MediaTek's octa-core MT6795 shows up in benchmarks, breathes down Snapdragon 810's neck

MediaTek's octa-core MT6795 shows up in benchmarks, breathes down Snapdragon 810's neck
It looks like MediaTek has come up with a mobile application processor that can go head to head with the mighty Qualcomm Snapdragon 810. Although it's prone to aggressive throttling, and it's fabricated on a 20nm process that's growing older, Qualcomm's 64-bit SoC is still a beast and a baseline for 2015 high-end mobile performance. So it's good news from MediaTek that its new MT6795 octa-core processor can comfortably spar with the big boys, as its GeekBench test result shows. The MT6795 scored 886 points in the single-core test, and 4536 points in the multi-score test. Meanwhile, the Snapdragon 810, as we got to know from LG G Flex 2 benchmarks, hit a much better 1144 points in single-core performance, but loses with 4345 points in multi-core.

Now, the deal with these benchmarks is that they are wholly dependant on the devices they are being run on. Obviously, we can't make any final conclusions from them, as we don't know how the MT6795 was tested. We can see it was running inside an Android Lollipop device with 3GB of RAM, which means it had an edge or two against the 2GB RAM, 1080p-resolution LG G Flex 2, which housed the Snapdragon 810. However, the MT6795's score sounds promising in terms of raw performance. 

MediaTek's newest chip will enter mass production in late March, so we take it we can look forward to a second quarter full of hot-rodded Chinese smartphones!


via G For Games

FEATURED VIDEO

39 Comments

1. sram145

Posts: 127; Member since: Dec 18, 2014

tat pic is ultimate :P

24. LifeSucks

Posts: 54; Member since: Jan 15, 2015

MT6795 is a 8x A53 chip just like MT6752. See post 23 for a deeper explanation.

2. rd_nest

Posts: 1656; Member since: Jun 06, 2010

Pathetic single core scores. As best this is a mid range SOC. Having 8 or 16 cores is useless when per core performance is so pathetic.

21. alex3run

Posts: 715; Member since: May 18, 2014

And what? Why do you pay attention only th the single-threaded performance? Don't you think multicore performance matters too?

25. hemedans

Posts: 755; Member since: Jun 01, 2013

how many android apps has been optimized to use multicore?

30. alex3run

Posts: 715; Member since: May 18, 2014

There are some.

38. joey_sfb

Posts: 6794; Member since: Mar 29, 2012

All apps running on minimum Android 3.0 Honeycomb can benefit from multi-cores without modifing their apps code. Details answer, read this ->http://community.arm.com/docs/DOC-7920

3. Tritinum

Posts: 471; Member since: May 06, 2014

a bit over exaggerated title :\

4. yoosufmuneer

Posts: 1518; Member since: Feb 14, 2015

Better than 810 but Exynos 7420 just smashed them both.

7. yoosufmuneer

Posts: 1518; Member since: Feb 14, 2015

in multi-core

9. slh163

Posts: 140; Member since: May 19, 2014

and single core too. 1500 in single

12. yoosufmuneer

Posts: 1518; Member since: Feb 14, 2015

Hell YEAH!

34. j2001m

Posts: 3061; Member since: Apr 28, 2014

And that 1500 was on a 8 core setup and not just when there is 2 or 3 cores, so that score kills all others doa without even trying including Apple, I be thinking a big part of the muti core kill is also down to the removal of the lag from core to core that samsung as sorted and the 14nm and also the gpu is a kill of all other phones going

5. TheGunnyPT

Posts: 252; Member since: Feb 12, 2015

Well... Personally not a fan of Mediated, but they did said they wanted to compete on high end.. But let's be honest no one who wants a high end phone wants Mediatek on board

19. datagrab

Posts: 49; Member since: Feb 13, 2015

^This. They already have a rep for SOCs for cheap mid range devices on the market.

6. TyrionLannister unregistered

How is the single core score so poor? With cortex A57@2 GHz, this should reach about 1200. However, just wait till you see the GPU scores. It can't even match the adreno 330 in SD 800.

