Jump to content


Photo

ATM7029: an ARM Cortex A9 or A5?


  • Please log in to reply
99 replies to this topic

#1 Tzul

Tzul

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 429 posts

Posted 10 January 2013 - 12:18 PM

I've heard rumors that the quad-core CPU of the Actions ATM7029 might not be an ARM Cortex A9 as claimed.
It could be an ARM Cortex A5, perhaps modified and improved...

What does this mean?
Well, the A5 is more energy efficient, but also older and slower per core and per clock (A9: 2.50 DMIPS, A5: 1.57 DMIPS).
Therefore, applications that don't take full advantage of all CPU cores might actually run faster on a real A9 dual core.


How can we find out if it is an A5 or A9?
Benchmarks should be able to tell. Also, if you have an ATM7029 based tablet, take a good file manager (doesn't even require root) and open the file /proc/cpuinfo.
Or from a terminal, you can run the command "cat /proc/cpuinfo" to display the contents of the file.

Here's what it looks like on an Amlogic AML8726-MX (dual-core Cortex A9):

Processor       : ARMv7 Processor rev 0 (v7l)
processor       : 0
BogoMIPS        : 2627.08

processor       : 1
BogoMIPS        : 2627.08

Features        : swp half thumb fastmult vfp edsp neon vfpv3
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x3
CPU part        : 0xc09
CPU revision    : 0

Hardware        : Amlogic Meson6 g06 customer platform
Revision        : 0020
Serial          : 000000000000000c

The "CPU part" number identifies the CPU type. The ARM Cortex CPUs should have 0xC0? with "?" being the type number, e.g. 9 for an A9 or f (hexadecimal) for an A15.
The ATM7029 supposedly reports 0xc05, thus an A5.

Can anyone here with a Hero II confirm or deny this?

Edited by Tzul, 10 January 2013 - 01:13 PM.

  • Arbalath likes this

#2 janesh

janesh

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 436 posts
  • Locationthe netherlands

Posted 10 January 2013 - 01:17 PM

From my hero ii that i almost sold.

Processor : ARMv7 Processor rev 1 (v7l)
processor : 0
BogoMIPS : 1114.46

processor : 1
BogoMIPS : 1114.46

processor : 2
BogoMIPS : 1114.46

processor : 3
BogoMIPS : 1114.46

Features : swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x0
CPU part : 0xc05
CPU revision : 1

Hardware : gs702a
Revision : 0000
Serial : 0000000000000000

Now my Onda V812


Processor : ARMv7 Processor rev 3 (v7l)
processor : 0
BogoMIPS : 1725.93

processor : 1
BogoMIPS : 1725.93

Features : swp half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x0
CPU part : 0xc07
CPU revision : 3

Hardware : sun6i
Revision : 0000
Serial : 00000000000000000000000000000000

And now my Ployer Momo8 IPS

Processor : ARMv7 Processor rev 0 (v7l)
processor : 0
BogoMIPS : 3214.95

processor : 1
BogoMIPS : 3214.95

Features : swp half thumb fastmult vfp edsp neon vfpv3
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc09
CPU revision : 0

Hardware : RK30board
Revision : 0000
Serial : 0000000000000000

Edited by janesh, 10 January 2013 - 01:37 PM.

  • Tzul likes this
On order, nothing, and am i am finished with Chinese cpu's, i prefer my exynos devices.
Favorites, in order of use, Samsung galaxy note 8 inch, Ramos W42, Ployer Momo11 quad exynos 4412.
Sold last months, PiPo U1, Teclast P85, PiPo S2, Teclast P85HD, Cube mini U30GT, Yuandao Vido N80IPS, Ainol Novo 10 Hero ii: very bad, Ployer Momo12, Onda V812, Yuandao Vido N80 RK3188, Teclast P88, Google Nexus 7-8Gb, PiPo M8pro RK3188, Chuwi V88 RK3188, Samsung galaxy tab 7.7, Ployer Momo8 IPS, Yuandao Window N70, BB playbook.
And: You know we all became mathematicians for the same reason: we were lazy.

#3 Tzul

Tzul

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 429 posts

Posted 10 January 2013 - 01:36 PM

Thanks!

So it's true, the ATM7029 has the 0xc05 ID as used for ARM Cortex A5 CPUs.
Yet another case of false advertising. The ID could be a mistake, but I doubt it.

By the way: don't read too much into the BogoMips values, they vary greatly depending on the current CPU clock speed.

