Do you finally have access to the resources?
Reply
Quote
+1
____________
Reply
Quote
Do you finally have access to the resources?
AMD are being a bit weird here, this is like promising a kid a bike for christmas, but not telling him which christmas.
Reply
Quote
Personally, this whole thing is
(1) either horribly planned and logistics are executed in the worst way imaginable OR
(2) I just cannot imagine the complexity going into that project but AMD ridiculoulsy oversold their promise to help by offering access to hardware resources temporarily OR
(3) it is starting to feel rightfully so just as some huge PR campaign
I hope it is one of the first two reasons that I mentioned, bc that means that eventually they WILL get access to the hardware. - Just at a later point in time ....
The story continues ....
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
I don't really know the details, anyway, we are still waiting. The actual hardware should be provided by Penguin Computing (as explained here: https://www.amd.com/en/corporate/hpc-fund) and the infrastructure should be brand new.
Reply
Quote
The pandemic will be over by the time you get those resources lol
____________
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
(apocalyptic mode on) there will be another one (apocalyptic mode off)
The pandemic will be over by the time you get those resources lol
Reply
Quote
(apocalyptic mode on) there will be another one (apocalyptic mode off)
The pandemic will be over by the time you get those resources lol
The government already has the plans set out for the next one, this was just a test of obedience, and we failed like sheep.
Reply
Quote
I don't want to make this an annoying habit to inquire on a monthly basis whether the promised hardware can yet be accessed or not, but I am starting to question AMD's commitment at this stage. Has the project lead already exerted a bit of pressure on this topic?
... Just curious.
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
I don't want to make this an annoying habit to inquire on a monthly basis whether the promised hardware can yet be accessed or not, but I am starting to question AMD's commitment at this stage. Has the project lead already exerted a bit of pressure on this topic?
... Just curious.
Well, we actually got access to one of the promised PODs (computational nodes, as described in a previous post) at the beginning of April, it took a couple of weeks to configure the system (with the help of the guys of Penguin Computing). We are still learning how to use it the proper way (the queuing system etc.). I simply forgot to tell everybody here about this, sorry...
Reply
Quote
That’s awesome! I am glad AMD finally granted access to the promised hardware. Hopefully that’ll boost the overall computational power, albeit temporarily :)
I would love to read a short follow up once the hardware is fully operational!
Cheers mate!
Reply
Quote
Hey valterc, hope that everything is going smoothly and you could slowly transition to the new compute nodes. Are they already up and running? Will they run on a special unique infrastructure (own work queue etc.) or will they also be visible on the server page due to a rising compute average?
Don't wanna be the annoying person here always pushing this request, but it's been almost another 3 weeks. Would highly appreciate any update!
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
Hi bozz4science, we don't plan to move TN-Grid server's components to the AMD computational node, so you won't see it on the server's status page. We recently bought one server for this purpose. The AMD node will probably be stressed this September, focusing on ML techniques, when we will start classes at the University. Right now we do just small experiments on it, also using BOINC.
Reply
Quote
Thanks for providing a timely feedback. Much appreciated!
Any way to keep us in the loop with the running of the experiments until September? Would love to hear about is being experimented with in the meantime. Maybe a short update by mid-July/August and a dedicated thread about the deployment of the AMD node in September.
Reply
Quote
Any news?
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
One software technician of the University of Trento will shortly attend a dedicated training course offered by AMD about HIP/ROCm, this will eventually provide us some needed skills in order to develop a GPU application
Reply
Quote
One software technician of the University of Trento will shortly attend a dedicated training course offered by AMD about HIP/ROCm, this will eventually provide us some needed skills in order to develop a GPU application
That's an awersome news
Reply
Quote
One software technician of the University of Trento will shortly attend a dedicated training course offered by AMD about HIP/ROCm, this will eventually provide us some needed skills in order to develop a GPU application
HIP/ROCm have serious hardware limitations on amd gpus. For ex. ROCm code will work in full functionality mode only with CDNA gpus and Radeon vii. AMD doesn't guarantee stable calculations on other older AMD chips.
Reply
Quote
One software technician of the University of Trento will shortly attend a dedicated training course offered by AMD about HIP/ROCm, this will eventually provide us some needed skills in order to develop a GPU application
Any update on that application?
I'm guessing that you no longer have access to the AMD HPC. Were they useful to your research?
____________
Reply
Quote
valtercProject administrator
Project tester
Send message
Joined: 30 Oct 13
Posts: 623
Credit: 34,677,535
RAC: 3
Unfortunately no news about the GPU application, it seems that it's not an easy task, also probably not worthy...
We had received an extension from AMD so, yes, we are still using their computing resources, mainly for ML work related to single cell data (we'll like to play with this kind of data using gene@home in the future), here the challenge is to start with single cell data and try to transform it into something suitable for our PC-IM algorithm, we'll see..
Reply
Quote