WCCF Tech claims both AMD's and NVIDIA's GPUs have been delayed due to capacity problems with the 20nm and 16nm process, respectively. AMD's next high-end GPUs will be made on the 20nm process but early capacity is being gobbled up by mobile chip makers like Apple and Qualcomm, leaving little capacity for other firms. NVIDIA on the other hand switched straight to 16nm but is facing the same issues as the limited capacity of that node is reportedly also being drained by the same mobile giants.
According to the site, AMD's 20nm GPUs have been delayed by a month or two from February/March to April/May 2015 whereas NVIDIA's 16nm GPUs will be something for 2016 and likely not even early 2016. The site also speculates NVIDIA may cancel its 16nm Maxwell plans and go straight to Pascal.
This news doesn't mean there won't be any new cards until the arrival of 16/20nm parts though. NVIDIA still has the big GM200 chip up its sleeve and AMD is expected to introduce an enhanced Hawaii GPU and a full unlocked Tonga before the arrival of the flagship Fiji XT.
Just like Qualcomm and Apple we may see Nvidia introduce 16nm Tegra mobile SOCs in late 2015. This development is interesting in the sense that it may drive Nvidia to cancel the 16nm Maxwell shrinks and go straight to Pascal instead for 2016. Either way, unsurprisingly there won’t be any 16nm GPUs until 2016. Even early 2016 is unlikely at this point based on TSMC’s own projected 16nm wafer revenue. However we don’t expect Nvidia to remain idle. The company still has one last ace to play in 2015 and that’s GM200.
If you remember back in September we had reported that the first of AMD’s 20nm R9 300 series graphics cards would arrive between mid February and early March of 2015. The delay has shifted the schedule by about two months. So the same 20nm chips would now arrive sometime between April and May of next year.