Edited by Tzul, 10 January 2013 - 01:37 PM.


#4 janesh

janesh

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 436 posts
  • Locationthe netherlands

Posted 10 January 2013 - 01:51 PM

I allready sold my hero ii, because the device is not fast enough for me, but now i must think about it. :angry:

http://detail.tmall....&id=16768981007

Below a quote on a ainol TMALL website where they are listing 'cortex A9'.

NOVO10英雄四核搭载SMP对称高性能四核ARM cortex A9 family架构芯片ATM7029,最高频率达1.5GHz,整体性能较双核处理器提升了200%。四核3D GPU与一颗专有的独立2D GPU分离技术,展现领先同级的图形渲染实力,四核3D GPU帮助运行大型3D游戏,2D GPU更好地提升3D GPU在像素填充上的品质,让你穿梭于各类大型3D游戏或者办公、绘图软件之间游刃有余。

Edited by janesh, 10 January 2013 - 02:14 PM.

On order, nothing, and am i am finished with Chinese cpu's, i prefer my exynos devices.
Favorites, in order of use, Samsung galaxy note 8 inch, Ramos W42, Ployer Momo11 quad exynos 4412.
Sold last months, PiPo U1, Teclast P85, PiPo S2, Teclast P85HD, Cube mini U30GT, Yuandao Vido N80IPS, Ainol Novo 10 Hero ii: very bad, Ployer Momo12, Onda V812, Yuandao Vido N80 RK3188, Teclast P88, Google Nexus 7-8Gb, PiPo M8pro RK3188, Chuwi V88 RK3188, Samsung galaxy tab 7.7, Ployer Momo8 IPS, Yuandao Window N70, BB playbook.
And: You know we all became mathematicians for the same reason: we were lazy.

#5 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 10 January 2013 - 03:06 PM

Janesh ran Antutu 3 on Hero II & Onda V812. Hero II showed very strong performance for integer, float & RAM.
http://www.slatedroi...post__p__476881

ATM7029 still comes out with higher results for those even if you scale up V812's # by 20%. ie. V812 quad Cortex A7 @ 1 GHz vs ATM7029 quad Cortex A5 or A9 @ 1.2 GHz. = 20% clock speed difference.

The A5 is a weaker ARM chip and CPU performance would be lower. Only 2 explanations. #1. Cortex A5 & not A9 or #2. Actions screwed up the part value and put in A5 instead of A9.

CPU benchmarks will prove which of these for sure. Right now, Antutu 3 results by Janesh are pointing to #2.

Edited by Nickos-V, 10 January 2013 - 03:09 PM.


*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#6 Tzul

Tzul

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 429 posts

Posted 10 January 2013 - 03:55 PM

Janesh ran Antutu 3 on Hero II & Onda V812. Hero II showed very strong performance for integer, float & RAM.
http://www.slatedroi...post__p__476881

ATM7029 still comes out with higher results for those even if you scale up V812's # by 20%. ie. V812 quad Cortex A7 @ 1 GHz vs ATM7029 quad Cortex A5 or A9 @ 1.2 GHz. = 20% clock speed difference.

Performance doesn't scale linearly with clock speed across different models.
The published info says A5 = 1.57 DMIPS, A7 = 1.9 DMIPS, A9 = 2.5 DMIPS. But those are only estimates and can vary with different implementations.
Also, do Antutu scores scale linearly? Is a CPU with 2000 points twice as fast as one with 1000 points?


The A5 is a weaker ARM chip and CPU performance would be lower. Only 2 explanations. #1. Cortex A5 & not A9 or #2. Actions screwed up the part value and put in A5 instead of A9.

CPU benchmarks will prove which of these for sure. Right now, Antutu 3 results by Janesh are pointing to #2.

As I said, it's entirely possible that the CPU ID is an error. However, I think it doesn't bode well for Action's chips if they can't even get the ID register right. ;)
It's also possible that they took a vanilla A5 and tweaked their implementation to make it better and faster. Maybe the GPU is a GC1000 "plus" for the same reason.

Here's one more thing to consider:
In the cpuinfo dumps above you can see that janesh's and my Cortex A9 models (Rockchip and Amlogic) support VFP v3 but not VFP v4 (that's the vector floating point architecture).
The Cortex A5 (and Cortex A7) do support VFP v4. And so does the ATM7029.

What the official ARM manuals say:
Cortex A5 FPU
Cortex A9 FPU

Edited by Tzul, 10 January 2013 - 04:02 PM.


