AnTuTu: Galaxy S7 with Snapdragon GPU scores 20% higher than the Exynos version
1. Sidewinder (Posts: 429; Member since: 15 Jan 2015)
I bet the Exynos was either downclocked or was at power saving mode when the test was carried out.
6. hung2900 (Posts: 956; Member since: 02 Mar 2012)
So basically you "accepted" this right?
Is the information above reliable enough for me to believe?
10. Nathan_ingx (Posts: 3787; Member since: 07 Mar 2012)
Which application notified you that I accepted?
I'm I different to both until I see the real thing. Besides, Samsung isn't my first choice anyway.
15. Finalflash (Posts: 3104; Member since: 23 Jul 2013)
Yes it is believable and you Samsung fans need to get over it. The new version of Antutu also favors fewer but larger cores. It's going to give the multicore setups a lower score but then they score higher elsewhere. Stop being so defensive about such a meaningless number. Taken with all others it paints a meaningful picture but by itself it may as well be one pixel of a whole image that you desperately want to figure out.
8. JumpinJackROMFlash (Posts: 450; Member since: 10 Dec 2014)
Like Samsung wouldn't improve on the 74xx Exynos... Sigh...
11. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
Its Antutu anyways
Lets wait for GFXBench and 3DMark
Or more importantly IMO, lets wait to see which is better in terms of efficiency and sustained performance
49. adi9764 (Posts: 57; Member since: 16 Feb 2016)
It's just weird cause in the geekbench leaked scores from months ago for example the exynos variant scored way higher than in the recent ones
59. j2001m (Posts: 2943; Member since: 28 Apr 2014)
All the geakbench scores are all very low for the (f) ver of the phones
It's been reported that the Samsung CPU is in some type of limiting mode to stop the score from being showed (this is by bloggers that have used the phone), as the sd 820 is going to be the same in all phone that not happening on the other hand but I have noted some very bad scores for it, so does it under clock fast?
22. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
Even from the Exynos 5410 till now has been clocked lower to 800ghz to match Snapdragon. Or-else Exynos has always kicked the butt of Snapdragon.
2. johanbiff (Posts: 403; Member since: 31 Mar 2015)
My gut tells me that this isen`t the hole story, I think the 8890 will be the more powerful one in the end.
19. combatmedic870 (Posts: 631; Member since: 02 Sep 2015)
One would expect it to be with 2 more A72 cores...
24. vuyonc (Posts: 1014; Member since: 24 Feb 2014)
Huh? They both use the same number of custom A72 cores. Were you referring to the extra A53 cores from the Exynos?
26. Macready (Posts: 958; Member since: 08 Dec 2014)
It doesn't have A72 cores. Those are M1 custom cores.
33. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
The 8890 should've been made on Arm Cortex A15, rather than power efficient A57, and 14nm. The A15 were on big dies with 28nm. Could we imagine the A15 on the S6?
That's why Apple uses bigger cores rather than crappy tiny 14nm Pos.
46. johanbiff (Posts: 403; Member since: 31 Mar 2015)
you mean A53 right? no 57 cores on the 8890(mongoose)
65. jalebi2000 (Posts: 180; Member since: 01 Mar 2015)
8890 uses custom made cores coupled with 4more a53 cores in littlebig
4. Nathan_ingx (Posts: 3787; Member since: 07 Mar 2012)
First there was one version so people had to fight over the choices of OEMs and now two versions of the same, resulting in fans of the same product clashing.
9. TerryTerius (Posts: 1789; Member since: 10 Apr 2014)
Pretty much this. I don't know why anyone on here is even upset or gives a damn which chip is scoring higher on synthetic benchmarks that mean basically nothing in the real world.
47. Shocky (unregistered)
I doubt anyone is upset.
I doubt there will be much difference and it make no difference either way.
