Alder Lake-N is an offshoot of the series for the embedded area, but it has very little to do with the original. Because there are no big performance cores here, only the small E-Cores instead. However, these should be used efficiently, Intel promises TDPs from 7 watts for 8-core solutions with iGPU.
Whether Alder Lake-N is still an Intel Alder Lake at all is probably the first question. Because without large cores, it is no longer a hybrid design, rather the solutions are now what previous Atom processors were: A certain number of small cores, combined with an integrated graphics unit, which are marketed as Pentium and Celeron, among others.
Intel is issuing this in several variants, from the 021 version, which stands for 0 performance cores, 2 E cores and the GT1 graphics unit, to the 081 variant, i.e. still no performance cores, but eight efficiency Cores and the GT1 graphics solution. Intel also places a quad-core cluster in the middle. Like the dual-core variant, these are specified with 6 watts, the eight-core models are available with either 7 or 15 watts. A spec table in Coreboot reveals the first details, because the CPUs also have a turbo mode, the 6-watt variants then boost up to 25 watts, the 15-watt version even up to 35 watts PL2.
* VR Configurations for IA and GT domains for ADL-N SKU's. * Per doc#646929 ADL N Platform Design Guide -> Power_Map_Rev1p0 * * +—————-+———–+——-+——-+- ——–+————-+———-+ * | SKU | Settings | AC LL | DC LL | ICC MAX | TDC Current | TDC Time | * | | |(mOhms)|(mOhms)| (A) | (A) | (msec) | * +—————-+————————+——-+——-+— ——+————-+———-+ * | ADL-N 081(15W) | IA | 4.7 | 4.7 | 53 | 22 | 28000 | * + +———–+——-+——-+———+——— —-+———-+ * | | GT | 6.5 | 6.5 | 29 | 22 | 28000 | * +—————-+————————+——-+——-+— ——+————-+———-+ * | ADL-N 081(7W) | IA | 5.0 | 5.0 | 37 | 14 | 28000 | * + +———–+——-+——-+———+——— —-+———-+ * | | GT | 6.5 | 6.5 | 29 | 14 | 28000 | * +—————-+————————+——-+——-+— ——+————-+———-+ * | ADL-N 041(6W) | IA | 5.0 | 5.0 | 37 | 12 | 28000 | * + Pentium +———–+——-+——-+———+——– —–+———-+ * | | GT | 6.5 | 6.5 | 29 | 12 | 28000 | * +—————-+————————+——-+——-+— ——+————-+———-+ * | ADL-N 041(6W) | IA | 5.0 | 5.0 | 37 | 12 | 28000 | * + Celeron +———–+——-+——-+———+——– —–+———-+ * | | GT | 6.5 | 6.5 | 26 | 12 | 28000 | * +—————-+————————+——-+——-+— ——+————-+———-+ * | ADL-N 021(6W) | IA | 5.0 | 5.0 | 27 | 10 | 28000 | * + +———–+——-+——-+———+——— —-+———-+ * | | GT | 6.5 | 6.5 | 23 | 10 | 28000 | * +—————-+————————+——-+——-+— ——+————-+———-+ */
Raptor Lake brings more L2 and L3 cache
Intel is following the expected path with the Alder Lake successor Intel Raptor Lake. As in the server area, the performance cores are now equipped with 2 MB L2 cache per core, which is exactly what the Golden Cove cores used in Alder Lake are already doing in Sapphire Rapids, while in the consumer area they have been reduced to 1.25 MByte are neutered. Even with the E cores, the full expansion of the maximum possible 4 MB L2 cache for four cores has not been used so far, but only 2 MB per four cores. In the end, Intel was everywhere with the handbrake on at Alder Lake, which makes adjustments for the refresh very easy.
Now, it appears.
About p core L2 cache latency, non-accuracy test result shown in pic2. https://t.co/Gf8o2V04Pa pic.twitter.com/mLarcP9lhl— Raichu (@OneRaichu) May 18, 2022
Due to the architecture and structure of the hybrid chips, the L3 cache will also increase. Because the additional double cluster of E-Cores, 16 instead of 8 cores, always bring an additional slice of L3 cache (LLC) with a size of 3 MB per 4 cores. All in all, Raptor Lake ends up with 36 MB of L3 cache in the maximum configuration.
Commercial is now instead of 44 MB for Alder Lake-S we are talking about 68 MB of cache, just the maximum value that comes from simply adding 8 × 2 MB of L2 cache of the P cores, plus 4 × 4 MB of the E cores and the shared 36 MByte L3 cache results. The effect of the changed caches will ultimately be seen in the second half of this year, when the processors will be presented. In the end, of course, this is not possible without additional costs for Intel. Since the production in 10 nm alias “Intel 7” is identical, the CPU die is inevitably larger. The large version of the Alder Lake-S has an area of 215 mm², double E-Cores and the adapted caches everywhere will allow the CPU to grow accordingly. Since it is still unknown what else Intel will change, an approximate numerical value is still pure speculation.