As part of the CES cavalcade of announcements, after launching Core-M back in September, Intel is formally releasing their next element of the 14 nanometer story: Broadwell-U. As the iterative naming over Haswell-U suggests, Broadwell-U will focus on dual-core 15W and 28W units from Celeron to Core i7 using 12 to 48 ­execution units for the integrated graphics. A Broadwell-U processor should drop into any existing Haswell-U equivalent design (i3 to i3) due to pin and architecture compatibility, albeit with a firmware update.

As with any node change, the reduction to 14nm affords the usual benefits: more transistors per unit area, lower power consumption for a given design, or the potential to increase performance. Ryan covered the details of Intel’s 14nm architecture back as part of the IDF launch, as well as a good deal of the Broadwell architecture itself. The launch today is in essence a specification list with a few extra details, along with potential release dates for Broadwell-U products. The CPUs are already shipping to partners for their designs.

There will be several combinations possible throughout the Broadwell line, but the most important distinctions are:

28W with GT3, Iris 6100 Graphics (48 Execution Units)
15W with GT3, HD 6000 Graphics (48 Execution Units)
15W with GT2, HD 5500 Graphics (23 Execution Units for low i3, 24 for others)
15W with GT1, HD (Broadwell) Graphics (12 Execution Units)

The graphics move up to Generation 8, and a lot of architectural detail into this was given by Intel and IDF San Francisco in September 2014 of which some of the important points are highlighted here.

The New SKUs

Without further delay, the list of the new processors is as follows:

CPU Cores Base Freq (GHz) 1C (GHz) 2C (Ghz) EUs GPU Base /
Max Freq (GHz)
LPDDR3
/ DDR3 Support (MHz)
L3 Cache cTDP Down vPro 1K $
Intel Broadwell-U
28W + Iris 6100 Graphics
Core i7-5557U 2 / 4 3.1 3.4 3.4 48 300/1100 1866/1600 4MB 23W No $426
Core i5-5287U 2 / 4 2.9 3.3 3.3 48 300/1100 1866/1600 3MB 23W No $315
Core i5-5257U 2 / 4 2.7 3.1 3.1 48 300/1050 1866/1600 3MB 23W No $315
Core i3-5157U 2 / 4 2.5 2.5 2.5 48 300/1000 1866/1600 3MB 23W No $315
Intel Broadwell-U
15W + HD 6000 Graphics
Core i7-5650U 2 / 4 2.2 3.2 3.1 48 300/1000 1866/1600 4MB 9.5W Yes $426
Core i7-5550U 2 / 4 2.0 3.0 2.9 48 300/1000 1866/1600 4MB 9.5W No $426
Core i5-5350U 2 / 4 1.8 2.9 2.7 48 300/1000 1866/1600 3MB 9.5W Yes $315
Core i5-5250U 2 / 4 1.6 2.7 2.5 48 300/950 1866/1600 3MB 9.5W No $315
Intel Broadwell-U
15W + HD 5500 Graphics
Core i7-5600U 2 / 4 2.6 3.2 3.1 24 300/950 1600/1600 4MB 7.5W Yes $393
Core i7-5500U 2 / 4 2.4 3.0 2.9 24 300/950 1600/1600 4MB 7.5W No $393
Core i5-5300U 2 / 4 2.3 2.9 2.7 24 300/900 1600/1600 3MB 7.5W Yes $281
Core i5-5200U 2 / 4 2.2 2.7 2.5 24 300/900 1600/1600 3MB 7.5W No $281
Core i3-5010U 2 / 4 2.1 2.1 2.1 23 300/900 1600/1600 3MB 10W No $281
Core i3-5005U 2 / 4 2.0 2.0 2.0 23 300/850 1600/1600 3MB 10W No $275
Intel Broadwell-U
15W + HD (Broadwell)
Pentium 3805U 2 / 2 1.9 1.9 1.9 12 100/800 1600/1600 2MB 10W No $161
Celeron 3755U 2 / 2 1.7 1.7 1.7 12 100/800 1600/1600 2MB 10W No $107
Celeron 3205U 2 / 2 1.5 1.5 1.5 12 100/800 1600/1600 2MB 10W No $107