Well if the gap is 20%+ for either SoC it will effect my next device. :D
7. Biernot (unregistered)
hahahah Benchmarks xD
12. Wiencon (Posts: 1882; Member since: 06 Aug 2014)
Really, are there any hardcore gamers on phones?
13. NoToFanboys (Posts: 1575; Member since: 03 Oct 2015)
Nope, but graphical improvements are always welcome.
14. RebelwithoutaClue (Posts: 2908; Member since: 05 Apr 2013)
Playing candy crush saga 24/7 is pretty hardcore
25. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
I play at 50Fps on Dungeon hunter, GrandTheftAuto, RealRacing on my rooted extreme overclocked SonyZ5.
27. ibend (Posts: 4639; Member since: 30 Sep 2014)
still cant beat my liquid nitrogen cooled overclocked HTC M9, it run crysis at 60fps
38. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
What's you're Gpu overclocked to 900mhz?
I need to replace my Kernel on my SonyZ5. There hardly any extreme Roms and Kernels available. Sucks.
My phone goes through battery fast rather than standard Z5, but I carry a powerbank which gives me 2500mah, so doesn't bother me.
44. tedkord (Posts: 11706; Member since: 17 Jun 2009)
We're talking Justin Bieber hardcore there.
23. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
Wildblood at 60fps. I have rooted Z5 with increased Cpu/ Gpu and voltage. I'm getting 50+ fps.
17. TyrionLannister (unregistered)
According to the result, exynos 8890 scores 20k less on GPU. That's bulls**t as that would mean that 820 will be 60% more powerful than 8890 in GPU.
Alse according to antutu, snapdragon 810 is more powerful than exynos 8890. LMAO
Anyway, here's another leak saying both have around equal GPU. Which should be expected:
And PA, the 8890 is nothing but a generic architecture. It's a highly complex GPU. You have seen how good the mp4 variant of mate 8 scores. Multiply them by 3.
20. Arch_Fiend (Posts: 2501; Member since: 03 Oct 2015)
3 times the cores does not equal 3 times the performance, just sayin.
The Mi5 didn't score especially high on GFX, with Sd820 and 1080p screen it barely beat the iPhone 6s plus by 1-3 FPS.
We'll just have to wait until all these devices can be tested to know the true story. somehow all these leaks just don't seem all that accurate.
30. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
It COULD be pretty close to 3x the 950's T880-MP4
More "GPU cores"/clusters can mean an almost linearly improvement in GPU performance
But there's other things:
*Thermals/efficiency, Samsung might downclock for thermal/efficiency improvements
*TSMC's new 16FF+ vs Samsung's new 14LPP
*Samsung's E8890 probably has more memory bandwidth (but is it enough for 3x the cores)
*Kirin's 3 year old ARM CCI-400 vs Samsung's new custom SCI interconnect
If you dont believe me that GPUs scale linearly, see the A9X/A8X and A9/A8
The A9X/A8X just have double the "GPU cores"/clusters than the A9/A8
The A9X/A8X have about double the GPU performance of the A9/A8
34. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
BTW: I'm saying "GPU cores" because there's not really such a thing
GPUs are not CPUs
40. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
There are Kernels which you can have both Cpu,Gpu running in series. That's double the Gpu.
56. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
Depends on the manufacturer
Imagination calls them clusters (Imagination used to advertise number of ALUs)
ARM calls them shader cores
Qualcomm doesn't refer to those, they just use number of ALUs
NVIDIA calls them Maxwell Streaming Multiprocessors (but NVIDIA advertise number of CUDA cores)
60. alex3run (Posts: 713; Member since: 18 May 2014)
ARM's and IMG's shader cores are not just ALUs.
66. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
Yep, that's right
Also NVIDIA's Maxwell Streaming Multiprocessors aren't just ALUs too
I only mentioned ALUs for Qualcomm, since AFAIK Qualcomm dont mention other details
And for Imagination, since they used to advertise by the number of ALUs, but now they've switched to advertising number of clusters now
67. alex3run (Posts: 713; Member since: 18 May 2014)
Qualcomm difn't unveil even the ALU count so... Also many things apart from ALUs are important for gaming performance.
