AMD processors currently enjoy a performance lead in gaming (something you probably wouldn’t expect back in the days of the first Ryzen processors), thanks in large part to the 3D V-Cache technology increasing the L3 cache capacity to 96MB, which accommodates a significant portion of the CPU’s give. The logical question is whether or when Intel will come up with something similar to match AMD’s lead. After all, he once had something similar in the form of Broadwell processors with eDRAM.
It turns out that Intel has such plans and will even see the implementation quite soon. Der8auer and Bens Hardware spoke to Intel Director of Technology Communications Florian Maislinger on YouTube and asked about this topic. He confirmed that Intel is also preparing processors with massive cache capacities, but it will be in a different form. And unfortunately for the field of games, this technology is not yet coming to PC processors, only to servers.
Maislinger said that Intel is preparing large L3 caches in servers, so instead of competing for X3D Ryzens, it will be processors that compete with Epics with 3D V-Cache (AMD refers to them as the “X” series). However, since Intel, unlike AMD, does not use the same silicon in desktop and server processors at the same time, this larger cache will not be transferred to Core (or Core Ultra) processors at the same time.
The first will be Clearwater Forest
The server processors with “V-Cache from Intel” should apparently be the Clearwater Forest generation Xeons planned for next year. They will be chiplet processors made up of a relatively large number of compute tiles. These should be mounted on underlying chips, but while in Lunar Lake or Arrow Lake processors the underlying tile is more or less passive and serves only as a connection layer, in Clearwater Forest processors this underlying tile will also be used to place the cache. So it will be a 3D cache in the full sense of the word. The change against AMD is that the chiplet with cache is not always one for each chiplet with CPU cores, but is common to several CPU chiplets, which should give advantages in core communication and data sharing.
It is not yet clear how many MB will be able to get into Clearwater Forest this way. Clearwater Forest is to use Intel’s 1.8nm process (Intel 18A technology) for compute tiles with CPU cores, and the chiplets will be connected using Hybrid Bonding technology. It looks like the architecture of the cores could be Skymont, so it would be an “E-Core” Xeon.
Unfortunately, there are no plans for a version of this technology in regular processors for (gaming) PCs. Or at least it’s not close enough to launch that Intel is willing to announce it. However, there is evidence that Intel at least considered implementation into PC processors. Already in the Meteor Lake processors, which also have a base tile, the possibility of integrating an extra cache into it was apparently considered, Intel patented this solution, and information about these plans or studies was leaked to the Internet, this technology apparently had the code name Adamantine Cache.
However, Meteor Lake (and neither Arrow Lake) ultimately did not pursue this option, perhaps the potential performance improvements were not enough to justify the higher production costs. But it doesn’t have to be like that forever, so one day 3D Cache may appear in the underlying tile in Intel chiplet processors. The fact that a piece of silicon is used, which must be in the processor anyway, is quite elegant about this solution, it could be an advantage against the current AMD solution, where the cache needs a completely new chiplet (which, however, is omitted in standard processors and thus makes them cheaper).
Gaming, too small a market for a dedicated CPU?
According to Maislinger, AMD’s 3D V-Cache processors are tailor-made for games aimed at a very specific group, with some compromises for other uses due to reduced clocks (however, this has been significantly mitigated with Zen 5). In the interview, it sounds a bit like the market for such specifically gaming processors is not big enough, which is not entirely logical, because given the company’s position, the potential number of customers for special gaming CPUs from Intel should always be higher than what AMD can count on. but for which the game market is obviously not so small that it is not worth paying attention to it.
But it can probably be understood that for tactical reasons Intel wants to have universal processors for PCs (and the approach where a game modification is derived from a universal processor, as AMD does, does not suit it for some reason). Perhaps it could also be more about the fact that the gaming market is not small, but there is not so much room in it for technologies that make the production of a processor more expensive and therefore require that the result can be sold at a high price, whereas in servers there is less of a problem when customers must pay extra.
Resources: Tom’s Hardware, Intel
Source: www.cnews.cz
