The history between these three companies goes back to two years ago, most of them either unaware of what we have uncovered based on the rumors we have concluded to be true or may be true according to those who have done so. posted on the Internet.
As we all know, NVIDIA was supposed to introduce its 12nm Turing from TSMC, while AMD finally introduced Navi has RDNA at 7nm, and formerly the Vega 20 with the Radeon VII at the same nanometers. But this was not originally planned by the elders … What happened next?
The story between NVIDIA and AMD with TSMC, or how bad the move is to market share
In 2018, and just a few months before the Train launched as construction and GPUs market, NVIDIA has been searching for power with its Samsung production facilities, seeking to break its fruitful relationship with TSMC for its new graphics cards.
The AMD's left compartment allows direct drive of its rivals: ensuring a large 7nm lifer protected by its CPUs and GPUs. After NVIDIA's crackdown between Koreans and Taiwanese, the company decided to come back with the latter, but as we said at the time, Apple and AMD had the largest number of orders that TSMC could make to companies on a monthly basis.
NVIDIA had no choice but, seeing the inaction of Samsung and its 7nm and further delays, they opted for the 12nm TSMC for Turing as an escape route. AMD has discovered two things about this: better prices at TSMC which focus on the company on the other hand have produced the NVIDIA version based on digital lithographic technology.
In recent months, AMD has been able to regain market share for NVIDIA and Navi, so it makes sense that Huang intends to launch Ampere as soon as possible to turn the tide.
Samsung was given a low price and a change of price at 5 nm
History will repeat itself with Ampere, even though it has some unique nuances. NVIDIA has tried again with Samsung and was behind TSMC again. The logical move is similar to finding the best prices recently because Samsung is competing in this category and not so much in direct node performance, not in frequency or usage.
In this regard the addition of the uncertainty that has occurred in Asia since December with COVID-19, has resulted in Ampere being delayed for several months and with this product having a reasonable delay. NVIDIA's move was therefore laid out late last year and in this case they are fast and smart.
As we announced two months ago, NVIDIA will use the 7 nm EUV node at the high end (not the original N7 TSMC or N7 +), while medium and low quality will use Samsung & # 39; s 7nm or 8nm as they are competitive pricing. At present we do not know exactly what the end-reference document means, GA100 and GA102? Or just the first one?
Besides, today it has been confirmed that NVIDIA has just booked a total of 5nm GPU orders, where rumors say it will be owned by TSMC and in view of its Hooper design. It is clear that they do not want to make the same mistake as Turing and aim to get ahead of AMD with it RDNA 3. The question will be that by 2022 Samsung will too 5nm EUV they will also become the money and pressure to force the TSMC to lower prices, we will see …
