Server for Xilinx Alveo Accelerator Cards

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

kingmouf

Member
Jun 15, 2016
44
6
8
43
Hi all,

I am trying to build a server for my university lab. The basic requirement for the server is that it has to support the installation of Xilinx Alveo accelerator cards and all the related software development and deployment tools.

I cannot build a grey box, for numerous reasons not in my hand, I need to find a server system from the likes of Dell, HPE etc. The second issue that I face is that in order to get support, it is almost a requirement to select one of the "qualified" servers, otherwise the usual answer to all issues is a typical "this is not one the supported/qualified servers for these cards".

As such, I think that the best match is the Dell PowerEdge R7525 AMD EPYC Rome server. And here comes the question on why I post here on DIY... Configuring from Dell inflates all prices to such a degree that it is simply unrealistic. As an example, buying the accelerator card as part of the configuration process from Dell, puts it in the ballpark of 11k EUR when Xilinx official distributors of said cards like AVnet, Mouser, DigiKey, price them at around 4k (note that these prices are without VAT so add another 24%...). Our budget is limited and such inflations are impossible to swallow... So I would like to know whether buying the server with a rather minimal configuration and upgrading parts like main memory, storage and accelerator cards makes sense. We have encountered difficulties with such endeavours in the past (for example Dell does not sell drive carriers without drives and buying carriers from ebay or used/refubrished is not an easy option with the bureaucracy here) and also had issues with incompatibilities or missing parts. Are there any resources that you can point to on compatible hardware, third party options, parts etc that I can use? Do you have experience with these issues? From your experience, do you think that it is a risk we should avoid? It would also be extremely useful if anyone has actually installed those accelerator cards on a Dell (or other) server and can share any potential issues.

Thank you in advance and please excuse me for all those questions!

ps. has anyone installed the active cooling version of the Alveo cards on standard workstations?