In a improvement that can shock no-one who is aware of computing historical past, it has been rumoured Intel’s next-gen desktop CPUs, codenamed Arrow Lake, will comprise an entire lot extra cache reminiscence. Like, doubtlessly a 75% improve in whole L2 cache for the highest Arrow Lake CPU. And we have seen what a fats wad of cache can do for gaming efficiency, proper?
Since introducing its hybrid core design with Alder Lake Intel has persistently elevated the extent of cache used, per core, in its processors. That is precisely the place you’d count on it to go while you stumble on the twin limits of energy and clock pace, and certainly it is the place the complete trade has gone, too.
AMD has made a reputation for itself jamming a ton of quick cache reminiscence onto each its processors and GPUs, with the X3D and Infinity Cache strategies of latest chip generations. Nvidia has been on the identical observe as properly with the Ada Lovelace structure. It’s primarily utilising the dual advantages of a brand new course of node, enabling increased clock speeds, and sticking a vastly bigger quantity of L2 cache into its GPUs to generate its gen-on-gen efficiency will increase.
Likewise, Intel is reportedly going so as to add one other 50% of L2 cache to the Efficiency cores of its upcoming Arrow Lake processors, with the older Alder Lake cores having 1.25MB, and every Raptor Lake core having 2MB. The rumours from HXL on Twitter (through Tom’s {Hardware}) are based mostly on a Bilibili message thread, and counsel that Arrow Lake will see 3MB of L2 cache per core.
If we’re anticipating the Arrow Lake chips on the Intel 20A course of node to high out on the similar eight Efficiency Core design, then you definately’ll be getting a full 24MB for these massive boi cores.
However Intel has doubled the variety of its Environment friendly cores from Alder Lake to Raptor Lake, and doubled the quantity of L2 cache at their disposal, too. Alder Lake initially had 2MB shared between every 4 core cluster of Environment friendly cores. Raptor Lake doubled that to 4MB—successfully 1MB for every core—and that allowed a chip such because the Core i9 13900K to hit 32MB for L2 cache alone.
The chance of Intel sticking to the very same Environment friendly core design for its peak Arrow Lake chip is low, and there are already rumours it may ship as many as 32 Environment friendly cores within the high desktop processors. Even when Intel did not improve the pool of L2 cache per four-core cluster, that may nonetheless imply 56MB of L2 cache for a theoretical Core i9 15900K.
| Header Cell – Column 0 | Core i9 15900K * | Core i9 13900K | Core i9 12900K |
|---|---|---|---|
| Efficiency cores | 8 | 8 | 8 |
| L2 cache per core | 3MB | 2MB | 1.25MB |
| Environment friendly cores | 32 | 16 | 8 |
| L2 cache per 4-core cluster | 4MB | 4MB | 2MB |
| Whole L2 cache | 56MB | 32MB | 14MB |
| L3 cache | 64MB | 36MB | 30MB |
| Row 6 – Cell 0 | * hypothesis | Row 6 – Cell 2 | Row 6 – Cell 3 |
All instructed, that may mark a 75% improve in L2 cache for the highest-spec Arrow Lake processor. And that is doubtlessly a conservative estimate given the possibilities of Intel opting to offer its upgraded Crestmont Environment friendly cores extra L2 cache per core are fairly excessive given the brand new 20A course of.
And with the L3 cache capacities often exceeding L2 (or what is the level?) then you definately’re taking a look at considerably extra cache all spherical. I would not be stunned to see Intel stick a shared 64MB of L3 onto these high chips.
The one obscure numbers we have seen to this point trace at a 21% efficiency improve over a Core i9 13900K for a equally specced Arrow Lake chip. And that is solely taking Intel’s inside numbers into consideration, which are not utilizing any game-centric benchmarks. With extra cores and extra cache, a Core i9 15900K may far exceed that 21% quantity.
However why is cache essential?
It is all about latency and the period of time it takes to tug knowledge from totally different ranges of reminiscence. In case your CPU can retailer all the knowledge it must course of a given process as bodily near the core as doable, the upper efficiency you will note. And that is all in regards to the measurement of the information. If it matches within the L1 cache (often the smallest and closest to the cores) then nice, if not the processor goes to the L2 cache, and if it misses that hit on the L2 cache then it goes to the slower L3 cache, after which into the comparatively glacial system reminiscence.
Learn extra: the way forward for CPUs
With extra L2 cache fewer directions have to exit to slower reminiscence swimming pools and that reinforces efficiency, most particularly in purposes similar to PC gaming. Take a look at the additional efficiency one thing just like the Ryzen 7 7800X3D delivers with its further tacked-on cache.
And, when all of the low-hanging fruit has been harvested, and clock speeds, energy calls for, and core counts cannot actually go a lot increased, that is while you stick extra cache in. It is one thing Dr Mark Dean—who was a part of the workforce who first cracked the 1GHz CPU barrier—instructed Jacob after they spoke again in 2021.
It is also one thing each engineer he not too long ago spoke to as a part of his wonderful The Way forward for CPUs collection stated—Cache is king.
