For some intrepid enthusiasts, this is the chip from Intel we’ve been waiting for. I foolishly predicted that Intel would never create an…
For some intrepid enthusiasts, this is the chip from Intel we’ve been waiting for. I foolishly predicted that Intel would never create an overclockable Core i3, because it had the potential to drive sales away from high-margin products. But here we are, Core i3-7350K in hand, and the ability to drive almost 5.0 GHz without too much trouble. This raises a couple of questions: just how close is it to the cheapest Core i5-7400 out of the box, which is only a few dollars more, and how close is it to the Core i7-2600K, a favorite chip among enthusiasts who have not yet upgraded.
For the previous generation of Intel products, we reviewed the Core i3-6100TE. This was a low-powered Core i3, but it was paired with the SuperMicro C7H170-M, a motherboard that enabled adjustment of the base frequency from 100 MHz to 130 MHz and up. This was the first chance in a number of generations to test what would essentially be an overclockable Core i3 product. There were many teething issues, as overclocking via the base frequency is less stable than by the multiplier, plus the motherboard options were not the easiest to work with.
However the final conclusion was that the Core i3 parts, while having two cores and hyperthreading, have difficulty keeping up with the Core i5 in tasks where having actual cores matter. For gaming, especially 1080p or Full-HD gaming, there was a small difference in favor of the Core i5, and the indication that with DX12 enabling multi-core gaming in a nicer way before, a gap would likely emerge over time (but mainly for high-end games for now). As a result, we postulated that Intel would never release a Core i3-K product: doing so would potentially drive sales away from the i5-K, as here is a part that could perform just as well for $60-$70 less, which users would have in pocket to upgrade a graphics card instead.
Intel, from the outside, seems very margin driven: keeping those numbers above 60-65% in every department seems to be key to progress. As a result certain projects come and go, and others with potential have a small grace period but the bottom line is performance and strategy. One could argue that potentially offering an easy route for users to migrate down from a Core i5-K to a Core i3-K could affect that bottom line: users are paying $60 less per unit, and that’s a $60 loss per sale. This is one way of looking at it – another is considering that the Core i3-K parts are dual-core by design, and cheaper to make.
Disclaimer: The data for 2+2/4+2 Kaby Lake and Skylake for the following analysis were derived from high quality images of each die, and thus subject to error.
Italicised numbers are calculated from images and represent more of a ‘peak’ die size. Data taken from delidded images varied wildly, from 105mm 2 to 130mm 2 for the 4+2, so should be taken under caution.
* Dual Core Westmere CPUs had a IGP built on 45nm
** Quad Core Westmere designs with IGP didn’t exist, so this is an estimate of what it would have been
*** Broadwell 4+2 designs don’t seem to exist. The two mobile parts (we don’t have access to) that fit this description are likely to be cut down 4+3e parts.
Taking data from available images on delidding the dual core parts and quad core parts, it’s clear that the new generation of Kaby Lake processors are physically smaller than Skylake. But the difference between the Core i5/i7 core design (known as 4+2) and the Core i3 design (known as 2+2) isn’t that much.
