HPC YEARBOOK 2021/22


go faster. It’s now grown to $20 million. And, as of the end of July, we’re at 24 institutions that have got either on-premise support or access to cloud nodes through the fund itself. I think, more fundamentally, it was


recognition that we had a role to play. We had a unique solution to offer, and that we need to do something quickly to be able to help. We are providing hardware – either


as on-premise, where we’re giving people server nodes – or clusters, based on various levels of grants. These are all very dense computing resources, so they’re all AMD Radeon Instinct driven (AMD ‘s brand of deep learning oriented GPUs), GPU server nodes. What we did was make a kind of an


allocation – we had a certain amount that we were going to supply to folks for on-premise, donating the equipment to their campus, or the research institution, wherever they’re located. We also set up an 80-node cloud


Why did AMD create this fund? Our senior team, led by Lisa Sue, our CEO, made a commitment at that point, recognising that here we are in a pandemic, we’re a global or multinational company where there’s a responsibility for us to help address it. We had a unique solution to offer; as a technology leader within the data centre, we had the tools and engineering talent that could really make a difference. And so they made a commitment at that time on April 2020, of $15 million worth of HPC equipment. We have these resources available, we want to accelerate research at some of the leading institutions that are already looking at addressing the pandemic. We wanted to know how we can help them


www.scientific-computing.com


with Penguin, and with support from DataBank that hosts it. And so, in that case, with the cloud, we can offer node access to institutions for 12 months. You can log in remotely from wherever your institution is, and have access to those same resources, dense computing resources. There are about over 250 server nodes donated as part of the programme, 80 of those were set up as a cloud. Some institutions got only on- premise support, some got a mixture of on-premise and cloud and some got cloud access only. But the $20 million, just to make a fine point, doesn’t include any of the engineering or support, that is just addressing the physical hardware that we donated.


Why did you choose to focus on GPU accelerated computing? There were a couple of instances where the institution’s specific workloads weren’t ported or optimised for GPUs yet.


But that’s, I think, maybe three donations out of the total, to where they requested CPU-only servers. The focus here was really trying to give GPU accelerated computing devices to these institutions. That is more optimised for AI and machine learning workloads, which is where we thought we can really make an impact on research. There is a diversity of workloads out


there, but the ones we’re really primarily supporting now are two camps, one


available, we want to accelerate


research at some of the leading institutions that are already looking at


can we help them go faster?


“


is machine learning and the other is HPC applications. So you have things like large-scale molecular dynamics applications some of those workloads that are now being used to look at spread modelling and variant tracing. We meet with these institutions


monthly, they’re coming to us with other kinds of workloads that they want to support.


And then we’re looking at how we


can prioritise engineering support to help them to accelerate their workloads. Instead of taking weeks or months, we’re down to days or hours.


Did AMD create and manage this fund alone or did you have partners? We had great partners, especially Penguin Computing, which built all the devices and distributed them. There was also Mellanox, which is now, as you know, Nvidia networking. Gigabyte also made a lot of concessions to make sure we could get this equipment out as fast as possible. And then Databank, which hosts the cloud service itself, has donated the cost of electricity to be able to run these high powered computing servers. So it definitely was a partnership that made this happen. There’s no way a global pandemic


can be addressed without real data science solutions being applied to it. This is an unprecedented time and


that required an unprecedented kind of response from us. l


21


addressing the pandemic, how


We have these resources


“


RaevskyLab/shutterstock


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42