In 2018, Nvidia finally launched their Turing range of graphics cards to the market, and given the release of the Pascal powered GeForce 10 series had been almost 2 years prior, people were extremely eager to see what Nvidia had been working on for the GeForce 20 series of GPUs. We now know that the answer is mostly Ray Tracing and RTX technology, with much of the other aspects of the GPU being built on the fundamentals of the companies Volta architecture.
This is just as true when we’re discussing the process that Turing was built on, as Nvidia opted to use the same TSMC produced 12nm FFN nodes used in Volta. Reduced leakages and other slight tweaks aside, it wasn’t exactly a revolution, and certainly, means (at least in the manufacturing process) will put Nvidia firmly behind AMD’s 7nm process produced by TSMC during 2019.
But, what of future architectures? Well according to the Japanese website, MyNavi, their sources claim that Nvidia will be a launch partner (along with IBM) of Samsung’s 7nm Process. For all intents and purposes, there are a few subtle differences between the two processes, but they are close enough to one another that it will largely be the actual architecture of the processor which will be the deciding factor on performance.
So then, Nvidia’s Turing architecture had been originally rumored to launch on the 7nm process of TSMC, but Nvidia clearly wasn’t happy with yields (likely because of the complex nature of the Turing) and opted to wait until the process was more mature (at least, that’s the rumor).
*UPDATE* – a supposed list of the enhancements fo ampere has leaked
What this means is that for 2019, AMD could (at least in theory) have the run of the place when it comes to the 7nm process GPUs… unless Nvidia opt to release a series of cards on 7nm for say the professional market (such as a new version of Volta, with the consumer GeForce using the 7nm of Samsung, which would be a little odd… but I suppose it is possible).
The other possibility (if we’re just spitballing) is Nvidia could launch a tweaked version of Turing (or another architecture) on the same 12nm process – but the issue there is that die sizes would start getting rather large.
So once again, assuming this report is accurate, the most likely scenario here is we’ll see GeForce 30 launch at some point in 2020, which means AMD have a great opportunity to pressure Nvidia on the pricing front with Vega and Navi graphics chips throughout the remainder of 2019.
2020 will be interesting for both Nvidia and AMD because we’ll finally see a third player in the market (well, at least when it comes to discrete graphics solutions) with Intel getting in on the act with the Intel Xe (the GPU formally known as Arctic Sound). It will also mean that Nvidia will be the only one of the two companies who doesn’t produce X86 CPUs, and this means that Intel and AMD could possibly launch a rather powerful series of APUs which could eat into a lot of the lower end market (think the current GeForce 1030).
In the short term, CES in going to take plus in just a few days, where we can expect a lot more details on AMD’s plans for graphics solutions. The market is very much crying out for competition in the 200 to 300 USD space, given that the cards currently sitting there are pretty old. After all, while the RX 580 (and its ilk) are no slouch, their underlying architecture is pretty old now, with Polaris debuting back in 2016 with the RX 400 series.
Other rumors tell us that the Vega VII could launch even for gamers, although this is far from confirmed. We can say that from reports Navi will be likely supporting PCIe 4.0, or at the very least multiple reports tell us that the X570 chipset (which is for the Ryzen 3000 Matisse AM4 platform) supports PCIe 4.0, and reports tell us that AMD wants to be the first to the market with both a chipset and graphics solution which is PCIe 4.0.
*UPDATE* – a supposed list of the enhancements fo ampere has leaked
