One of the most respected processor experts around is David Kanter, who has published numerous deep technical analyses of computer chips on his website Real World Tech as well as for the respected trade publication Microprocessor Report.
On Twitter, Kanter -- in a discussion with another user -- mentioned that he's "spent a while figuring out" why chipmaker Intel (INTC +4.67%) hasn't yet built a processor for desktop gaming computers that's essentially the company's current flagship Core i7-7700K gaming chip paired with Intel's custom-designed eDRAM memory, a technology which has been shown to boost chip performance across many workloads.
Image source: Intel.
Although only Intel knows the real reason that such a product has yet to appear on store shelves (and why it hasn't appeared on any leaked Intel processor roadmaps), I'd like to offer a potential explanation.
Such a chip would be great, but...
There's no doubt that such a chip would be quite good for gaming, as the 7700K is already a decent gaming processor and the addition of eDRAM would probably help to push performance up further.
Indeed, The Tech Report's testing of the Core i7 5775C gaming enthusiast chip -- which included eDRAM because it was essentially a repackaged high-performance notebook chip for desktops -- showed that the eDRAM helped improve CPU-related performance in several games significantly..
It's not all gravy. The integration of the eDRAM chip into the same package as the processor would almost certainly drive production costs up. The eDRAM chip (which would likely be the size of a mid-range smartphone applications processor) has a non-trivial manufacturing cost associated with it. And the added complexity of putting the eDRAM chip right next to the main processor could push up chip assembly and test costs.
However, for the purposes of this argument, let's suppose that a reasonable portion of the consumers interested in Core i7-7700K chips could be upsold to a higher-priced variant with the eDRAM built in -- in effect, a product that's gross-margin-percentage neutral at a higher average selling price.
Even viewed through that lens, it's not hard to see why Intel might not want to productize such a part.
How is eDRAM used in Intel's product lineup today?
Today, Intel sells low-power notebook chips with eDRAM integrated right alongside the main processing engine. The eDRAM is only available on versions of those notebook chips that include the company's Iris Plus graphics -- unsurprising, since the main purpose of the eDRAM is to alleviate potential memory bandwidth bottlenecks that would limit the performance of the larger and more expensive graphics engine integrated onto the chip.
In other words, the inclusion of eDRAM is a necessary cost to allow the value proposition of the higher-end Iris Plus graphics engine to shine through.
The high-end low-power notebook market is clearly smaller than the mass-market, lower-cost notebook market, but it's not hard to find examples of high-volume computers that incorporate these eDRAM-packing chips: the 13-inch MacBook Pro, the Core i7 variant of the Surface Pro 4, higher-end Dell XPS 13 models, and more.
Finding anywhere close to that kind of volume would be very tricky for a desktop chip with eDRAM but standard performance graphics.
Where would such a chip go?
It's important to remember that Intel's Core i7 7700K enthusiast desktop chip (as well as the 7600K and 7350K enthusiast chips) are based on the same silicon as the company's mass-market desktop and high-performance notebook chips. The only differences are that the chips are essentially the cream of the crop (they run at higher speeds out of the box than their non-enthusiast counterparts do) and are "multiplier unlocked," which means that users can push the chips to higher speeds than what Intel rates them at.
In other words, these enthusiast desktop products are low-risk, since the chips will be designed and built anyway, and high-return, since gamers and enthusiasts are willing to pay premiums for higher speeds and the unlocked multipliers.
But if Intel were to do a separate SKU that married the same silicon that powers the 7700K (and the plethora of related desktop and notebook chips), then it'd need a large enough market to sell the chips into to make the incremental product-development and inventory-management risk worthwhile.
The problem is, it'd be hard to find markets into which Intel could sell such a product and get paid a premium for it, especially since the eDRAM in this case would be pitched as helping to deliver a modest boost in CPU performance, rather than as fundamentally unlocking the capabilities of the graphics processor sitting there in the chip.
It would essentially be used in chips targeting higher-end gaming desktops and gaming laptops -- that's it. Those are reasonably sized and growing markets, but Intel probably doesn't think they're large enough to justify totally new chip designs.
