, looks like they have a lots to do, try to find help here with my installation problem:

///
Running Vectordash tests. This may take a few minutes.
Testing KVM support...
Testing Libvirt virtualization support...
setlocale: No such file or directory
Testing IOMMU support...
setlocale: No such file or directory
Testing Nvidia GPU 0 on PCI 05:00.0...
Testing Nvidia GPU 1 on PCI 06:00.0...
Testing Nvidia GPU 2 on PCI 09:00.0...
Testing Nvidia GPU 3 on PCI 0a:00.0...
libvirt: Storage Driver error : Cannot access storage file '/var/vectordash/roms/0a:00.0.rom' (as uid:0, gid:0): No such file or directory
  File "virt_utils.py", line 199, in create_vm
  File "/usr/lib/python3/dist-packages/libvirt.py", line 1035, in create
    if ret == -1: raise libvirtError ('virDomainCreate() failed', dom=self)
libvirt: QEMU Driver error : Requested operation is not valid: domain is not running
Requested operation is not valid: domain is not running
libvirt: QEMU Driver error : Requested operation is not valid: domain is not running
Testing all Nvidia GPUs...
libvirt: Storage Driver error : Cannot access storage file '/var/vectordash/roms/0a:00.0.rom' (as uid:0, gid:0): No such file or directory
  File "virt_utils.py", line 199, in create_vm
  File "/usr/lib/python3/dist-packages/libvirt.py", line 1035, in create
    if ret == -1: raise libvirtError ('virDomainCreate() failed', dom=self)
libvirt: QEMU Driver error : Requested operation is not valid: domain is not running
Requested operation is not valid: domain is not running
libvirt: QEMU Driver error : Requested operation is not valid: domain is not running
Testing connection to Vectordash server...

Test results:
KVM support:                                  Passed
Libvirt virtualization support:               Passed
IOMMU Support:                                Passed
Testing Nvidia GPU 0 on PCI 05:00.0:          Failed
Testing Nvidia GPU 1 on PCI 06:00.0:          Passed
Testing Nvidia GPU 2 on PCI 09:00.0:          Passed
Testing Nvidia GPU 3 on PCI 0a:00.0:          Failed
Testing all Nvidia GPUs:                      Failed
Connection to Vectordash Server:              Passed
Not all tests have been passed. Please contact Vectordash support.

///

Any thoughts?

TIA betaminer

Hardware is
4x 1080 ti 11 GB
Asus X99-e ws USB 3.1
Xeon E5-1650 v4
64 GB Ram
2 TB SSD on M2
4 TB HD SATA
Had a chat about it with Sharif, should be the minimum hardware expectations

maybe you could check nebula ai, which is going to release end of September.
http://nebula-ai.com/

Looks like they are still working on a solution that will work on the majority of hardware:

https://www.youtube.com/watch?v=0G-27pkVL7w

Tried reaching out to them like 5 times now, no answer whatsoever if anyone gets a hold of them..let me know

Would love some more info myself..

So what happened to Vectordash? I heard about it 3 months ago and was quite excited, thought this might be a future path for my GPU mining farm.

Did the project get abandoned? Team left? Met some regulatory, legal, or technical hurdle? The last step I've got to with Vectordash is getting the technical requirement for the mining rigs through their email. I decided to see how things go first before I upgrade my mining rig CPU's from celeron to i5, for all of my hundreds of GPU's.

The Vectordash Twitter, Facebook page and official website has not been updated for months already. Is the project still developing or dead?

Do you plan to release the boards in time fo the acorn release?

I hope to be able to start verifying the performance difference between 16x and 8x on real AI workloads within a few weeks.
I'm also looking for boards with lots of well spaced slots with full bandwidth and I'm not alone.

I suggest Threadripper and PCI switches (e.g. PLX) like what was done on some X99 boards in the past.
Other important factors for AI use cases is provisioning for large amount of RAM. Typically, you want 2x the amount of memory on the GPU.
Storage is another important parameter so bandwidth has to be reserved for that. If the compute models can't be fed data from storage as fast as it's being processed, all is for nothing.

I am definitely interested in seeing what you can come up with for a mining motherboard with more PCIE lanes, for future mining needs, rendering, ai research, or whatever else.  The more options the better for all of us.

Appreciate the feedback!
The Epyc mobo should be around 100USD more than the X399 version. We are leaning towards the Epyc based platform ourself too due to the 128x PCIe lanes it has. I am also talking with the people behind the upcoming Acorn FPGAs also to get their feedback and to make sure our new motherboard will play well with the Acorns. Smiley

I'd definitely grab those boards if they were made. But I will prefer EPYC, you never know when a task may benefit from 16 rather than 8 lanes. Is it possible to estimate how much more will the EPYC mobo cost compared to threadripper? CPU wise it's about $100.

Here's a discussion on pcie lanes by OTOY users: https://www.youtube.com/watch?v=XrOBvz454fU&t=3485s

Support for pcie burification, and using the remainder lanes for m2 pcie slots may also be important for people getting the acorn fpga accelerator.

Well finally someone uses the lanes of the Threadripper.
I would also love to see one for Intel X299 as well.

Hey hey, Octominer here. We manufacture and sell specialised mining hardware for crypto miners.

We are currently working on developing our NEXT GEN riser free motherboard that will be more focused on AI and Rendering tasks also and not just mining for mining.   

I would love to hear some feedback from you guys on which kind of motherboard would you think would be perfect for Vectordash or AI/Rendering rigs? Currently we are considering making a riser free motherboard with X16 slots based on either AMD Threadripper X399 platform or AMD EPYC server CPU platform. Advantage of the threadripper is that it is cheaper and the CPU clock speeds are much higher. Advantage of the AMD EPYC is that it has 128 PCIe lanes compare to the threadrippers 64 PCIe lanes and more cores. Which makes it possible to run 7x GPUs at full 16x speed on the EPYC motherboard, while the threadripper motherboard can support a maximum of 3x GPU running at 16x full speed or 7x GPUs running at 8x speed.

I would love to hear some real life feedback on the topic of 16x vs 8x PCIe speed when it comes to rendering and AI tasks. From what I understand for OTOYs OctaneRender 8x speed is optimal and there is only a 1-2% difference in performance when comparing 8x to 16x PCIe speed. Does anybody know how is the 8x vs 16x PCIe performance difference when it comes to AI tasks and Vectordash?

Why did they not use something legit like https://boinc.berkeley.edu/ to submit their project there? Instead, they offer unrealistic gains.
"High-throughput computing with BOINC
Scientists : use BOINC to create a volunteer computing project , giving you the power of thousands of CPUs and GPUs.
Universities : use BOINC to create a Virtual Campus Supercomputing Center .
Companies : use BOINC for desktop Grid computing ."

I guess we know the answer.

I reached out to them a few times in March and April and received no response whatsoever. I have later  seen a youtube video from one guy who quoted their requirements for PCs they emailed to somepeople (I was not lucky to get this email). By the way they set a minimum requirement at one GPU per machine which means you would need  around 100 PCs alltogether for them to consider you. I though if they are serious I could make that investment provided there are any guarantees from their side that they even launch. I contacted them again saying I could meet those requirements and asked for more info. It was a month ago. Still no answer.
It does not look serious at all to me. With that type of attitude developers if there are any and their project simply do not deserve your time.

Anyone tried it out?