Last week, Windows Report posted a (for now internal) presentation by Qualcomm that showcases Snapdragon X Elite processors for PCs. It will be a SoC manufactured using a 4nm process consisting of 12 Oryon processor cores running at 3.8GHz and a single- or dual-core 4.3GHz boost.
Snapdragon X Elite (Qualcomm via WindowsReport)
The integrated Adreno GPU is expected to achieve a performance of 4.6 TFLOPS (less than half of the GeForce RTX 4050). The memory controller supports a maximum of 64 GB LPDDR5x-8533, which currently means memories soldered to the board (with Samsung’s LPCAMM modules, a swappable configuration can also be considered in the future). The width of the memory bus corresponds to 128 bits (in the words of a classic desktop, two channels, although in fact it is eight narrow 16-bit channels), which means a throughput of 136 GB/s. If we take into account the available data on integrated graphics and data throughput, it may be the gaming performance side of the competition for APUs Rembrandtplus minus, so probably no extreme gaming solution in 2024.
In terms of multimedia, the equipment is about as expected: There is hardware video playback up to 4K60 in h.264, h.265 (HEVC), VP9 and AV1 formats. The chip can compress the same formats with the exception of VP9. From the video outputs, eDP v1.4b and DP v1.4 are integrated.
However, according to recent Reuters findings, Qualcomm is not the only one with an ARM SoC for PCs in the pipeline. Both Nvidia and AMD are expected to work on their own solutions. Both companies have experience with the ARM architecture, but neither has succeeded (whether they tried more or less) in PCs at the CPU level.
Grace Superchip, tj. 2× ARM SoC (Nvidia)
Nvidia entered the ARM SoC with Tegra, which with different generations targeted different segments including tablets, mobiles, laptops and cars. In notebooks, which belong to the segment of personal computers, Nvidia probably gave up first. Today, ARM architecture is trying to be applied in the segment of servers and systems for AI acceleration, where it is supposed to function as a CPU within modules with AI-accelerating GPUs. In retrospect, it is essentially still a single evolutionary line originating from Tegra.
Project SkyBridge (AMD)
AMD made several more or less unrelated attempts related to management changes in the (post) bulldozer era, when each leadership figure had their own idea, scrapped previous plans and started their own. We can mention the project SkyBridge, which was supposed to be a platform covering both x86 and ARM architectures, i.e. a board in which both x86 processors and ARM processors could be installed. The project was terminated before it was released because the market was moving in a different direction and there was not enough demand for such a solution.
AMD Seattle / Opteron A1100 (AMD)
The Opteron A1100 turned out significantly better, but with a major hitch (Seattle). Eight Cortex-A57 cores, 8MB L3 cache, a pair of 10Gbit ethernet interfaces and a memory controller supporting ECC and handling DDR3 and DDR4 with a total capacity of up to 128 GB was quite an interesting undertaking in its time. The problem was the late release. Already in the summer of 2014, AMD had samples and developer kits on customers’ desks, documentation was released, release looked imminent. But then there was complete silence for 1.5 years, and the real release happened only at the very end of 2015, when these processors faced stronger competition than they would have had in the fall of 2014.
AMD K12 (AMD)
The third major ARM project was the K12 core, tinkered with by Jim Keller during his time at AMD, while a team led by M. Clark and S. Plummer worked on x86 Zenu. But at that time, AMD did not have the finances to start the production of two product lines and it got priority Zenin other words Keller’s K12 remained in the drawer.
Why are AMD and Nvidia now seeing room for another attempt? The reason will probably be the fact that Microsoft and Qualcomm concluded an agreement in which both companies pledged that their collaboration on ARM / PC / Windows solutions would be exclusive. This exclusivity will expire next year. AMD and Nvidia can either see the emerging space for their own solutions, or they simply do not want to risk that their competitors will see this space and they will remain unprepared, without their own product.
As history shows, plans in this regard, no matter how ambitious, can turn out in any way, and from the current point of view it is impossible to say whether it is the future of the PC platform or another dead end.
2023-10-24 22:05:11
#AMD #Nvidia #responding #Qualcomm #quietly #working #ARM #processor #PCs
