I think both Nvidia and AMD are doing exactly that. The way all these leaks keep changing. It's almost daily now.
Yes, only natural for them to toss out disinfo to leakers from time to time. This is the way I view them as their launch dates approach.
Siemens and NVIDIA to Build Industrial Metaverse - Digital Engineering 24/7 (digitalengineering247.com) June 29, 2022 “We have to make the digital twin as realistic as possible and bring it as close as possible to the real world,” said Busch. For Siemens customers, the design and simulation software NX is where their industrial equipment, machinery, and vehicles are conceived and defined. To breathe life into them as digital twins, the company needs an immersive virtual world where they can live. NVIDIA has already built such a world: NVIDIA Omniverse. ... NVIDIA has been promoting Omniverse as the ideal digital environment for digital twins. With built-in physics, machine learning, and real-time visuals, Omniverse can be used to simulate the electromechanical behaviors of industrial equipment. Infrastructure company Bentley is one of the early adopters of Omniverse to host digital twins. Enterprise NVIDIA Omniverse users include Amazon, Kroger, Lowe’s, and PepsiCo. Since Omniverse partnerships are not exclusive, many manufacturers will conceivably develop and maintain their own metaverses to house their digital twins. ... Developing and maintaining a digital twin is a herculean task. It involves creating a dynamic 3D replica of a real-world object in the virtual world, establishing live links to real-time data, and maintaining a history of its peaks, lulls, and breakdowns. Therefore, once brought to life inside a metaverse, transferring it to a different environment is highly unlikely. Viewed in this light, depositing a digital twin into a metaverse is a lifelong commitment, and the locked in effect is unavoidable.
In another thread I mentioned how Samsung's reduced orders for components and parts for their production of several products (TVs, smartphones, etc) is basically industry-wide and will eventually include Apple and other big chip clients of TSMC who will similarly faced a reduced need for chips. Looks like Nvidia (and AMD) as well is trying to reduce their chip orders from TSMC s well but is locked in with them currently but may have option to delay shipments. https://twitter.com/chiakokhua/status/1542732724443938816
as I suspected, even with low or no crypto demand it's gonna take until summer next year before we can get good cards at decent prices.
Just cant hide my delight in how Nvidia screwed themselves so badly with the $ billions paid upfront to TSMC to secure enough chip supply for new RTX 4000s to avoid the shortages that affected Ampere in last couple years. They can delay it, but cant cancel their contract. I really see no way out for them other than to bite the bullet on RTX 4000 pricing plans.
Their best solution is to find someone that wants those waffers. Maybe AMD can use some of those to make more Zen4.
Mindful this is still rumor, but why not keep the production schedule and just put the chips into inventory. Then use as planned or on a delayed schedule? Also Hopper could be a targeted source since they use the same process. It's not as if Nvidia was going to a new node in 2023.
The whole PC market is in a slow-down. Asus and Gigabyte expect their motherboard sales to be down by 25%. Doubtful AMD would want to commit to more CPU production than they had already planned.
Fwiw... The NVIDIA GeForce RTX 4090 will use 128 SMs of the 144 SMs for a total of 16,384 CUDA cores. The GPU will come packed with 96 MB of L2 cache and a total of 384 ROPs which is simply insane. The graphics card is expected to feature the latest TSMC 4N process node which is an optimized version of the 5nm process node and will rock some impressive clock speeds. The leaker mentions the base clock at 2235 MHz, boost clock at 2520 MHz, and the actual max boost clocks to be over 2.75 GHz which might be a confirmation of the 2.8 GHz+ clock speeds. It looks like custom models might be pushing this over the 2.9 GHz range too. That's a: 60% Increase In Base Clock (2235 MHz RTX 4090 vs 1395 MHz RTX 3090) 49% Increase In Boost Clock (2520 MHz RTX 4090 vs 1695 MHz RTX 3090) 33% Increase In Max Clocks (~2800 MHz RTX 4090 vs ~2100 MHz RTX 3090) That gives us slightly over 90 TFLOPs of compute horsepower so the rumored 100 TFLOPs figure may also be possible now with a full-fat AD102 configuration rocking similar or higher clocks... https://wccftech.com/nvidia-geforce...s-leaked-flagship-over-2-75-ghz-rumor/?beta=1