Shipments of GPUs are being slowed down or suspended in light of a slowdown in demand driven by cryptocurrency miners:
Taiwan-based graphics card makers including Gigabyte Technology, Micro-Star International (MSI) and TUL are expected to see their shipments for April plunge over 40% on month, as many clients have suspended taking shipments in response to drastic slowdown in demand for cryptocurrency mining machines, according to industry sources.
Channel distributors and larger mining farm operators have cut orders with makers of mining graphics cards and mining motherboards or asked them to suspend shipments due to the crypto mining craze waning abruptly from the beginning of April, the sources said.
Quite a few mining farm operators have even stopped purchasing graphic cards, as they are awaiting the rollout of Ethereum mining machines by China's Bitmain in the third quarter of 2018. They anticipate mining rewards to pick up gradually in the third quarter, as Bitcoin and Ethereum values may rebound following sharp declines seen in early 2018, the sources indicated.
Previously: AMD GPU Supply Exhausted By Cryptocurrency Mining, AIBs Now Directly Advertising To Miners
Cryptocoin GPU Bubble?
Ethereum Mining Craze Leads to GPU Shortages
Used GPUs Flood the Market as Ethereum's Price Crashes Below $150
Cryptocurrency Mining Wipes Out Vega 64 Stock
GPU Cryptomining Hurting SETI and Other Astronomy Projects
Related: AMD Profits in Q3 2017
(Score: 3, Interesting) by fritsd on Friday April 27 2018, @06:54PM (1 child)
Well I'm interested in experimenting with Tensorflow.
Any tips as to which cheap second hand graphics cards are good for it?
And is it true that Tensorflow doesn't run on AMD Radeon GPUs? :-(
BTW I get only OpenCL segfaults nowadays on my Radeon 5870 :-( are there any new-ish versions of Mesa / OpenCL / libclc / Clang / LLVM that work?
Stupid question, I know, but it's much too complex for me to debug on my own in my free time.
(Score: 2) by tibman on Friday April 27 2018, @10:58PM
Looks like you don't even need a GPU to use TensorFlow. CPU only versions are available. Probably a good way to get started.
https://github.com/tensorflow/tensorflow [github.com]
SN won't survive on lurkers alone. Write comments.