As we know from previous leaks, Intel plans a series of architectures in just 5 years that will leave everything behind in record time. And is that the company has pending work to such an extent that it has no choice but to push and catch up to compete against Apple and AMD. What is filtered today is related to its architecture Arrow lake, which is expected by the end of 2023 and where it will compete against the bet of those of Cupertino in low consumption and at the same time with Zen 5 of AMD.
The fifteenth generation Core Arrow Lake and its iGPU will be very powerful
The information that is handled today affirms that Intel will have in its hands a new micro architecture and also a completely new general architecture, but based on big.LITTLE once again (Alder Lake will be the first in a few months). The code names are Lion cove (High Performance Cores) and Skymont (Cores for efficiency).
What must be clear from here is that at all times we are going to talk about CPUs for laptops, where leaks for desktop are more rare, but there is some brushstroke. Therefore, we talk about Arrow Lake-P (and not S) which would come with surprise from the hand of TSMC at 3 nm.
Yes, no Intel 3 or similar process – TSMC would make these processors on its 3nm node with EORV (possibly High-EUV). What can we expect with all this? The filtration ensures that it will have at most a configuration of 6 high-performance cores and 8 of efficiency, which gives us a total count of 14 cores and 20 threads. Today it seems quite a lot for laptop CPUs, but the reality is that it will not be far from Alder Lake-P and Raptor Lake-P, so at this point Intel does not seem surprising.
If you don’t increase the number of cores and threads, where is the improvement?
Well, curiously in the architectures, in the reduction of the transistors by 3 nm and especially by a new generation of iGPUs based on Xe, where it is specified that they would arrive as GT3. Here the surprise is of such a caliber that it is convenient to take a little air and open your mind for the future, since it is specified that in its maximum configuration they will have up to 320 EU, which means obtaining nothing less than 2560 Shaders.
To put us with this data: this would mean more real Shaders than a RTX 3060 Ti current. If we take into account the frequency and efficiency improvements of the node and given the current frequency trend in iGPUs, it is more than likely that we are seeing iGPUs hovering around the 2000 MHz, almost nothing.
The filtration concludes with a key fact, and that is that these new Xe GT3 would have a total area in the chip of 80 mm² and that another auxiliary called ADM would also be included next to said die, where it is speculated that it would be a kind of exchange cache between Cores and iGPU in the purest style AMD V-Cache.