#7 robertnl

robertnl

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 36 posts

Posted 10 January 2013 - 05:01 PM

After Janesh posted this on a Dutch treatd i did put my Baidu skills to the work (they are bad :D).

Found this:

Propaganda written A9Family (Manufacturer marketing is going far)

Note, fleely translated partly with translate partly by my interpertaion after translating in 3 languaes

So what is inpliced here is that ATM has lyed and as Ainol stated A9 famaly you could call A7, A8, A15 famaly to as they do use the same instruction set.
So after some more research this could be true, Ainol should skip the Hero or give a explination fast as this will couse lots and lost of dmg to there brand.


In Benchmarks the difference between the to (A5 VS A9) is not so far away (less then a 100 points as you can see).
Posted Image

And the ATM score on 1.5GHZ? (not completely sure as translate screws up some things ;)!)
Posted Image


Sources for the above:: http://tieba.baidu.com/p/2084867680 and http://tieba.baidu.com/p/2084867680


Then i went to the Ainol forums and found a link to the Imp3 forums (will do some more research there later):
http://bbs.imp3.net/...880138-1-1.html

In short, certainly not A9.
If it is really A9, killed does not add a damn ambiguous family! - Chinese people is inexcusable traditional personality!


Gun God I just get a small waterway messages in the group, and true and false do not know
this atm7029 fact, is not arm architecture a5
but former mips by underlying convert analog into an arm like the actual performance close to a5.
That the bottom of the Actions chip false


So when i read this tread it gives me more or less the following options:
- MIPS structure with a conversion layer (highly doubt it)
- Cortex A5
- Modifyed cortex A5 (Like qualcomm did with Krait) So this way it could be part A9, part A5 but its highly doubtfull as it would take lots of effort and research to make something like this.
- it is A9 (doubt it)

Edited by robertnl, 10 January 2013 - 05:03 PM.

  • Tzul likes this

#8 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 10 January 2013 - 05:38 PM

OK. Good points Tzul.

vfp v4 is not supported on Cortex A9 with ARM's spec. Actions may have decided to make a "modified" Cortex A9 chip which includes vfp v4. ie. Cortex A9 + vfpv4. In order for software to make use of vfp v4 instructions, Actions would have to change the chip ID to A5 (or A7) because A9 does not support vfp v4. Or Actions made a 2nd error by including vfp v4 in the description. Or Actions has done a really amazing job tweaking the performance out of Cortex A5 architecture. Or Actions made Cortex A5 chip and found a way to "fool" Antutu benchmark to achieve better CPU results.

Whether Cortex A9 or A5 is really kinda irrelevant. What really matters is the CPU performance in the benchmarks. Presently, ATM7029 has a very strong showing in Antutu RAM, INT & FLOAT. Though we "should" test with other CPU benchmarks to confirm those to be true and accurate.

My N80 (RK3066) 2 x 1.6 GHz A9s w/Antutu 3
RAM: 1886
INT: 2520
FLOAT: 1825

Total CPU clock difference between Hero II & N80 = 50% (4.8 GHz VS 3.2 GHz)
We can see that ATM7029 scores very well against my N80. ie. does 43% better RAM score, 43% INT, 56% FLOAT. Notice that if you scale up my N80 results by 50% they would be very similar to ATM7029?

Take into account that Antutu is not super accurate either and chip design between RK3066 & ATM7029 are different too. ie. the results still seem "fairly" linear and are within acceptable differences

You missed my point before. Antutu "should" scale CPU results linearly for the same chip (ie. A31). A 1.2 GHz A31 should scale up linearly for CPU performance compared to A31 @ 1.0 GHz. I was scaling up the results for V812 (Cortex A7) & comparing those "scaled" V812 CPU results to ATM7029 Antutu results. The results "should" scale up linear based on clock speed when using the "same" ARM chip (ie. A31) because #1. the architecture is the same, #2. the chip design is exactly the same and #3 the same tests are being run. Imagine two Intel Core i3s (1st gen) chips. 1 @ 1.0 GHz & the other at 2.0 GHz. The only difference being their clock speeds. In theory, the CPU results would scale up linearly between the 1.0 & 2.0 GHz chips running the same tests. The 2 GHz chip should get 2X the CPU performance. ie. that would only make sense.

V812......V812 (@ 1.2 GHz)....Hero II
1 GHz.....1.2 GHz (scaled)....1.2 GHz
1547......1856.......................2690
2876......3451.......................3615
1830......2196.......................2845
*Notice how Hero II beats V812 (scaled). If ATM7029 is Cortex A5, then it should be scoring less and not more than scaled A31 (Cortex A7) chip.

