[Tzul]

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)<br />
processor       : 0<br />
BogoMIPS        : 2627.08<br />
<br />
processor       : 1<br />
BogoMIPS        : 2627.08<br />
<br />
Features        : swp half thumb fastmult vfp edsp neon vfpv3<br />
CPU implementer : 0x41<br />
CPU architecture: 7<br />
CPU variant     : 0x3<br />
CPU part        : 0xc09<br />
CPU revision    : 0<br />
<br />
Hardware        : Amlogic Meson6 g06 customer platform<br />
Revision        : 0020<br />
Serial          : 000000000000000c<br />

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?

 

[Tzul]

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.

 

[Tzul]

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

 

[Tzul]

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).

 

[Tzul]

Does that mean that Rockchip RK3066, 1.6GHz, Cortex A9 dual core is better choice than ATM7029, Cortex A9 Quad core 1.5GHz, Vivante GC1000 GPU.

Yes, that's what it means.
A Rockchip RK3066 will outperform an ATM7029 in most use cases, because it has a faster CPU (even though it's just dual core) and a better GPU.

However, it's your decision.
If you like the Venus better for some reason, then go for it.
Just be aware that even though it's sold as a Cortex-A9, it's actually Cortex-A5. And that the Vivante GC1000 is rather weak and has compatibility problems in some games.

 

[Tzul]

No idea how Antutu gives it such a high score when other benchmarks do not. I am disappointed that Antutu ranks it very high. Gives false illusion that it is really great when it is not.

So, Antutu Benchmark was just updated to version 3.1.1.
And I've been told that the Hero 2's score dropped from 12000 to 6200 or so. Can somebody confirm?
My Aurora 2 (dual-core A9) still gets 9200.

Maybe Antutu fixed whatever caused the ATM7029 to score so high in RAM/CPU. Unless the new Antutu version actually has a bug there. But I doubt that, as it doesn't contradict other benchmarks anymore.

 

[Tzul]

Update on the current situation:

It looks like Actions / Ainol removed the "CPU part" line from /proc/cpuinfo in their latest firmwares!
A bit late for cover up tactics, makes them only look worse.

Some background on this:
ARM CPUs have a couple of ID registers, the CPU part number (c05 for Cortex-A5, c07 for Cortex-A7, etc.) is given by the Main ID Register (MIDR).
According to ARM's documentation, the MIDR is "only accessible in privileged modes", which probably means that only the kernel can read it, Android apps can't (maybe with root?).
So, that's why there's /proc/cpuinfo, a virtual file used as an interface to query the CPU info from the kernel. The source code responsible for generating the file's content is located in /arch/arm/kernel/setup.c which uses the "read_cpuid_id()" function from /arch/arm/include/asm/cputype.h to get the MIDR value.

Here's the relevant code from Amlogic's kernel source:

seq_printf(m, "CPU part\t: 0x%03x\n", (read_cpuid_id() >> 4) & 0xfff);

Actions most likely removed / commented out this part. Once they release the source code for the Hero II, Venus, etc., custom firmware developers should definitely take a look at this and undo the tampering.

PS:
An admin on Ainol's official forum has written this related post. Is anyone here fluent enough in Chinese and English to give a summary? Machine translation isn't quite good enough yet.

 

[Tzul]

The Amlogic chip is a 1.5ghz part that Ainol underclock because their implementation of it gets too hot. I don't really blame Amlogic for that, its still their highest supported speed if you give it an opportunity to cool itself and enough juice.

This "too hot" card is often pulled as an excuse. Is there any proof for that? I'm guessing that the absence of 1.5GHz on the AML8726-MX has more to do with stability and power consumption than with heat. According to its quick reference manual, the AML8726-MX's CPU may run with "up to 1.5GHz" and may get quite hot.

They haven't even listed an absolute maximum temperature there. The "recommended" maximum is 125°C. However, that's the internal junction temperature. The outside / ambience of the chip must always be quite a bit less than that during operation, due to imperfect heat conductivity.
Does Ainol even actually make the firmware? I've had the impression that it's done by Amlogic, especially the lowlevel stuff. In that case, Amlogic would underclock their implementation themselves.