11. Nebice

Posts: 68; Member since: Jan 05, 2015

phones arent just for gaming :D :D

15. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

Because they are NOT A57, everyone is telling that and they are WRONG. the cores in the MT6795 are A53 clocked at high frequency. How do I know that? For several reasons: 1) If they were A57@2GHz the scores should be north of 1200 2) A57 can't be clocked that high in 28nm 3) Mediatek DOESN'T say at any point in time they are A57...http://www.mediatek.com/en/products/mobile-communications/smartphone1/mt6795/ 4) Scores are what you would expect of an A53 clocked over 1.7 MHz of the MT6752 (A53 cores).http://browser.primatelabs.com/geekbench3/search?dir=desc&page=5&q=MT6752+meizu&sort=score

18. TyrionLannister unregistered

This is interesting. But do you have a credible source/die shot of confirming it. As even the SD 810 can get these scores( even worse) after throttling: http://browser.primatelabs.com/geekbench3/1940580 However, that's an insanely high clock for A53 at 28 nm. They would be consuming power like crazy at that clock.

23. LifeSucks

Posts: 54; Member since: Jan 15, 2015

You want credibility? According to Anandtech's Exynos 5433 A57/A53 analysis article: 1) 4x A53 @ 1.3 GHz (on 20nm) consume 847 mW. So 8x A53s @ 1.3 GHz (on 20 nm) should consume around1500-1600 mW. 8x A53s @ 2.2 GHz (on 20 nm) should consume around 4 Watts. 2) 1x (just one) A57 at 1.9 GHz consumes 2 freaking WATTS. ALONE. And this on the 20 nm process. 4x A57 at 1.9 GHz consume 7.65 freakin watts on the 20 nm process. 2.2 GHz 4x A57 ON 28 nm would be using over 10 W for the CPU ALONE. Add in GPU and rest of SoC and you have a chip that LITERALLY MELTS your phone. So you see A57 is a MONSTER in terms of the amount of power it consumes. There is no way in hell they are using A57s on that chip. 3) From what I can tell. Mediatek, just like Qualcomm 615 are using two 4x A53 clusters that have different max clocks. Remember that on Snapdragon 615 one quad A53 cluster can go upto 1.7 GHz but the other cluster is limited to only 1.0 GHz. So TECHNICALLY it is still heterogeneous. 4) On their official website they mention "True Octa Core" while referring to MT6795. This is the same branding that was used for MT6592 which had 8x A7s.

26. LifeSucks

Posts: 54; Member since: Jan 15, 2015

If you see the imdividual tests and compare single core performance to multi core performance: AES: Multi Core Score = 6.6x single core score TwoFish: Multi core score = 7.4x (ALMOST 8x) single core score SHA-1: Multi core score = 7.1x single core score So whether you like it or not, it is a high clocked 8x A53 setup. And I think I may have been wrong about comparing MT6795 to S615 because all 8 cores are able to clock upto 2 GHz. But this begs the question. What is the talk about HMP on MT6795?

27. TyrionLannister unregistered

If only power consumption increased linearly.

32. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

And this SoC can go much lower than that:http://browser.primatelabs.com/geekbench3/1444870

8. realjjj

Posts: 375; Member since: Jan 28, 2014

That's a poor score actually (10% higher than the MT6752) but that's ok since it's on a dev board and the software is clearly not final.The final version should do better.

10. itsdeepak4u2000

Posts: 3718; Member since: Nov 03, 2012

Nice from MTK.

13. bugsbunny00

Posts: 2264; Member since: Jun 07, 2013

And exynos will be breathing downing mediatek's neck..

14. shahrooz

Posts: 792; Member since: Sep 17, 2013

I wish the same thing would happen to pc market too, so sad.

16. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

I really wonder how a 30% inferior score can be considered "breathing on the neck". Even the 2 years old Snapdragon 800 has a similar single core score than this SoC.

17. yoosufmuneer

Posts: 1518; Member since: Feb 14, 2015

SD810 is a huge flop. HTC and LG are among the first victims. An average person buying this phone would only blame the Manufacturer and the Software not the SoC.

22. boosook

Posts: 1442; Member since: Nov 19, 2012

An average person won't have any problem and won't notice any slowdown, since we're talking about an extremely fast chipset and the average user doesn't waste his time running batches of benchmarks to see when the thermal throttling kicks in...

Latest Stories

This copy is for your personal, non-commercial use only. You can order presentation-ready copies for distribution to your colleagues, clients or customers at https://www.parsintl.com/phonearena or use the Reprints & Permissions tool that appears at the bottom of each web page. Visit https://www.parsintl.com/ for samples and additional information.