Intel ready to say goodbye to Hyper-Threading with Core Ultra 200?

Before the end of the year Intel will present the processors Arrow Lake for Desktop PCalso introducing into that market sector a disaggregated chiplet-based design (or tile), similarly to what was done with the Core Ultra 100 “Meteor Lake” dedicated mainly to the mobile sector.

The Core Ultra 200, this is the name that the new chips should have, will therefore represent a new starting point for Intel’s desktop solutions: based on new architectures (Lion Cove for P-Core and Skymont for E-Core) and more advanced production processes (Intel 20A for Compute Tile and other recent TSMC processes), Core Ultra 200 solutions will require a new motherboard LGA 1851 socket. Arrow Lake CPUs will also be Intel’s first desktop series to integrate an NPU.

Another new element should be represented by the absence of Hyper-Threading technologyor Intel’s definition of Simultaneous Multithreading (SMT) which allows multiple threads to use the same core, making more efficient use of resources: in the desktop context we have seen CPUs with a maximum of 2 threads per core.

Based on several leaks in recent months, https://twitter.com/momomo_us/status/1779159569933258760, It looks like the Core Ultra 200 won’t support Hyper-Threading, therefore all appearing in the “1 core, 1 thread” configuration. Arrow Lake processors should offer up to 24 cores, consisting of 8 P-cores and 16 E-cores. Two driver logs linked to Linux systems seem to confirm the rumors once again.

In the case of the 20-core model, the configuration should include 8 P-cores and 12 E-cores, mirroring the configurations of the existing Core i9-14900K and Core i7-14700K with the 24 core: the main difference with current CPUs , architecture and design aside, it seems to all intents and purposes to be the absence of Hyper-Threading, which, moreover, in the Cores from the 12th to the 14th generation only affected the P-cores due to the different architectures used.

The decision to abandon HT could be linked on one hand to security reasons (numerous attacks on CPUs take advantage of the technology), on the other hand role of E-cores in multi-threaded loads, capable of delivering high performance with efficiency. At the same time, it is good to remember that in the past there has been talk of the arrival, but only in a few generations, of a new solution called Rentable Units which should improve thread management.

We will have to wait for the release of the processors – in the second half of the year – to understand how the increasingly probable absence of Hyper-Threading will impact performance. We should have some more information on this at Computex 2024 at the beginning of June.