Now that Broadwell’s protracted , awkward launch is complete , the company would like you to turn your attention to yet another new architecture that’s right around the corner: Skylake. Intel insisted that the problems with the new 14nm manufacturing process that so delayed Broadwell wouldn’t affect Skylake, and it appears to be holding to that promise even though it makes some of the Broadwell CPUs look more than a little pointless.
Today the company is releasing just a dribble of Skylake information centered around a few specific products: the high-end overclockable Core i7-6700K and i5-6600K, and the accompanying overclocker-friendly Z170 chipset. We’ve got some performance numbers for the CPU and its new integrated Intel HD 530 GPU over here (after a few years of four-digit model numbers for GPUs, it looks like Intel is switching to a three-digit system in Skylake). Here, we’ll focus primarily on an overview of the Z170 chipset and the DDR4 memory controller included in the Skylake CPUs, since they have important implications for storage and integrated graphics speed.
|Know your codenames|
|Codename and year||Process||Prominent consumer CPU branding||Tick/tock|
|Westmere (2010)||32nm||Core i3/i5/i7||Tick (new process)|
|Sandy Bridge (2011)||32nm||Second-generation Core i3/i5/i7||Tock (new architecture)|
|Ivy Bridge (2012)||22nm||Third-generation Core i3/i5/i7||Tick|
|Haswell (2013)||22nm||Fourth-generation Core i3/i5/i7||Tock|
|Broadwell (2014/2015)||14nm||Fifth-generation Core i3/i5/i7, Core M||Tick|
|Skylake (2015)||14nm||Sixth-generation Core i3/i5/i7, Core M||Tock|
|Kaby Lake (2016)||14nm||TBA||Neither tick nor tock|
Intel will be doing a full Skylake architecture disclosure at the Intel Developers Forum later this month, and both Peter Bright and I will be there to get all the details—expect more about the full range of Skylake processors then, and look forward to at least a partial launch of Skylake at some point in Q3 of 2015.
To head some of you off at the pass, no, DDR4 support isn’t technically a feature of the 100-series chipsets themselves; memory controllers in mainstream CPUs were moved from the chipset to the main CPU die years ago to reduce latency. RAM support still feels like a chipset-adjacent topic, though, so we’ll talk about it here for a bit.
DDR4 follows the same trends set by DDR2 and DDR3: lower power usage (1.2V, down from 1.5 or 1.35V for DDR3), higher data transfer rates, and marginally higher CAS latency . DDR4 also promises greater density, meaning we could soon see higher memory capacities crammed onto a single DIMM—third-party PC makers advertising Skylake desktops say that the chips support up to 64GB of RAM, up from 32GB with Haswell and Broadwell.
Skylake’s memory controller can support either 1600MHz DDR3L or 2133MHz DDR4, but since the two memory standards use different slots it’s likely that most motherboards and PCs will either support one or the other.
The memory manufacturers have been making DDR4 parts for a while, but so far they’ve been limited to use in servers and high-end workstations. Since it’s still less commonly used than DDR3, it tends to be more expensive as of this writing—a glance at Newegg says that 16GB of 1600MHz DDR3 (two 8GB DIMMs) starts at around $75 or $80 while the same amount of 2133MHz DDR4 costs between $100 and $110. As DDR4 becomes more widely used, RAM makers will increase their production and drive down prices, just as they did when DDR2 and DDR3 went mainstream.
We’d expect DDR4 to be the most important to users of Intel’s integrated GPUs, which respond well to increases in memory bandwidth. A handful of Broadwell CPUs supported 1866MHz DDR3 instead of the standard 1600MHz; that small bump in memory speed improved graphics performance by around ten percent .
PCI Express 3.0, and what it means for SSDs
Like the 9-series chipsets, the Z170 supports storage connected via PCI Express rather than the older, slower SATA interface. The difference is that The Z170 includes PCI Express 3.0 lanes instead of PCI Express 2.0 lanes.
Up until now, Intel’s chipsets have topped out at PCI Express 2.0. Its higher-end (usually quad-core) platforms supported PCI Express 3.0, but that was because Intel built PCIe 3.0 lanes directly into those processors for use with high-end video cards. Ivy Bridge was the first architecture to include this feature, but Haswell and Broadwell did it this way too. And high-end Skylake CPUs like the ones launching today continue to include 16 of those PCIe 3.0 lanes for use primarily with graphics cards—you can use one card with 16 lanes, two cards with eight lanes, or three cards by giving one card eight lanes and the other two cards four lanes.
For a long time, skipping PCIe 3.0 in the chipset was fine. The other things that you would connect to a chipset—Ethernet and wireless controllers, TV tuner cards, external sound cards, and other peripherals—didn’t need more bandwidth. Solid-state hard drives have changed that, and as they’ve improved they’ve continued to suck down every scrap of bandwidth we can give them.