My guess is that Actions created a Cortex A9 w/vfp v4 and changed the chip ID to Cortex A5 so software uses vfp v4 instructions. But, this all assumes and relies heavily on Antutu results being correct and not manipulated in anyway! More CPU testing will prove whether I am right or not with this "guess" or if Antutu was somehow fooled.


PS, the biggest problem I have seen with ATM7029 is 3D performance because they went with GC1000. ATM7029 chip would have been a winner if it had Mali400MP4 GPU. I would strongly advise people to avoid this chip just because of the very weak 3D performance.

*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#9 Tzul

Tzul

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 429 posts

Posted 10 January 2013 - 09:46 PM

So when i read this tread it gives me more or less the following options:
- MIPS structure with a conversion layer (highly doubt it)
- Cortex A5
- Modifyed cortex A5 (Like qualcomm did with Krait) So this way it could be part A9, part A5 but its highly doubtfull as it would take lots of effort and research to make something like this.
- it is A9 (doubt it)


Thanks for your contribution.
That's a good a summary of the possibilities there.




vfp v4 is not supported on Cortex A9 with ARM's spec. Actions may have decided to make a "modified" Cortex A9 chip which includes vfp v4. ie. Cortex A9 + vfpv4. In order for software to make use of vfp v4 instructions, Actions would have to change the chip ID to A5 (or A7) because A9 does not support vfp v4.

Yes, a modified A9 is possible, just like a modified A5. But as robertnl quoted, both are unlikely. Well, Qualcomm and Apple created their own Cortex A9 variants, does Actions play in the same league? They would certainly stand out in this list.
However, Actions wouldn't have changed the CPU part number for that reason (to indicate the presence of vfpv4). That might have been done decades ago, but nowadays CPUs have CPUID sort of commands or registers that indicate what features and capabilities are and are not supported (see MVFR1 and FPSID in the ARM manuals).


Whether Cortex A9 or A5 is really kinda irrelevant. What really matters is the CPU performance in the benchmarks. Presently, ATM7029 has a very strong showing in Antutu RAM, INT & FLOAT. Though we "should" test with other CPU benchmarks to confirm those to be true and accurate.

I agree, the name is irrelevant, the performance counts. The performance in real world use, not synthetic benchmarks. ;)
Still, they shouldn't lie to us customers. If it is a CPU based on ARM Cortex A5, then call it that (or "A5 plus" if you must) and point out that it doesn't have to hide behind A9s. But don't call it Cortex A9 if it isn't one.


My N80 (RK3066) 2 x 1.6 GHz A9s w/Antutu 3
RAM: 1886
INT: 2520
FLOAT: 1825

Total CPU clock difference between Hero II & N80 = 50% (4.8 GHz VS 3.2 GHz)
We can see that ATM7029 scores very well against my N80. ie. does 43% better RAM score, 43% INT, 56% FLOAT. Notice that if you scale up my N80 results by 50% they would be very similar to ATM7029?

Take into account that Antutu is not super accurate either and chip design between RK3066 & ATM7029 are different too. ie. the results still seem "fairly" linear and are within acceptable differences


You're right, the ATM7029 really does score well in RAM, int and float. Whereas the Allwinner A31 disappoints, except in int maybe. But that could just mean that Allwinner's A7 implementation sucks. It'd be interesting to see benchmarks of A7s by other implementers.

Anyway, I did a quick test with my Aurora 2 (AML8726-MX Cortex A9 dual-core). Forced it to 800 MHz (actually 792 MHz) and ran the Antutu 3 CPU test, then again at 1320 MHz.

      | 792MHz  |  1320MHz ||  792MHz * 1.67 || 792MHz * 2.02
------+---------+----------++----------------++--------------
RAM   |   1023  |     1590 ||  1708          || 2066
INT   |   1258  |     2089 ||  2101          || 2541
FLOAT |    881  |     1480 ||  1471          || 1780

CPU integer and floating point scores really are proportional to the frequency. But the RAM score certainly is not, and that makes perfect sense - the RAM doesn't actually get faster, only the CPU instructions accessing it do, and maybe the CPU's caches.

And while the int and float "792MHz * 2.02" values are reasonably similar to your RK3066 values, that doesn't necessarily have to be the case. Again, a customized Cortex A9 implementation (Qualcomm and Apple) could yield significantly different results. But Amlogic's and Rockchip's implementations seem to be comparable vanilla Cortex A9s.


