Recently, the issue of the lack of video memory in many GeForce RTX 30 video cards has been very actively discussed. And the best way to check whether this is so seems to have already been found. Enthusiast Paulo Gomes took the RTX 3070, which comes with 8GB of memory, and added another 8GB to it. The modified adapter worked fine after certain settings, except that the authors took only a remake of Resident Evil 4 as a test.

In their case, the game consumed 11 GB of memory, and this is at a resolution of 2560 x 1080 pixels. As you can see, the difference is very large. It is not very correct to evaluate it by screenshots, it is better to watch the video. In the case of a regular RTX 3070, the average performance jumps a lot, but formally it is at a good level of 50-70 fps. However, the problem is that FPS 1% and FPS 0.1% are 7 fps! This means that periodically the performance drops to these values ​​for a moment, which makes the overall gaming experience ragged and uncomfortable. In the case of the modified RTX 3070, the same figures are 64 fps and 48 fps, respectively, that is, there are no problems with smoothness.

Simply put, if the RTX 3070 had 16 GB of memory, or at least 12 GB of memory, it would perform much better. The problem is that we are talking about a 3D card from just two years ago, which at one time came out at a price of $ 500, and a game that was launched only in Full HD +.

Apparently, Intel Meteor Lake processors will indeed receive a fourth-level cache for iGPUs. Previously, this information was found in a Linux patch, but now Intel has received a patent for the corresponding solution.

The patent was filed back in 2020. Intel itself calls the same L4 cache Adamantine. Apparently, it is quite important for Meteor Lake, since it got its own name.

There is no such information in the patent, but Moore’s Law is Dead says that Intel is currently testing variants of such a cache from 128 to 512 MB, although it could technically be several gigabytes.

As a reminder, the Adamantine cache will only be available for iGPUs. This, firstly, will allow the graphics core to work faster, since it will not always be necessary to access slower RAM, and secondly, it will leave the third-level cache at the disposal of the processor part.

CPU Meteor Lake should be released at the end of this year.

GeForce RTX 4090 is by far the most powerful graphics card in the world. But in some ways it is inferior to the RTX 3090 Ti. And it was this feature that allowed the CENS overclocker to set another world record, precisely using the RTX 3090 Ti.

More precisely, two RTX 3090 Ti, since that very feature is SLI support. Nvidia flagships of the past generation have it and are absent in the new one. And two EVGA RTX 3090 Ti in the hands of an enthusiast were overclocked to 2715 MHz on the core, which allowed them to score a record 36,730 points in 3DMark Port Royal.

By comparison, the best RTX 4090 score is 30,761 points, which means two RTX 3090 Tis are 20% faster.

The enthusiast also used a Core i9-13900KS overclocked to 5.8 GHz and DDR5 memory, which he overclocked to 8500 MHz.