69. vincelongman (Posts: 4444; Member since: 10 Feb 2013)
I've seen Qualcomm's ALU count online
Not sure if they are official
ALUs are just one factor
Although ALUs probably the most important factor
39. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
To answer you're question is to root.
The standard commercial gpu on phones run at 600mhz. I overclocked mine to 900mhz, and with Samsung phones there are Kernels which you can insanely raise them to 1200mhz. That's extreme gpu performance.
32. TyrionLannister (unregistered)
3 times the GPU cores equal 3 times the performance. Read my comment #28.
21. Zack_2014 (Posts: 378; Member since: 25 Mar 2014)
You should know that increasing the cores from 4 to 12 doesn't signify 3x performance. Sigh
28. TyrionLannister (unregistered)
If you knew anything about how GPUs work, you wouldn't be saying that. GPU workloads are highly parallel. That's why nvidia cuda based GPUs have 100s of cores.
Every polygon can be calculated in parallel, and hence even 100 cores will have exactly 100x the performance of a single core. Same for rendering pixels or rasterizing them.
GPUs don't work the same as CPUs. You don't depend on kernel threads here to utilise multi cores.
29. vuyonc (Posts: 1014; Member since: 24 Feb 2014)
Bruh, multiplying the GPU cores by 3 while retaining the same frequency would burn a hole through your hand!
36. TyrionLannister (unregistered)
That's why samsung will tone it to around 700 MHz vs 900 on kirin. They will also lower the voltage leading to about the same power consumption while having around 3x the performance due to faster bandwidth and newer process node.
Ah the benefits of going wide. Power is proportional to f^2*V. So decreasing the frequency to 700 MHz and voltage linearly, it should consume only 40% more power than kirin. 650 MHz should reduce the gap to less than 10%.
55. vuyonc (Posts: 1014; Member since: 24 Feb 2014)
Bandwidth is also clockspeed dependent so it might not necessarily triple along with the number of cores. The overall performance/watt should still be high :) Can't wait to see how Samsung's 14nm LPP compares to 14nm LPE and 16nm FF+ nonetheless.
18. Arch_Fiend (Posts: 2501; Member since: 03 Oct 2015)
I called this, Qualcomm be on point with their GPU game bruh.
35. TyrionLannister (unregistered)
31. ibend (Posts: 4639; Member since: 30 Sep 2014)
lets wait battery life, temperature, and performance test from some legit reviewer (maybe next month :-/)
37. Shocky (unregistered)
Right, so Exynos 8890 has a weaker GPU than SN810 and Exynos 7422.
That makes perfect sense, could Antutu be any more unreliable ?
41. S.R.K. (banned) (Posts: 678; Member since: 11 Feb 2016)
Because it'sdownclocked 100mhz lower.
43. Shocky (unregistered)
I just tried this on my Galaxy S6 and after downloading 6.04 my score is down about 11k points than it was previously.
Antutu is inaccurate and inconsistent, it's insulting to even call this a benchmark.
50. jellmoo (Posts: 1595; Member since: 31 Oct 2011)
I just ran AnTuTu on my device. The score I got was: PurpleDolphinCheesePillow
Makes about as much sense as AnTuTu ever does.
51. aryanfr1 (Posts: 121; Member since: 24 Jan 2013)
Don't understand when Samsung was using power vr it was toe to toe with appi or better. Why they choose Mali? Power vr is always the best and that makes the appi GPU so dam fast.
52. Shocky (unregistered)
Don't know, would have to ask Samsung this question but like all things I would guess it comes down to cost one way or the other.
Samsung are apparently developing their own GPU, doesn't really explain why they would choose Mali but it's interesting all the same. Hopefully we will see something from them in the next few years, but then again nothing could come of this. :)