There are varied causes being leveraged at this, none with substantial grounds aside from “reported from industry sources”. NVIDIA on the lookout for higher yields is without doubt one of the appointed causes, as is its historical past as a TSMC buyer. NVIDIA should not have too excessive a price porting its manufacturing to TSMC when it comes to design adjustments to the silicon degree in order to cater to completely different traits of TSMC’s 7 nm, as a result of the corporate’s GA100 GPU (Ampere for the non-consumer market) is already manufactured at TSMC. The subsequent a part of this publish is mere (comparatively knowledgeable) hypothesis, so take that with a saltier disposition than what got here earlier than.
That NVIDIA is seeking to tier its manufacturing course of throughout high-end and the remainder of its product stack (with 7 nm for high-end and eight nm for the remainder of it) would turn into a headache for themselves and for customers, ought to NVIDIA simply have two suppliers for a similar graphics merchandise. There would probably be want for some adjustments within the energy supply designs, there are a number of recent high quality assurance exams that must be taken for the brand new silicon, and NVIDIA would set itself up for authorized troubles ought to they only silently replace the manufacturing course of on high-end fashions – not solely would early adopters be understandably miffed about their product having developed over time, as there might be some claims relating to eight nm-based fashions being purchased after the 7 nm ones are launched. And if NVIDIA have been to place a sticker on retail containers updating the eight nm to 7 nm, nicely, then any person might simply decline to buy any eight nm playing cards, and solely search for the 7 nm variations, which could depart NVIDIA with an actual immovable provide drawback.
No. If this report checks out, NVIDIA will probably launch the newly produced top-end Ampere playing cards (we’re considering RTX 3090, RTX 3080 and RTX 3070) in 7 nm Super variations, taking a web page from their RTX 20-series guide. The introduction of a higher-performing product in-built 7 nm inside an entire new sequence would defend NVIDIA from authorized troubles whereas permitting them to publicly announce the transition. This would maintain early adopters “happy” within the respect that this can be a entire new product launch – customers would obtain that significantly better than by feeling that they have been beta testers for NVIDIA’s tango with Samsung, as reception for the 20-series Super playing cards exhibits. The utilization of this new course of would additionally enable NVIDIA to enhance efficiency additional over the unique 30-series playing cards, as a result of decrease leakage and better potential operational frequencies – maybe along with NVIDIA’s 20-series technique of trickling down greater GPU designs.
