MediaTek added two additional products to its premiere deca-core SoC family today. The Helio X23 and X27 join the previously released Helio X20 and X25, adding additional performance levels for OEMs to choose from.

All four SoCs use the same deca-core CPU configuration—two ARM Cortex-A72 cores for heavy workloads that require higher performance and two quad-core Cortex-A53 clusters, using different DVFS curves, for medium to light workloads where lower-power operation is desirable. MediaTek’s custom-developed, power-aware scheduler, CorePilot 3.0, still handles thread distribution. The differentiating factor is clock frequency, with peak frequencies increasing from 2.1GHz / 1.85GHz / 1.4GHz (A72 / A53 / A53) for the X20 to 2.6GHz / 2.0GHz / 1.6GHz for the new X27. The ARM Mali-T880MP4 GPU runs at up to 780MHz in the X20 and X23 and up to 850MHz in the X25. The new X27 bumps the GPU frequency even higher to 875MHz, just shy of the 900MHz peak frequency used by HiSilicon’s Kirin 950/955 that also uses a Mali-T880MP4 GPU.

MediaTek Helio X2x Family
SoC MediaTek
Helio X20
(MT6797)
MediaTek
Helio X23
MediaTek
Helio X25
(MT6797T)
MediaTek
Helio X27
(MT6797X)
CPU 2x Cortex-A72
@2.1GHz

4x Cortex-A53 @1.85GHz

4x Cortex-A53 @1.4GHz
2x Cortex-A72
@2.3GHz

4x Cortex-A53 @1.85GHz

4x Cortex-A53 @1.4GHz
2x Cortex-A72
@2.5GHz

4x Cortex-A53 @2.0GHz

4x Cortex-A53 @1.55GHz
2x Cortex-A72
@2.6GHz

4x Cortex-A53 @2.0GHz

4x Cortex-A53 @1.6GHz
GPU ARM Mali-T880MP4
@780MHz
ARM Mali-T880MP4
@850MHz
ARM Mali-T880MP4
@875MHz
Memory
Controller
2x 32-bit @ 933MHz
LPDDR3

14.9GB/s b/w
Encode/
Decode
encode:
2160p30
H.264 / HEVC w/HDR

decode:
2160p30 10-bit
H.264 / HEVC / VP9
Camera/ISP Dual ISP
32MP @ 24fps (single camera)
or
13MP + 13MP @ 30fps (dual camera)
Integrated
Modem
 LTE Category 6
DL = 300Mbps
2x20MHz CA, 64-QAM

UL = 50Mbps
1x20MHz CA, 16-QAM

FDD-LTE / TD-LTE / TD-SCDMA / WCDMA / CDMA / GSM
Mfc. Process 20SoC (planar)

Both the Helio X23 and X27 are manufactured on the same TSMC 20nm process as the X20 and X25. MediaTek has not made any implementation changes to the layout or cell libraries to push the X27 to higher frequencies, which makes the fact that its A72 cores can hit 2.6GHz on a 20nm planar process even more interesting considering ARM only targets frequencies this high on 14nm/16nm FinFET processes.

Other blocks within the X23 and X27, including MediaTek’s Imagiq ISP and integrated Category 6 modem, remain unchanged. The ISP still includes support for dual camera sensors, including color + monochrome configurations that can capture more light than a single color sensor, potentially improving low-light photography. There’s also a 3D depth engine built in to enable depth-of-field post-processing effects.

One new feature added to the X23 and X27 is EnergySmart Screen power-saving technology. Part of MediaTek’s MiraVision display suite, it modifies various display parameters in response to changes in ambient lighting and screen content to reduce display power consumption by up to 25%.

Products using the new Helio X23 and X27 SoCs have yet to be announced, but MediaTek says smartphones using the new chips will be available soon.

Comments Locked

17 Comments

View All Comments

  • SaolDan - Thursday, December 1, 2016 - link

    Neat!!
  • shabby - Thursday, December 1, 2016 - link

    Binning to the eXtreme!
  • jjj - Thursday, December 1, 2016 - link

    The 2.1GHz was MT6797m with X20@ 2.3GHz being MT6797. Now the X23 gets the D suffix.
  • Valantar - Thursday, December 1, 2016 - link

    Hm. Seems reasonable that the X20 and X25 are being replaced by these, no? Given how incredibly closely specced they are, this seems like Mediatek doing a "Haswell Refresh". No way they'll keep marketing two sets of high-ish-end SoCs with a scant 100mHz between them.

    Kind of makes sense, as the A72 is being replaced in (some) high-end chips by the A73, while the 20nm production process is ever more obsolete. Fitting to do a refresh to keep these chips relevant in the "lower high-end" or "high-end midrange".
  • serendip - Friday, December 2, 2016 - link

    Highest thing is how do these compare yo Snapdragon equivalents? Decacore Mediatek chips seem to score well on benchmarks but they're not as speedy in the real world, along with using more power than Qualcomm chips on a larger process node.
  • Valantar - Friday, December 2, 2016 - link

    True - but that's a given with the process advantage of the Snapdragon 820/821 and the Exynos 8890. What would be interesting here would be to compare actual operating frequencies in real-world scenarios, and power consumption/heat under these conditions. Considering the performance of the Kirin 950/955, the A72 seems like a decent core.
  • serendip - Friday, December 2, 2016 - link

    I'm actually thinking about the Snapdragon 65x chips on 28nm. The A72 cores on those pack quite a punch while overall SoC power consumption is quite low because of the A53 cores used for most processing. I'm still not convinced about Mediatek's decacore designs - they're not good enough for flagships while comparable Snapdragon chips have better performance *and* lower power consumption.
  • name99 - Friday, December 2, 2016 - link

    I think it's interesting as a general point to consider the fact of the 20nm here.
    Ever since ATI and nV decided to skip 20nm, the internet jumped to the conclusion that 20nm was "worthless", a mistake by TSMC, and was destined to be a "short-lived node" (whatever that means).

    After two years experience I think we can conclude from this something about internet hysteria and how it is driven by a few large bullet-points ('nV doesn't want to use TSMC 20nm. OMG, that means 20nm must be useless for ALL purposes.").
    Obviously the leading edge has moved past 20nm, and by Q2, maybe even Q1 2017, the leading edge will have moved past 16nm FF on both TSMC and Samsung. But that was always going to happen; the point is what happened after the leading edge moved on. And it's interesting to see that 20nm continues to have some life in it, because the entire world doesn't need to run on the leading edge.

    More interesting, I think, is the question of why GloFo has been so unable to exploit this on-going interest at 20nm to offer a process that's cost competitive with TSMC and retains the simplicity but uses FD-SOI to match FF performance. They've been talking this up for what, a year?, two years?, now but nothing ever seems to happen.
  • RaichuPls - Friday, December 2, 2016 - link

    Any idea why they're still using that underpowered GPU? The Galaxy S7 has a MP12 configuration...
  • MrSpadge - Friday, December 2, 2016 - link

    Because it's very probably an unchanged layout. Maybe purely binned chips, maybe with some physical layout tweaks. Offering these doesn't cost thme much, whereas including a stronger GPU (what for?) would require a serious redesign in the sense that lots of things would have to be checked again.

Log in

Don't have an account? Sign up now