Video cards with a small amount of memory can sometimes still perform well in modern games, if we are talking about the flagship solutions of yesteryear. However, now there is at least one game where accelerators with less than 4 GB of memory simply will not work. This is Halo Infinite.

A fresh patch for this game now simply does not allow you to run it on a PC with video cards that have less than 4 GB of memory. At the same time, such changes are not mentioned in the description of the patch.

It can be noted that the requirement in the form of at least 4 GB of memory has always been in the game, but usually such requirements are conditional. In particular, Halo Infinite ran on inappropriate 3D maps prior to the patch. For example, there are a lot of videos on the Web where the game is tested, including on a GeForce GTX 1050 with 2 GB of memory, which is quite enough to play even in Full HD with minimal graphics quality settings. But there is a much more powerful GeForce GTX 1060 3GB, which feels quite tolerable even in more demanding games.

There is a possibility that the restriction is just a mistake and will be removed soon. But not at all a fact.

Nvidia introduced the H100 NVL accelerator, which, apparently, was created specifically for training language models. Servers equipped with H100 NVL GPUs deliver up to 12x the performance of the GPT-175B over DGX A100 systems while maintaining low latency in power-constrained data center environments, according to Nvidia’s website.

In fact, the H100 NVL is a pair of H100s on the same board. We are talking about two GPUs connected by an NVLink interface, thanks to which Nvidia says 68 TFLOPS in double precision calculations (FP64), 143 TFLOPS (Tensor Core FP64), 7916 TOPS (IN8), 188 GB of memory with a bandwidth of 7.8 TB /With. All this allows you to process up to 175 billion ChatGPT parameters in real time.

Nvidia also announced the launch of inference platforms for large language models and generative AI workloads. We are talking about four inference platforms at once, optimized for a diverse set of rapidly developing applications of generative AI.

Each platform combines a full stack of inference software, the company’s latest GPUs, from gaming Ada Lovelace to dedicated Hoppers, and the Grace Hopper Superchip.

Mastercard has announced a partnership with Australian stablecoin platform Stables to provide USDC payments in the APAC (Asia Pacific) region.

The new Mastercard will be connected to the Stables crypto wallet, which only supports stablecoins. Card users will be able to pay at any outlets that support Mastercard cards. Funds will be debited from the balance of the Stables crypto wallet with automatic conversion to USDC.

“Mastercard is committed to delivering innovative payment solutions that give cardholders the freedom to spend their assets where, how and when they want. Stables is building a solution for the Web3 sector using Mastercard’s global payment network. The platform also uses various analytical tools, including CipherTrace and Ekata services to ensure trust and security”, ? said the head of Mastercard Australia on fintech Callan Hogan (Kallan Hogan).

Stables co-founder and CEO Erez Rachamim said the initiative was driven by the need for a “simple and affordable means” to leverage digital assets. In addition, these efforts are accompanied by an increased desire of digital asset owners to use a debit card to spend digital currencies. In a global survey by Marqeta, 82% of participants showed interest in such a solution. In addition, according to the survey, 55% of clients in Asia Pacific (APAC) are more comfortable investing in digital assets backed by reliable companies.

Using the Stables wallet, you can also place deposits in other stablecoins such as Tether (USDT) and Binance USD (BUSD). However, all deposits are automatically converted into USDC without commission. The integration is expected to go live in June, in the second quarter of 2023. Mastercard payments in USDC will be available in Australia first. However, later this payment method will be available in Europe, the US, the UK and most of the Asia-Pacific region.