By the way, seeing "4.8 GHz VS 3.2 GHz" makes me cringe. ;)
With a multi core CPU, you cannot simply do "number of cores" times "clock frequency" equals "virtual single core with equal performance". Because in reality, not everything is perfectly parallelizable and not every resource is actually multiplied (more on that here).

It works here because the benchmarks of Antutu are highly parallelizable - well, for int and float at least. That's to be expected, they're probably just number crunching without any dependencies, so there's no need for the code running on the different cores to communicate and synchronize, as you would have in most real world applications.

Tested it on my AML8726-MX by deactivating one core:

      | 1200MHz x1 | 1200MHz x2 || x1 * 2
------+------------+------------++-------
RAM   |  874       | 1444       || 1748
INT   |  964       | 1934       || 1928
FLOAT |  668       | 1338       || 1336



PS, the biggest problem I have seen with ATM7029 is 3D performance because they went with GC1000. ATM7029 chip would have been a winner if it had Mali400MP4 GPU. I would strongly advise people to avoid this chip just because of the very weak 3D performance.


I'd second that, but only if you want to play many 3D games (not everybody does). The ATM7029's CPU performance looks good, A5 or not, and it apparently has a nice video decoder onboard (there's a demo of a Hero II playing several videos simultaneously).

Edited by Tzul, 10 January 2013 - 09:48 PM.


#10 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 11 January 2013 - 12:35 AM

You need synthetic benchmarks otherwise you will never be able to measure and compare performance. =) For instance, gaming benchmarks are the best and most accurate but they test out CPU, RAM & GPU together. To get individual values for those requires running synthetic tests. Whether it be computing PI or testing with Antutu or CF-Bench, etc. Though relying on just a single benchmark is never good. Better to run and compare against a couple of different ones.

I still believe modified Cortex A9. So, they would not be lying then. =)
Though could be modified Cortex A5 but I doubt that because the performance in Antutu CPU tests looks really great. Finding Antutu 3 values for Cortex A5 chip and scaling them would help prove this.

Let's forget about A31 for now and just compare to RK3066 & 8726-MX instead.

INT & FLOAT scale up linearly with CPU frequency (in Antutu) as you have also proven. RAM not 1:1 but it does benefit from higher clock speed too. ie. 100% frequency increase = 100% increase to INT, 100% increase to FLOAT & 83-90% increase, when working with dual cores, to RAM.

Notice how ATM7029 scores similar to 729 * 2.87 for INT but much higher for FLOAT (792 * 3.23). ie. Action really improved the FPU performance on their chip probably by going with vfpv4. That is why their chip appears to be Cortex A9 w/vfpv4 to me. For ATM7029: INT compares to dual A9 @ 2.27 GHz and FLOAT compares to dual A9 @ 2.56 GHz.

      | 792MHz  |  1320MHz ||  792MHz * 1.67 || 792MHz * 2.02 || 792 * 2.87          || ATM7029
                                             (Dual A9 @ 1.6 GHz) (Dual A9 @ 2.27 GHz)   quad 1.2 GHz
------+---------+----------++----------------++--------------------------------------------------
RAM   |   1023  |     1590 ||  1708          || 2066          ||  
INT   |   1258  |     2089 ||  2101          || 2541          ||  3610               || 3615
FLOAT |    881  |     1480 ||  1471          || 1780          ||  2528               || 2845

You have proven, in your previous post, that INT & FLOAT scale linearly per frequency & per core in Antutu tests. In that case, for INT, we can do, 2 x 2.27 = 4.54 GHz Cortex A9 and 2 x 2.56 = 5.12 GHz for FLOAT. ie. INT comes very close to 4.8 GHz A9 and FLOAT results are similar to 5.1 GHz A9.

True, that creating a more unique chip design than the standard Cortex A9 one from ARM could change performance but only to a point. ie. 5-10% CPU performance difference is possible between different Cortex A9 chips @ same clock frequency. Only so much extra performance you can squeeze out of a certain chip architecture with extra tweaks.

By the way, seeing "4.8 GHz VS 3.2 GHz" makes me cringe. ;)

I have been working with SMP machines since 2003 so I know that is not true. ie. app first has to be multi-threaded otherwise only 1 core gets used. 2nd, even then, the app may only achieve maximum of 70-85% performance increase per additional core because of the app's code and/or OS. 2 cores does not mean 2X the performance. =) Though some OSes, like BeOS/Haiku, will get 95% efficiency (or higher).

