AMD RTG Project Canis Flagship Leaked: An Intel & AMD Desktop GPU Joint Venture Politics makes for strange bedfellows, because my enemy's enemy becomes my best friend! Nvidia has such a huge lead over both Intel and AMD in the area of AI that the latest rumor here strongly implies the two long time rivals have joined forces. With the exception of their prices and graphics, Intel has every advantage over AMD imaginable, but neither of them can touch Nvidia's AI and graphics. Intel is so far ahead of everyone in manufacturing processing that Nvidia can't keep up and has developed their own high speed asics to compete with Nvidia's "Tensor Cores" designed by IBM. AMD brings their infinity fabric and Navi architecture which allows them to scale multiple gpu chips, rather than relying on more expensive larger single chips. What this article does not mention, is that it would mean that Intel-AMD graphics cards could be made much more low powered and, eventually, jambed onto a 30w chip for laptops, with up to 50 teraflops of compute! On one stupid APU a few inches in diameter! All of these companies are fighting over AI research, with Nvidia merely being the first to transition most of their business into AI applications thanks to their powerhouse graphics cards. He who controls the AI controls the machines and, apparently, Intel has no intention of allowing Nvidia to dominate the market any longer. Praise Be To the Gods of Video Gaming!!! The high tech giants are about to try and undercut each other's technology while, right now, even a mere 20 teraflops of compute is more than enough for almost anyone's needs, including ray tracing, VR, and 4k gaming at extremely high frame rates. A 50 teraflop video card would be almost 40 teraflops more than the current GTX 1080 Ti and almost half that of the 110 teraflop Titan V, possibly costing in the neighborhood of $1,500.oo for 20 or more teraflops, which is the enthusiast market, but prices are cut in half every three years. That would mean PC gamers finally have some incentive not to buy a damned console, with the PS5 rumored to be coming out with 20 teraflops of performance, meaning that a PC would require double that to produce the same graphics, which is expensive on PCs today by even enthusiast standards. Assuming there is any truth to the rumor, we may hear confirmation in the summer when Nvidia starts to push their new Ampere graphics cards. These corporations always attempt to take wind out of each other's sales. AMD and Intel have already combined their cpu and gpu technologies with impressive results, but this would take it to the next level and eliminate Nvidia's huge advantage over AMD, not only being the underdog, but not having their own manufacturing capacity and technology.
Intel Is Developing A Desktop Gaming GPU To Fight Nvidia, AMD Its weird to think of Forbes as a high tech magazine and reliable source for such information, but money does all the talking worth listening to these days. Intel is famous for their classic brute force engineering approaches and "Good Enough" attitude. That attitude came back to bite them in the ass when it came to graphics cards because they are more of an analog technology. Focusing more on business needs, Intel assumed their lead in manufacturing would allow them to catch up to the competition, but ran into an unexpected limitation in silicon with their Knights Bridge gpu turning out to be too complex and an energy pig. As a result, graphics cards were the one major area where Intel could not easily stomp on their competition. How Nvidia might respond remains a mystery, but Intel's approach is all too predictable. They already own the patents on a number of fundamental technologies, including the first stochastic self-learning neuromorphic chip, and the question is how they will combine all these technologies in their graphics cards. First they will focus on optimizing the more traditional silicon transistors, and then see about integrating their other technologies and this could be the first step towards developing a completely new PC architecture based on AI.
As if to confirm that something is happening, we now have rumors that AMD Navi will merely be as powerful as a gtx 1080, and the ability to combine more than one using infinity fabric may have to wait until next year. Even the rumor mill, it seems, is going absolutely bananas which could be an indication nobody has any real idea what the fuck is going on. What seems self-evident, is Nvidia is working on releasing a new gpu that can support real time ray tracing, while Intel is working to produce their own Arctic Sound gpu, with the remaining question being, exactly what the hell is AMD doing? This summer the cheaper and faster HBM4 will be coming on the market along with GDDR6, while Vega appears to love vram, but AMD insists they will not be releasing any new consumer graphics cards based on the new 7nm process. If Nvidia continues to dominate the gpu market so thoroughly, prices will not be coming down fast and their next Ampere video cards could be the last new additions we see from them until 2020, when I expect all of this crap may suddenly hit the fan at once. The gtx 1080 architecture is phenomenal and a good base line architecture, with the newer graphics cards essentially making it scalable, so that smaller cheaper chips can be combined. However, the smaller the node or features on a chip become, the harder it has proven to tweak them for performance. But, according to Intel, it should be possible to get two cheap gpu chips up to around 20-40 teraflops.
