Is anyone doing real machine learning work with AMD GPUs?32gb of ram mean ML types will happily pay 6 figures for this thing.
Is anyone doing real machine learning work with AMD GPUs?
Somebody is buying them at least.
https://www.nasdaq.com/aspx/call-tr...u-on-q4-2017-results-earnings-call-transcript
"Our professional graphics business, had its best quarter ever, based on growing data center sales, highlighted by strong Radeon Instinct, MI25 sales to a major cloud provider."
And there was this.
https://www.amd.com/en-us/press-releases/Pages/amd-and-baidu-2017aug28.aspx
They bought Radeon Instinct for A.I. applications:But are they buying them for machine learning applications, or other kinds of HPC?
Baidu said:We are delighted to work with AMD to introduce AMD Radeon Instinct products into our datacenter and AI program, to build a more flexible and powerful computing platform that empowers AI products and accelerates the development of the global industry.
Yeah it is a bit academic until one knows the scale of the implementation and sales in actual revenue or some kind of figure, most will do small node implementation as part of their R&D alignment just like several other cloud/AI/hyperscale companies before deciding on committing to large scale services/nodes.But are they buying them for machine learning applications, or other kinds of HPC?
Also, AMD gets their best professional graphics quarter almost every quarter, it seems, but it never appears to amount to much in the end. I sometimes get the feeling that they sold one card during the quarter when they created this business unit, two in the second, three in the third, etc.
Vega as in Vega FE/64/56/MI25 or Vega as in Vega FE/64/56/MI25 +Ryzen 3 2200G with Radeon Vega Graphics +Ryzen 5 2400G with Radeon Vega Graphics +Intel Core ix-8xxxG Processor with Radeon RX Vega M GH/GL graphics
AMD later, at its post event round-table, clarified that it was referring to "Fiji," or the chip that went into building the Radeon R9 Fury X, R9 Nano, etc., and comparing its sales with that of products based on the "Vega 10" silicon.
Thanks, should've read this myself in the first place.According to the person who wrote the article, it's Vega 10 > 10x Fiji.
Growth in shipments of "Vega" based graphics cards is triggered by the crypto-mining industry, and for all intents and purposes, AMD considers the "Vega 10" silicon to be a commercial success.
Lisa Su said:The high end Vega GPU that we launched in August of last year, if you look at just August until now we have over 10x the number of gamers that are using Vega vs our previous generation in GPU.
Over 10x, and we are just starting with Vega overall.
In what way?That seems contradictory as to what's reported by techpowerup.com. I'm confused.
In the way, that Su said (shortened): „10x Vegas in the hands of gamers“ while the other guy at the roundtable _reportedly_ said „Growth in shipments of "Vega" based graphics cards is triggered by the crypto-mining industry“. That's two different audiences in my books. Or is my understanding of english off again? It's kinda hot in the office, so I would not rule that out.
Obviously financially for AMD it's still a success I guess.
My thinking as well. Steam may have its flaws but did Fiji ever show up in the top 20 of GPUs? From a volume and profit point of view, Fiji was probably one of their biggest flops in years.I just don't think 10x Fiji is really that impressive given the limited sales and production of Fiji and the large amount of Vega sales attributed to mining. At least from a gamer perspective on the success of the product. Obviously financially for AMD it's still a success I guess.