Edited by Nickos-V, 11 January 2013 - 12:57 AM.


*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#11 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 11 January 2013 - 02:14 AM

PS, Antutu maybe wrong with the #s too. Vellamo Metal gives ATM7029 score of 251. My N80 gets 500 which is double in score. The Metal tests appear to be CPU & memory tests & maybe disk access too.
http://www.quicinc.com/vellamo/

It makes no sense why ATM7029 scores very well in Antutu CPU & RAM tests and poorly in Metal benchmark. That is why other CPU benches are needed to know what is really going on here. ie. two benchmarks and both saying different things. Believe Antutu, then looks like Cortex A9. Believe Metal, then Cortex A5. Who should we believe?

I still tell people to avoid ATM7029 because should always buy a balanced computer/tablet. ie. RK3066 has way better balance than ATM7029 between CPU & GPU power.


Edit: I probably would believe Vellamo over Antutu benchmark because made by Qualcomm. Also, many newer devices benched in Metal get 450-700 score. Like Note II with score of 619.
https://play.google....nZlbGxhbW8iXQ..

So, very likely that ATM7029 is really quad Cortex A5 like you said. More CPU tests have to be run to know for certain though.

Edited by Nickos-V, 11 January 2013 - 02:49 AM.


*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#12 Ali DSD

Ali DSD

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 160 posts

Posted 11 January 2013 - 06:10 AM

PS, Antutu maybe wrong with the #s too. Vellamo Metal gives ATM7029 score of 251. My N80 gets 500 which is double in score. The Metal tests appear to be CPU & memory tests & maybe disk access too.
http://www.quicinc.com/vellamo/

It makes no sense why ATM7029 scores very well in Antutu CPU & RAM tests and poorly in Metal benchmark. That is why other CPU benches are needed to know what is really going on here. ie. two benchmarks and both saying different things. Believe Antutu, then looks like Cortex A9. Believe Metal, then Cortex A5. Who should we believe?

I still tell people to avoid ATM7029 because should always buy a balanced computer/tablet. ie. RK3066 has way better balance than ATM7029 between CPU & GPU power.


Edit: I probably would believe Vellamo over Antutu benchmark because made by Qualcomm. Also, many newer devices benched in Metal get 450-700 score. Like Note II with score of 619.
https://play.google....nZlbGxhbW8iXQ..

So, very likely that ATM7029 is really quad Cortex A5 like you said. More CPU tests have to be run to know for certain though.



I just saw the add of Novo 7 Venus on caveat emptor.com, its showing :
CPU ATM7029 (Cortex-A9)
Graphic Hardware Mali-400, 1080p decoding,Supporting OpenGL ES2.0, ES1.0, OpenVG1.1

So Is this ATM7029 the different from Hero II , because it have Mali-400. I am confused, Nikos and Tzul, please give your opinion.

Thanks

#13 AC Slater

AC Slater

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 180 posts
  • LocationAustria

Posted 11 January 2013 - 07:38 AM

@ Ali: there has been very poor marketing, with lack of precise descriptions and only vague fingerpointing what's inside.

First it looked as if they would use ARM Mali-400 of even Mali-450. Then it was supposed to be the Vivante GC1000, but it had 4 cores and dedicated core for 2D, so the 'GC1000 Plus' was born, whatever that really is.
Since there was no clear answers, they could even have gone for somekind of PowerVR-solution.

Now people are actually getting their tablets in, it sure seems to lack graphic processing in the 3D department. It's being outperformed by the RK3066 SoC which uses Mali-400MP quadcore.
So using the power of deduction, you would have to agree it is in fact not a Mali, nobody ever claimed something about PowerVR, so lets skip that one and assume it is in fact a Vivante GC1000 Plus variant.
Though even that is not guaranteed, because on the tablet self, it says it contains a 'S5 Multicore Engine' build/sold by 'Advanced Graphics Corporation'.

