This year, at the first-ever digital CES, Intel is announcing a slate of new processors across multiple product families. There are new updates coming for the budget Pentium Silver and Celeron families, new 11th Generation vPro and Evo vPro systems, upcoming eight-core Tiger Lake CPUs, new higher-end Tiger Lake quad-cores intended for gaming, and new data on upcoming products like Rocket Lake. Here, we’ll focus mostly on the new CPU announcements.
Intel is putting a lot of emphasis on the Chromebook market, with the first 11th Gen CPUs launching in that space later this quarter. When Chromebooks first appeared, they were typically powered by ARM CPUs, but we’ve seen a number of x86 devices launch over the past few years. Chromebooks have been particularly popular in the age of COVID-19; PC market reports for Q3 2019 showed a 1.9x growth in shipments in just a single year. AMD has also been quite interested in this market of late, so we’ll see how things evolve this year. We might see a low-end grudge match between the two companies in this space if the Chromebook market stays red-hot.
Tiger Lake Goes High Power
As far as Tiger Lake is concerned, Intel has unveiled a new Tiger Lake gaming platform (H35), and teased upcoming Tiger Lake eight-core CPUs. There are three new TGL quad-core chips — the Core i7-11375H “Special Edition,” the Core i7-11370H, and the Core i5-11300H. All of these are 35W chips:
Note: Intel has stopped providing base frequencies for its processors at the 15W TDPs it has previously reported. Instead, it now reports the CPU’s minimum frequency if run in cTDP Down mode (28W) and cTDP Up mode (35W). This is a consumer-unfriendly change that obfuscates basic information about the minimum operating frequencies that customers should expect. Without knowing whether a laptop is designed to operate in one TDP range or the other, the consumer has no way to compare expected performance. The end-user deserves to be aware of the expected minimum CPU clock in all cases.
One of the problems with mobile phones is the degree to which they obfuscate information about their own actual clock speeds, and often market chips based solely on boost clocks. This is not a trend the PC industry should copy.
With that said, these CPUs offer the advantages you’d expect from a higher TDP — higher boost clocks and likely more sustained turbos, delivering higher overall frequencies. They won’t make a tremendous difference, but they’ll probably deliver better experiences under load.
Meanwhile, Intel still isn’t ready to talk about its eight-core Tiger Lake CPU by name, but it’s willing to confirm the existence of the chip. This CPU will offer up to eight cores of Willow Cove performance, with full support for PCIe 4.0 and 20 lanes of support. Since GPUs don’t currently benefit from an x16 PCIe 4.0 connection — PCIe 3.0 can feed any modern card — it’ll be interesting to see if we see any companies using some of those lanes to enable multiple x4 linkages for M.2 storage arrays in mobile as opposed to using them solely for GPUs. Sadly, we’ll probably see a lot more GPU-focused usage of lanes rather than M.2, but one can hope.
Presumably, the eight-core TGL CPU will drop into a TDP range between 35W – 65W. Intel’s 10th Gen Core i9-10980HK has a base frequency of 2.43GHz but a TDP range of 45W – 65W, so the upcoming TGL CPU might come in anywhere in this space. With top-end frequencies of up to 5GHz, we’re guessing this isn’t a low-power chip.
This eight-core CPU will represent the most-advanced, highest-power variant of an Intel architecture you can buy in 2021, even though it may not be the fastest in absolute terms. Rocket Lake, which we’ll discuss below, is based on Cypress Cove, aka Sunny Cove, aka the CPU inside of Ice Lake — not Tiger Lake.
Rocket Lake Arrives
Finally, we have the imminent arrival of Rocket Lake, which Intel has also confirmed as arriving in Q1 2021. The Core i9-11900K is capable of 5.3GHz turbo and a 4.8GHz all-core turbo, and it will ship with support for DDR4-3200, up from DDR4-2933. It is typically possible to operate Intel CPUs at higher RAM frequencies than this, but Intel chips don’t benefit from higher RAM clocks to the same degree that AMD chips do.
Rocket Lake will be backward compatible with the Intel 400 series, and Comet Lake-S CPUs will be compatible with 500 series motherboards, except Celeron CPUs with just 2MB of cache. Those specific 10th Gen CPUs will not be compatible. This shouldn’t affect many people, since anyone with a bottom-end Celeron on a 10th Gen motherboard should have plenty of upgrade room within the Comet Lake family to make a new chip worthwhile without needing to swap motherboard platforms.
Intel is forecasting a 1.19x IPC improvement for Rocket Lake, which would be in line with our own expectations and indicates the full benefit of Ice Lake’s Sunny Cove backported successfully to Rocket Lake’s 14nm process. There will be some new features to improve chipset bandwidth, add AV1 decode, and improved integrated graphics performance compared with previous generation CPUs. Intel also believes it can tie or exceed AMD’s game performance with this next chip:
It’s possible the company can indeed pull this off. AMD’s Zen 3 architecture is faster than Intel’s current 14nm CPUs across the board, with the possible exception of some well-tuned AVX-512 workloads, but gaming is probably the place where AMD’s leadership position is weakest. Intel may not be able to slug it out with 12-16 core CPUs in absolute performance, but strong single-threaded perf could let the company reclaim gaming. We’ll see in a few months how accurate the benchmarks above, provided by Intel, are compared to independently verified results.
Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption
Intel's higher-end desktop CPU TDPs no longer communicate anything useful about the CPUs power consumption under load.
VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties
VIA and Zhaoxin are deepening their strategic partnership with additional IP transfers, intended to accelerate long-term product development.
Nvidia Unveils ‘Grace’ Deep-Learning CPU for Supercomputing Applications
Nvidia is already capitalizing on its ARM acquisition with a massively powerful new CPU-plus-GPU combination that it claims will speed up the training of large machine-learning models by a factor of 10.
How L1 and L2 CPU Caches Work, and Why They’re an Essential Part of Modern Chips
Ever been curious how L1 and L2 cache work? We're glad you asked. Here, we deep dive into the structure and nature of one of computing's most fundamental designs and innovations.