Breaking News

Leaked Benchmarks Suggest Intel Will Drop Hyper-Threading from Core i7

For most of the past decade, Intel has followed a fairly steady set of rules when speccing out its Core i3, i5, and i7 processors. Until last year, Core i3 chips were dual-cores with Hyper-Threading enabled, Core i5 CPUs lacked Hyper-Threading (aka Symmetric Multi-Threading), and Core i7 CPUs were quad-cores with HT enabled. Last year, Intel changed things up, with the Core i3 family bumping to quad cores (no HT), the Core i5 jumped to six cores (also no HT), and the Core i7 retained six cores with Hyper-Threading. Only now, if leaked benchmarks are accurate, HT is going away from the i7 as well.

The data comes from a set of Core i7-9700K results that popped up in SiSoft Sandra and shows the chips as having 8 CPU cores and threads, with no Hyper-Threading in sight. The peak turbo frequency would be 4.9GHz, and that we’d see a Core i9-9900K launch at the same time, Ars Technica reports, with a full eight cores and 16 threads.

Original story by WCCFTech

Let’s assume that Intel’s Hyper-Threading is worth, on average, about 1.2x additional performance. On a six-core CPU, you’d expect HT to add roughly the equivalent of 1.2x “real” CPU cores. At first glance, this doesn’t seem like much of an improvement over the Core i7-8700K — an eight-core chip without HT and a 4.9GHz peak turbo may only be slightly faster than an 8700K. Stripping HT may explain the smaller L3 cache relative to core counts — the Core i5 chips that lack HT offer 6 cores and 9MB of L3, while the Core i7-8700KSEEAMAZON_ET_135 See Amazon ET commerce fields 12MB. The new 9700K supposedly uses 12MB as well, despite adding two additional cores.

Assuming, as always, that these rumors are true, I think we can intuit a bit about what Intel’s strategy with this move would be. It’s not clear how much faster these new Core i7’s would be than the 8700K, but let’s assume Intel can eke out, say, 10 percent more performance from the higher core count, eight physical cores, and slight clock speed kick. That should put the 9700K on par with the 2700X again, if not ahead of it, and priced at roughly $350.

Now, Intel rolls up with the Core i9-9900K — an eight-core / 16-thread HT-enabled CPU that’s a further step faster than the Core i7-9700K. It’ll beat the 2700X in single-thread and multi-threaded code, having matched it core-for-core. It’ll be priced to match, at $450.
AMD and Intel Are Playing CPU Chicken

At this point, both AMD and Intel are basically daring the other to do something drastic to their high-end desktop CPUSEEAMAZON_ET_135 See Amazon ET commerce product lines. Last year, Intel blinked first with the short-lived Kaby Lake X platform that had no reason to exist and vanished into the ether in short order. AMD’s Threadripper has pounded on Intel’s HEDT pricing by offering far more cores for less cash; Ryzen’s best relative performance against Intel is the $1000 Threadripper 1950X against the 10-core Core i9-7900X. With a 32-core Threadripper coming soon, AMD has dared Intel to trim its per-core pricing or be buried under an onslaught of dramatically higher core count chips that offer much better performance.

But by positioning the Core i9-9900K at the top of its own stack in an eight-core / 16-thread configuration, Intel is basically daring AMD to try and bring Ryzen CPUs with higher core counts to the desktop as well. It’s not clear if such chips would fit into AMD’s existing socket infrastructure or not (Threadripper motherboards are typically more expensive than your standard AM4 products). There’s unquestionably room in AMD’s product line for a higher core count chip — the eight-core 1900X is $329, while the 12-core Threadripper 1920X is $785. But taking on Intel’s eight-core means AMD either needs a 10-core chip that can drop into that $450 price range or it needs a hell of a price cut on a 12-core chip.

And this raises another question: Which company currently has more long-term headroom? Unofficial rumors suggest Zen 2 will target a 10-15 percent IPC uplift, but AMD is still trying to close the gap with Intel overall, not surpass it. With Intel stuck on 14nm the momentum advantage is very much on AMD’s side of the equation, but given the fundamental problems with improving silicon performance, it isn’t clear how much daylight the two companies can create between each other, in the final analysis. We’re going to need to see someone jockeying with Intel for the pole position before we can tell to what extent Intel’s silicon scaling problems are unique to Intel or common to the entire industry. Right now, we think those problems tend to be common to the industry and a function of clock speeds and material properties. Based on how the next 18 months plays out, that could change.

No comments