Core Features



To reiterate, Broadwell-E is a ‘Tick’ revision in Intel’s CPU roadmap, meaning that it's essentially a die-shrink of the previous architecture. Like Skylake therefore it is built on Intel’s 14nm Tri-Gate Transistor process, but shouldn’t see major IPC improvements over Haswell-E. It remains to be seen if the release will be a repeat of Ivybridge-E – something greeted with a relative lack of enthusiasm from the enthusiast community - or welcomed with glee by those on the cutting edge of requirements.



The Core i7-6950X is a 10-core CPU with 25MB L3 Cache, a steep step up over the i7-5960X’s 8-cores and 20MB L3 Cache. In common with all CPUs on this platform the 6950X supports Hyperthreading, doubling the number of supported threads per core from one to two (up to a maximum of 20 on the 6950X). The cores are clocked at a base clock speed of 3.0GHz, matching the i7-5960X, and will boost to 3.5GHz through the use of Intel Turbo Boost Technology 2.0, once again matching the 5960X.

Physically, the Broadwell-E CPU is a relatively large one with particularly dense transistor arrangement thanks to the use of 14nm Tri-Gate Transistor technology. It's believed that the CPU shares all the hallmarks of the entry-level Broadwell-EP model, which has a 246mm^2 die and total of 3.4bn transistors compared with 356mm^2 and 2.6nb respectively on Haswell-E. As a result it's likely to get distinctly hot under the collar, cramming more into a smaller surface area. This may have a knock-on effect on temperatures, and will mean that Intel's choice of TIM and headspreader will be critical to successful overclocking.



For the first time on the HEDT platform Broadwell-E will have native support for quad-channel DDR4-2400. This speaks to a slightly more robust memory controller on the new CPUs, which will hopefully have the knock-on effect of higher stable overclocks. DDR4 itself has become distinctly more affordable in the past 20 months, and one can hope that additional flexibility in compatible memory types brings with it both higher performance and more consumer choice.

Fully equipped CPUs in the family are kitted out with 40 PCI-E 3.0 lanes supported in 2 * 16x + 1 * 8x configuration, allowing the platform to natively support quad-GPU configurations without the need for PCE-E bridge chips. Only the Core i7-6800K is restricted in this aspect, limited to 28 lanes from the CPU but nonetheless still well suited to multi-GPU setups.

All Broadwell-E CPUs are compatible with the LGA 2011v3 socket, and are hence supported by X99 motherboards currently on the market (although a BIOS update might be required in some instances).



Three additional SKUs will be launching alongside the flagship Extreme Edition – the Core i7-6900K, i7-6850K and i7-6800K. Chief differentiating factors between these SKUs are the number of cores (either 8 or 6) and number of PCI-Express lanes (40 or 28). In this they mirror the Haswell-E release, offering distinct performance levels and features suitable for a fairly wide range of use cases.

Compared with Haswell-E, Broadwell-E increases both the base clock speed and Turbo mode by one or two bins, i.e. 100-200MHz. In itself this won’t make a major difference – even on paper it’s only between 3 and 7% boost over the previous generation – but assuming a zero IPC change over Haswell-E we would still expect to see a slight performance improvement for the new generation. Of course, at the top end and with ten fully unlocked cores, the Extreme Edition part bucks the trend.

Naturally the most intriguing part remains the ‘low-end’ Core i7-6800K. Priced relatively attractively at $412 MSRP, it encroaches slightly on the Skylake i7-6700K and could certainly appear to be a ‘bang for your buck’ winner depending on out-the-box and overclocked performance. That’s especially the case given the maturity of the X99 platform and quality of even ‘affordable’ models in that range, with feature specs that rival more expensive Z170 designs.

We should also note that Intel are gradually reducing the complexity of their desktop CPU naming scheme. Whilst more confusion is typically associated with the mobile PC CPU landscape, the more straightforward conventions in the 6000-series of CPUs should be to everyone’s benefit, especially if new models are released in the range during (for instance) a Skylake refresh later this year.

Turbo Boost Max Technology 3.0

Turbo Boost Technology has been a fixture of Intel CPUs since the Nehelem days, allowing a CPU to dynamically overclock beyond a ‘base’ frequency based on a number of factors including core temperatures, power draw and the number of cores in operation. Turbo Boost Tech. 2.0, introduced with Sandybridge in 2011, smoothed out the overclocking curve and let the CPU draw more power than its theoretical peak TDP so long as temperatures remained within specified limits.

Turbo Boost Max Technology 3.0 is exclusive to Broadwell-E, offering per-core Turbo Ratio settings rather than a blanket guidance. The exact magnitude of the Turbo scaling will differ from part to part as the core is monitored according to internal limits for power draw and consumption. Whilst taking advantage of this additional headroom Turbo Boost Max Technology 3.0 brings to the table is optional, its worth noting that doing so doesn’t void the warranty.

The Intel Turbo Boost Max Technology 3.0 application also allows the user to assign an applicaton to specific cores such that the most demanding applications are assigned to the fastest cores. Although a small thing, this level of control is another aspect of Broadwell-E which sets it apart from Haswell-E and could be a major boon in 'Mega Tasking'.

Turbo Boost Max Technology 3.0 requires both a specific driver and application to support it, but will eventually be integrated into Windows-based OS as standard. The tech should be supported by all X99 motherboards updated to the latest BIOS, but that may depend on the whims of your motherboard vendor.

Mega Tasking



Greater control over thread allocation to specific cores, alongside an increase in the number of cores exposed to the operating system, will also have a complementary affect on performance in so-called mega-tasking workloads. Defined as a group of demanding tasks which are performed simultaneously, a core aspect of Broadwell-E is reducing the waiting time between workloads to optimise the use of resources.

A classic ‘Mega-Task’ example is online game streaming, where the user may be running a game at a high resolution, streaming the game via Twitch, and encoding the video for later editing as a standalone video. Intel estimates that with the Core i7-6950X encoding times are reduced by up to 25% whilst simultaneously gaming in 4K and streaming in 1080, but in all fairness this may be down to an increased core count and improved scheduling/resource allocation.

New Overclocking Features

Finally, Broadwell-E brings with it a trio of overclocking features which weren’t a part of the Haswell-E lineup.