There are some clear patterns in the product line. Every unit apart from the Pentium and Celerons has hyperthreading, putting most of the line in a dual core, quad thread scenario. This also ties in with the Pentium and Celeron’s use of HD (Broadwell) graphics, which is a 24 EU design with half of each subslice disabled. The speeds of the Pentium and Celerons are also cut back, despite the 15W TDP and high cTDP down, ensuring that these are the bargain basement units of the line.

vPro will only be enabled on i7-56x0U and i5-53x0U series, giving a range if HD 6000 or HD 5500 is needed, however there is no vPro Iris 6100 part being released. The HD 5500 parts will have a cTDP Down of half their original TDP, allowing 7.5W designs to also take advantage of Broadwell-U.

The Core i3 15W SKUs have an odd combination involving 23 EUs rather than the 24 EUs that the die is designed with, presumably in order to keep yields higher it gives Intel a chance to still sell those with a single defect. This produces a lop-sided EU design within the configuration, which has its own implications, and we are requesting more detail from Intel as to how this is managed in the firmware.

A positive point for 6x00 series graphics SKUs is the memory compatibility on LPDDR3, with these units (having an 5 or an 8 in the 00x0 name) allowing 1866 MHz memory. As our previous Haswell desktop memory testing has shown a small bump away from 1600 MHz DRAM can give a good performance boost when it comes to graphics, especially when the memory speed between CPU and DRAM is the main bottleneck. I would be interested in exploring the difference with this for sure.

It might come across as somewhat surprising that a 15W CPU like the i7-5650U has a 2.2 GHz base frequency but then a 3.2 GHz to 3.1 GHz operating window, and yet the i7-5557U has a 3.1 GHz base with 3.4 GHz operating for almost double the TDP. Apart from the slight increase in CPU and GPU frequency, it is hard to account for such a jump without point at the i7-5650U and saying that ultimately it is the more efficient bin of the CPUs. So while the 28W models will get the glory in terms of performance, there are a number of models that can offer just under that performance but for just over half the power rating. This obviously levels battery life for the more efficient design as a significant jump, depending on how the system as a whole is used.

The Dies and Packaging

Broadwell-U will be derived from two main dies. The larger design contains the full 48 EU (two common slices with 6x8 EU sub-slices all in) configuration for 1.9 billion transistors in 133 mm2, while the 24 EU design (one common slice, 3x8 sub-slices) will measure 1.3 billion transistors in 82 mm2.

This puts the size of one common slice  with 3x8 sub-slices at 600 million transistors / 49 mm2, and thus the die without the graphics subsystem at all at 700 million transistors for 33 mm2. This would mean the cores, the Last Level Cache, the IO and memory controller all fit into the 700 million.

Compared to Haswell-U, Intel provided the following data:

Broadwell-U with HD 5500 (24 EU) has 240M more transistors than Haswell-U with HD 4400 (20 EU)
Broadwell-U with HD 6000 (48 EU) has 600M more transistors than Haswell-U with HD 5000 (40 EU)

Unfortunately calculating the increases for separate parts is a little more difficult than just comparing numbers due to the different elements of the new graphics, known as Intel Gen 8.

In terms of the packaging for the dies, we also have some shots of those to share:

On the left is the 2+2 configuration, giving two cores and GT2 (24 EUs), while on the right is the 2+3 package. The silicon on top is the Platform Controller Hub, discussed later.