What I can tell you in short is so much:

- Never go to a reseller/webstore if you want 100% correct information. A great deal of the time they have no clue what they're actually selling and just copying information from 'somewhere' without checking.
- The Ainol Tmall-stores advertised the Hero II with 'Mail400MP4' processor
Posted Image
Notice the misspelling too, not really to be trusted
- Ainol.com has a quadcore-page: http://www.ainol.com...olquad/page.htm , it says 'GC1000 plus'
- The tablet itself shows the GPU as 'S5 Multicore Engine' by 'Advanced Graphics Corporation'

So basicly, I don't think anybody can give you a 100% guaranteed correct answer. :pardon:
MyTablets: PiPO Max-M8 Pro & M7 Pro 3G | Ainol Novo7 Elf II (cracked screen), Fire & Novo10 Hero II (dead) | Archos 101IT G8 (weird lines on screen)
MySmartphones: Jiayu G4T Adv | FAEA F2 | Neomobile N002 (N02-M)
MyAndroidTV: MINIX NEO X5 | Mele F10
MyGaming devices: Archos GamePad | OUYA
Previously owned: Jiayu G2 & G3 | Neomobile N003 | HTC Desire Z | FNF ifive X2 | Ainol Novo7 Aurora II | Hyundai A7HD
Proud other gear: ASUS N56VZ laptop | Panasonic Lumix DMC-GH2 & LX-7 camera's | Earson ER-151 BT speaker | Brainwavz HM9 headphone | BOSE QC15
Soon: -
Interested in: powerful tablets (8.9"/9.4") & smartphones (≤ 5.0")
My blogposts:
Buying from Taobao / Tmall (directly from manufacturers, thru an agent) <<

#14 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 11 January 2013 - 08:52 AM

So Is this ATM7029 the different from Hero II , because it have Mali-400.

ATM7029 is one chip with certain specs. You cannot have ATM7029 with Mali400 and another with GC1000. Some sellers are lying or do not know the chip specs. Chinese sellers only care about selling goods. Some tablet makers are to blame too because I have seen Ainol & Ramos say Mali400 GPU which is not true.

Action very likely re-branded Vivante's GC1000 GPU. Half or more of the sites say GC1000 for GPU. If you look here you will see the GPU comes up as:
Advanced Graphics Corporation
S5 Multicore
http://www.glbenchma...re&testgroup=gl

I have never heard of Advanced Graphics Corp. before ATM7029. That leads me to believe a re-brand of another GPU.

Here is GlBenchmark compare of Novo10 Quad VS Cube U30GT-H:
http://www.glbenchma...hnology U30GT-H

Many tests fail with newer GLbenchmark but can at least see:
                          Novo10 Quad     U30GT-H
2.1 Egypt Classic            
Offscreen (1080P)            2529          6642

2.1 Egypt Classic            3456          6797
  Onscreen
ie.
in offscreen mode, at 1920x1080, U30GT is 2.63X (163%) more powerful at HD resolution
in onscreen mode, 1280x800, U30GT is 1.97X (97%) more powerful

I wish there were more GLBenchmark "game" type benches to look at but that one will just have to do.


PS, and now it looks like the CPU could very likely be Cortex A5 too which is way weaker than Cortex A9. This will add another very good reason to avoid ATM7029 chip. ie. weak GPU & CPU.

Edited by Nickos-V, 11 January 2013 - 08:55 AM.


*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#15 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 11 January 2013 - 09:16 AM

If you want strong ARM chip today go with one of:
RK3066, Exynos Quad or A31

In few months time, you can add RK3188 to that list.

Tablets with those ARM chips are the ones worth buying. Forget about ATM7029 & 8726-MX chips. The ARM chip is the most important part. It determines performance. 2nd is the tablet brand.

Edited by Nickos-V, 11 January 2013 - 09:16 AM.

  • janesh likes this

*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#16 Ali DSD

Ali DSD

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 160 posts

Posted 11 January 2013 - 09:45 AM

Thank you Nickos and AC Slater, the information about GPU is very vague on different web sites.
Currently I have bought RK3066, and its really very powerfull then the Amlogic 8726. I would wait for RK3188..
Thanks

#17 janesh

janesh

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 436 posts
  • Locationthe netherlands

Posted 11 January 2013 - 10:16 AM

If you want strong ARM chip today go with one of:
RK3066, Exynos Quad or A31

Right Nickos, my thoughts, but with my experience with RK3066 and now with A31, i will never buy a device with this chips and a retina screen.

I just did the vellamo benchmark with three devices, here the results.

Onda V812, Metal 372, HTML5 1120
Hero ii, Metal 254, HTML5. 942
Ployer momo8,Metal 514, HTML5 1563

Also disappointing for me since i own the hero ii is the small amount of Ram, listed as 1 Gb, and only 718 Mb available, while the ployer momo8 and teclast P88 both have 876 Mb.

For me it was clear after one day of use, i am going to sell this device.

Edited by janesh, 11 January 2013 - 10:29 AM.