So far, the only OEM shipping an SSD connected to PCI Express 3.0 is Apple, who presumably used some of the processor lanes for the SSD in the otherwise underwhelming 2015 15-inch Retina MacBook Pro refresh . Here’s a chart from that review—compare the read speeds of the 13-inch model, which uses four PCIe 2.0 lanes, to the PCIe 3.0 lanes in the 15-inch model (the write speeds in the 13-inch model are lower in part because it’s a lower-capacity drive; 256GB and 512GB drives tend to have better write performance since the controller can write to more NAND chips at once).
PCI Express 3.0 is theoretically twice as fast as 2.0, and while the new 15-inch rMBP doesn’t quite get there, it’s not too far off. Now imagine this as a standard option on high-end PCs. If you’re using a SATA III SSD right now, the potential speed increase will be ridiculous. Just keep in mind that not all of the 100-series chipsets may include PCIe 3.0 lanes.
Other chipset features, and the rest of the family
For this section, we’ll supplement Intel’s official information with a few leaked slides from reasonably reliable sources. Those slides suggest that there will be a total of six 100-series desktop chipsets, with Z170 sitting on the top of the pile. These chipsets replace the full range of offerings from both the 8- and 9-series chipset families—the 9-series chipsets were intended mainly for enthusiasts, and Haswell’s long life (and the limited desktop Broadwell rollout) let Intel use the 8-series chipsets to fill other market gaps. As with Intel’s CPU lineup, the list of chipsets is segmented and needlessly convoluted, and there are plenty of differences you need to be aware of.
Of those six chipsets, five include PCI Express 3.0 lanes. Of those five, three chipsets include enough lanes to support multiple SSDs: The overclocker-targeted Z170 and the business-and-workstation-focused Q170 can each support up to three drives since they include 20 PCIe 3.0 lanes, while the more mainstream H170 and its 16 lanes can support two drives.
The B150 and Q150, more mainstream business-and-workstation-level chipsets, include eight and ten PCIe 3.0 lanes (respectively) but they appear to be reserved for other peripherals. The low-end H110 only includes six PCI Express 2.0 lanes, the same number as the H81 chipset it replaces.
The H170, Z170, and Q170 chipsets are the only ones that support RAID and the Smart Response Technology feature , which lets you use a smaller SSD to speed up a larger HDD. The final major difference, excepting enterprise-level features like vPro, is the number of USB ports each chipset supports. Z170 and Q170 support a total of 14 USB ports, and as many as 10 can be USB 3.0. H170 and Q150 also support a total of 14 ports, but only eight can be USB 3.0. B150 supports 12 ports and six USB 3.0 ports, and H110 supports 10 ports and four USB 3.0 ports.
The big chipset-related question that remains is which of these features will make it into the mobile versions of the Skylake chipsets, the ones that are integrated into the low-voltage Y- and U-series processors. Hopefully this information is included along with the rest of the architectural information revealed at IDF in a couple of weeks.
Listing image by Intel
- Asus ZenBook and VivoBook notebooks launched with 11th Gen Intel CPU and Iris Xe graphics
- Samsung launches flagship 980 Pro PCIe 4.0 SSD that can hit a blistering 7,000 MB/s
- Meet Lenovo’s Flagship Tiger Lake Laptops, the Yoga and IdeaPad 9i Series
- Samsung's next-gen 980 Pro SSD to deliver blistering speeds up to 7,000MB/s
- Best SSDs for gaming
- Intel Core i9-10850K CPU Benchmarks: Cheaper, but Nearly Identical to 10900K
- Intel Tiger Lake: Release Date, Specs, Benchmarks, and All We Know
- These will be the first Intel Comet Lake-S Z490 motherboards
- Intel Launches Tiger Lake: Up to 4.8 GHz, LPDDR4 Memory, Iris Xe Graphics up to 1.35 GHz
- Lenovo Yoga Slim 7 Tested: Ryzen 7 4800U Slams Intel (But You Can’t Buy It)
- Asus’ latest ZenBook laptops feature Intel’s 11th Gen CPUs and Thunderbolt 4 ports
- Deals roundup for Aug. 28
- Lenovo launches flagship Yoga 9i line, Legion Slim 7i, and slew of other devices
- Asus’ New Zenbooks Include Tiger Lake, 2:3 Aspect Ratio
- MSI Enters the Boardroom With Tiger Lake Summit Business Laptops
- The best cheap Dell laptop deals and prices for September 2020
- The best cheap MacBook sale prices and deals in September 2020
- Newegg Labor Day Sale 2020: The 5 best deals
- The best Dell XPS 13 and 15 deals and prices for September 2020
- The best Microsoft Surface Pro deals and bundles for September 2020
Intel’s 100-series chipsets: DDR4, PCIe 3.0 SSDs, and other Skylake supplements have 1601 words, post on arstechnica.com at August 5, 2015. This is cached page on Health Breaking News. If you want remove this page, please contact us.