Broadwell GPU Improvements
POST A COMMENT

85 Comments

View All Comments

  • KaarlisK - Monday, January 05, 2015 - link

    "It might come across as somewhat surprising that a 15W CPU like the i7-5650U has a 2.2 GHz base frequency but then a 3.2 GHz to 3.1 GHz operating window, and yet the i7-5557U has a 3.1 GHz base with 3.4 GHz operating for almost double the TDP. Apart from the slight increase in CPU and GPU frequency, it is hard to account for such a jump without point at the i7-5650U and saying that ultimately it is the more efficient bin of the CPUs."
    This is not surprising. This is used to increase the GPU performance. 28W CPUs have Iris 6100, 15W CPUs have HD 6000.
    In no way does TDP tell us anything about efficiency.
    Reply
  • aratosm - Monday, January 05, 2015 - link

    Iris 6100 vs HD 6000 are almost identical. The only difference is a slightly faster clock speed. I think the problem is, HD6000 will throttle more to stay in that power envelope. Reply
  • Topinio - Monday, January 05, 2015 - link

    Looks to me like the 23W ones (i.e. those with the 6100 graphics) will be the only ones to be capable of being near the max turbo clocks for long.

    Would also be interesting to know the AVX base and turbo clocks for these chips, to compare the possible 64b DP GFLOPS from the CPU cores to those listed on page 2 from the GPUs. Top bin is likely somewhere < 102 (vs 211 from GPU), but how much lower?
    Reply
  • MrSpadge - Monday, January 05, 2015 - link

    For the big Xeons the AVX base clock is typically 200 MHz below the regular base clock. They operate in a similar frequency & voltage range as the mobile chips (and are as power-limited as they are), so expect the same to apply here. Reply
  • hansmuff - Thursday, January 08, 2015 - link

    First time I read about AVX clocks, then found another mention in a previous Xeon CPU article. Is this a thing for Xeon only, or do the Haswell desktop chips throttle the clock with heavy AVX as well? Reply
  • naloj - Monday, January 05, 2015 - link

    A good example of this is in the throttling of the HD5000 in the 15W NUC i5-4250. You can get 40% better performance by changing the TDP settings from 25W short burst / 15W steady to 35W short burst / 31W steady. Reply
  • MrSpadge - Monday, January 05, 2015 - link

    "In no way does TDP tell us anything about efficiency."

    Agreed - TDP is far to crude for this. Intel Desktop CPUs often operate far below TDP, whereas mobile chips are throttled by it. How much? Depends on the laptop, environment temperature etc.

    So even though the 15 W CPUs quoted above are allowed to top out at 3+ GHz, they won't run at anywhere close to this frequency under sustained heavy load. The 28 W chips should have no trouble sustaining the speed, given adequate cooling.
    Reply
  • zepi - Monday, January 05, 2015 - link

    Iris 6100 + edram or at least DDR4 bandwidth increases would have made a terrific difference to "retina" and high-dpi ultrabooks / laptops, but now this upgrade is watered to irrelevancy.

    Nothing to see here...
    Reply
  • HungryTurkey - Wednesday, January 14, 2015 - link

    In a retina/hdpi environment, few applications would come close to saturating the bus. The EUs (even with the 6100) would bottleneck long before LPDDR3/DDR3 would. Reply
  • fokka - Thursday, January 08, 2015 - link

    as i see it the given tdp only ensures operation at base clocks and without a substantial graphics load. operation at turbo clocks requires to overstep the tdp until power draw and or temps are too high and the clock returns to the base frequency.

    if you look at it like this it's not surprising a 15w sku has a base clock of 2.2ghz ans a 28w sku 3.1ghz. that said the 28w tdp still looks "too high" for the frequency you get out of it, but i guess that this extra power/heat-budget is there for the sole reason so the 28w sku can operate at turbo clocks for longer without throttling down again, plus there is more headroom for a graphics load at the same time. this ensures, even with similar hardware and turbo-clocks, the 28w sku is allowed to produce more heat and in turn get more work done in the same time.

    that's the same reason core-m has very high turbo speeds, but can only turbo for a couple seconds until it's too hot and it "throttles" down to base clocks.
    Reply

Log in

Don't have an account? Sign up now