On order, nothing, and am i am finished with Chinese cpu's, i prefer my exynos devices.
Favorites, in order of use, Samsung galaxy note 8 inch, Ramos W42, Ployer Momo11 quad exynos 4412.
Sold last months, PiPo U1, Teclast P85, PiPo S2, Teclast P85HD, Cube mini U30GT, Yuandao Vido N80IPS, Ainol Novo 10 Hero ii: very bad, Ployer Momo12, Onda V812, Yuandao Vido N80 RK3188, Teclast P88, Google Nexus 7-8Gb, PiPo M8pro RK3188, Chuwi V88 RK3188, Samsung galaxy tab 7.7, Ployer Momo8 IPS, Yuandao Window N70, BB playbook.
And: You know we all became mathematicians for the same reason: we were lazy.

#18 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 12 January 2013 - 12:26 PM

I just did the vellamo benchmark with three devices, here the results.

Onda V812, Metal 372, HTML5 1120
Hero ii, Metal 254, HTML5. 942
Ployer momo8,Metal 514, HTML5 1563

Hi Janesh. Thanks for your benches.

You can see that ATM7029 is the worst of all 3 of those chips. RK3066 is the best. It also shows how ATM7029 does poorly in both CPU & GPU sides. Add to that, losing 166 MB of RAM over RK3066.

I believe Metal runs single threaded tests but cannot say for sure. If metal is single (or dual) threaded, that would explain why momo8 does so well compared to the other two tablets. But, even if multi-threaded, Cortex A5 & A7 would still lose because slower architectures, slower CPU speeds and because of efficiency.

HTML5 tests out GPU, CPU & Network stack. Metal is CPU, RAM & disk. Good seeing those tablet comparisons.

Many apps are dual or single threaded so a fast dual core Cortex A9 (or better, A15) will run better than a lower clocked quad CPU tablet! That is why RK3066 will give better CPU performance over A31 most of the time.

*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).

#19 janesh

janesh

    Advanced Member

  • Jr. Member
  • PipPipPip
  • 436 posts
  • Locationthe netherlands

Posted 13 January 2013 - 01:14 PM

http://tabletrepubli...-up-3024-2.html

Interesting page and topic on tabletrepublic, where some predict month agoo what is the matter with ATM7029.
On order, nothing, and am i am finished with Chinese cpu's, i prefer my exynos devices.
Favorites, in order of use, Samsung galaxy note 8 inch, Ramos W42, Ployer Momo11 quad exynos 4412.
Sold last months, PiPo U1, Teclast P85, PiPo S2, Teclast P85HD, Cube mini U30GT, Yuandao Vido N80IPS, Ainol Novo 10 Hero ii: very bad, Ployer Momo12, Onda V812, Yuandao Vido N80 RK3188, Teclast P88, Google Nexus 7-8Gb, PiPo M8pro RK3188, Chuwi V88 RK3188, Samsung galaxy tab 7.7, Ployer Momo8 IPS, Yuandao Window N70, BB playbook.
And: You know we all became mathematicians for the same reason: we were lazy.

#20 Nickos-V

Nickos-V

    Advanced Member

  • Hero Member
  • PipPipPip
  • 2,022 posts

Posted 13 January 2013 - 02:37 PM

http://tabletrepublic.com/forum/latest-tablet-news/ainols-quad-core-line-up-3024-2.html

Interesting page and topic on tabletrepublic, where some predict month agoo what is the matter with ATM7029.

Actually a chip designer told Frank the ATM7029 was using Cortex A7. Though it really turned out to be Cortex A5. A7 would still have been good but A5 is not. A5 is very weak.
http://tabletrepubli....html#post30151

Though I am very disappointed how Antutu gives this chip a very high score for INT & FLOAT tests with an A5 architecture. Antutu really surprised me because it outputs a really high total score for the Hero II even though it has weaker CPU & GPU. Go figure! That is why I have never been a big fan of Antutu benchmark.

Edited by Nickos-V, 13 January 2013 - 02:39 PM.


*MINIX NEO X5 (RK3066) - OK. Not fun because uses mouse cursor.
*YUANDAO N80 DUAL (RK3066) - Very good
*CUBE U30GT (RK3066)- Defective; buttons broke in 2-3 weeks; Bad quality control!!! I will never buy a Cube again!!! - Very good while working
*WOPAD I7 (RK2918)- About OK. Sluggish because has weaker ARM chip (RK2918) & less RAM (512 MB).




1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users