As soon as we managed to touch on the topic of AMD’s announcement at Computex 2022 a few days later, data appeared on the Web that the company could show us.

Unfortunately, this data does not apply to processors or video cards. It’s about chipsets. Sources say AMD will unveil at least three next-generation chipsets for the AM5 platform during its event. These will be the X670E, X670 and B650 chipsets. Judging by the names, we should also expect the A620, but there is no data yet. It is likely that such a chipset will simply be released later.

Let’s start with the older solution. AMD X670E will be the top chipset, and the letter E in the name stands for Extreme. It won’t differ from the X670 in terms of features and capabilities, but motherboards based on this chipset should provide PCIe 5.0 connectivity to both the GPU and the M.2 slot, or possibly multiple slots, while motherboards X670-based ones may use PCIe 4.0 instead. Whether there will be any other differences remains to be seen.

But the source confirms early reports that the X670 (respectively, the X670E) will be a dual-chip chipset with two identical chipsets called Promontory 21. And the B650 chipset will be the same chip.

The source had all the details about the new Nvidia 3D card, but not the long-awaited GeForce RTX 40 line, but the old GTX 16 series. Now that the cost of video cards has dropped to almost the recommended level, the premiere of the low-powered GeForce GTX 1630 looks strange. However, Nvidia certainly has its own ideas in this regard.

The GeForce GTX 1630 will be based on the old 12nm TU117 GPU in the TU117-150 version. According to the Techpowerup database, this will be the twenty-fifth version of the TU117, if you count the various versions of 3D accelerators for desktop PCs, consumer laptops and mobile workstations.

The TU117-150 will have 512 CUDA cores, the GPU will be paired with 4GB of 12GHz GDDR6 memory via a 64-bit bus. The frequency of the GPU itself will be 1.8 GHz. Due to the truncated bus, the memory bandwidth will be 96 GB / s – even less than a regular GeForce GTX 1650 with GDDR5 memory. At the same time, the TDP of the novelty is still the same 75 watts.

The official premiere of the GeForce GTX 1630 is expected on May 31st. It will probably take place as part of Nvidia’s press event, timed to coincide with Computex 2022. Perhaps the company will announce something more interesting that day. For example, GeForce RTX 40 3D cards. The latest data suggests that they may come out earlier than expected.

Despite the fact that Intel Raptor Lake processors should be released only at the very end of the year, they are increasingly appearing in various benchmarks. Today, the top-end Core i9-13900K lit up in the tests.

The software data confirms early rumors that the new CPUs will have much more cache memory. In particular, the Core i9-13900K has a total (L2 and L3) of 68 MB of cache. Apparently, we are talking about 36 MB of L3 cache and 32 MB of L2 cache. To be more precise, only large cores will have L3 cache, and small ones will also have L2.

Presumably, each large Raptor Cove core will have 2MB of L2 cache and 3MB of L3 cache. And each cluster of small cores will have 4 MB and 3 MB, respectively. The new CPUs will have eight large cores, and 16 small ones if we are talking about top models. As a result, Raptor Lake will have 55% more cache than Alder Lake.

However, this will still be comparable only to the Ryzen 5000, which came out a year and a half ago, and Intel processors will still be far from models with 3D V-Cache.