A potential shortage of GDDR6X memory could impact the availability of popular GeForce RTX 40 Series GPUs, including the RTX 4070, starting in August 2024. The shortage is attributed to an issue with a new batch of GDDR6X memory modules, affecting GPUs using this type of memory, such as the RTX 4070, 4070 SUPER, 4070 Ti SUPER, 4080 SUPER, and RTX 4090. While the issue may be temporary, it could impact gamers seeking to purchase these high-performance graphics cards.
NVIDIA is reportedly cutting production of its RTX 40 series GPUs by up to 50% in preparation for the release of the RTX 50 series. This move is likely to lead to shortages and price hikes for the RTX 4070, RTX 4080, and RTX 4090, as retailers will struggle to keep up with demand. Meanwhile, AIB manufacturers like ASUS, GIGABYTE, MSI, and ASRock are expected to benefit from the increased prices.
Meta’s Llama 3 405B model training took 54 days on a cluster of 16,384 NVIDIA H100 AI GPUs, facing numerous hardware challenges. The team overcame these hurdles with in-house tools and efficient strategies, achieving a 90% effective training time.
The year started with excitement for new PC hardware releases, but delays, instability issues, and a potential shift in release schedules have dampened enthusiasm. AMD’s RDNA 4 GPUs are rumored to be delayed, Nvidia’s RTX 50 series might face a similar fate, and Intel’s Battlemage graphics cards are shrouded in uncertainty. Even CPU launches have been impacted, with Intel facing crashing issues and AMD delaying its Zen 5 processors.
A surprising discovery in the PCI-ID database suggests Nvidia is reviving the RTX 3050 with an Ada Lovelace chip. This unusual combination could potentially bring DLSS 3 capabilities to an older generation card, but it’s likely a move to utilize leftover Ada components.
NVIDIA is reportedly developing a modified version of its Blackwell B20 AI GPU for the Chinese market, designed to comply with US export regulations. This comes as China seeks to develop its own AI processor capabilities. While the full-fledged B20 AI GPU is too powerful for export, a toned-down version is expected to be released in Q2 2025. This move signifies NVIDIA’s commitment to the Chinese market, where it has faced challenges due to US sanctions.
Leaked information suggests that AMD’s upcoming RDNA 4 architecture will significantly improve ray tracing performance, potentially challenging Nvidia’s dominance in this area. The leak details several key enhancements, including a doubled ray tracing intersect engine and optimizations for efficiency and speed. While AMD might not be able to surpass Nvidia’s top-tier GPUs, these advancements could make RDNA 4 a formidable contender in the mainstream market, especially if priced competitively.
Igor’s Lab has revealed a concerning issue with the RTX 40 series GPUs, claiming that many models are using cheap thermal paste that degrades quickly, leading to overheating and potential failure. This could have serious implications for the GPU market and impact gamers experiencing high temperatures.
ASRock has introduced two new Radeon GPUs designed for workstation users, the ASRock Radeon RX 7900 XTX Creator 24GB and ASRock Radeon RX 7900 XT Creator 20GB. These compact, two-slot graphics cards feature blower-fan cooling and a single 12V-2×6 power connector, making them ideal for multi-card parallel computing and AI workloads. The cards offer impressive performance for 4K gaming and professional applications.
A new toolkit called SCALE allows CUDA applications to run natively on AMD GPUs, potentially changing the landscape of GPU computing. Created by Spectral Compute, SCALE bridges the gap between CUDA and AMD’s RDNA GPUs, allowing developers to write code once and run it on various hardware platforms.