Anyway, Feiyu's latest "happy spring festival" firmware for Ainol Amlogic tablets has 1.5GHz unlocked, and it appears to run fine. On some tablets at least. There could very well be copies of the AML8726-MX that just aren't stable at 1.5GHz and reasonable voltages in their given tablet environment.
I'm sure the AML8726-MX could do 2GHz and more if you crank up the voltage and battle the heat with liquid nitrogen, but does that mean they should market the chip as 2.0GHz?

 

[Tzul]

For those that want to 3D game. Are they supposed to play 5 mins. at a time to allow 8726-MX to stay cool??? A chip should be able to run stable and slightly hot for at least 1 or more hours while 3D gaming. My desktop, laptop & RK3066 tablet can all do this but 1.5 GHz 8726-MX tablet certainly cannot. I hear it gets extremely hot while 3D gaming @ 1.5 GHz.

You heard wrong.
I know what "extremely hot" is. I once had a notebook with AMD Turion dual core CPU that reached 95°C under load, so hot that it started throttling the frequency to reduce heat.

As I mentioned before, I'm just testing Feiyu's "happy spring festival" JB 4.1.2 release on my Aurora 2.
CPU at 1512MHz, GPU at 333MHz or 400MHz. Antutu 3.1.1 score of 9942, Vellamo Metal 580. It does get a bit warm under stress, but so far nothing I'd call "extremely hot". And it ran absolutely stable so far, no crashes yet.
Also, keep in mind that most games are either CPU or GPU bound. It is very unlikely to come across a game that pushes both CPU and GPU to their limits at the same time. I've been running 10 minutes or so of the Epic Citadel benchmark, for example, which clearly is limited by GPU perfomance. Whether the CPU runs at 1320MHz or 1512MHz doesn't make a noticeable difference there.

The power consumption is quite alarming, though. I measured battery discharge currents of nearly 1.7A and power consumption of over 6W - with the screen at maximum brightness (which I barely use in practice).

 

[Tzul]

The Action ATM7029 GPU is Gc1000 Plus from Vivante, Not really same as GC1000. Quite hard to say whether ATM7029 utilize the a5/a9.

Have you somehow missed the first two pages of this topic?
It's crystal clear that the ATM7029 has a Cortex-A5 as CPU!

 

[Tzul]

It's known that eaven Amlogic overclocks when it notices Antutu is run.

Eaven the startup page of Antutu says this tablet is cheat.

Amlogic does not "overclock". It basically switches to performance mode when certain benchmarks are detected. I don't consider this a real cheat as you can activate performance mode yourself any time. As I said elsewhere, even Windows does this - it automatically switches to the "high performance" energy profile whenever the "Windows Experience Index" (WinSAT) benchmark is run.

And Antutu's new red warning message "your phone is fake" is just caused by wrong build.prop impersonation/fingerprint settings, as far as I know.

 

[Tzul]

I'm not convinced that Actions was somehow cheating on the Antutu benchmark, even though the GC1000 scored low on other benchmarks/games.

Actions are already cheating by selling their A5 CPU as "Cortex-A9 family". If Antutu attests the ATM7029 a CPU or GPU performance that it shouldn't have according to other benchmarks and real-world apps, then there's clearly something fishy going on...

 

[Tzul]

ATM7029 is ARM Cortex A5 SoC, it's 99% becouse on the site of Actions there is

- Supporting VFP v4(Vector Floating Point v4) architecture and compliant with the IEEE 754 standard for floating-point calculation

( http://www.actions-semi.com/en/productview.aspx?cat=98&id=108 )
And Cortex A9 can have ONLY VFPv3, then Cortex A5 can have VFPv4, but in allmos all devices it hav VFPv3
http://www.arm.com/products/processors/cortex-a/cortex-a5.php

We have discussed that fact back on page 1 already.

 

[Tzul]

One more piece of evidence:

Actions writes on their ATM7029 website that it has an "In-order pipeline with dynamic branch prediction equipped".
Well, a Cortex-A9 has a superior out-of-order pipeline, says ARM.

So if anybody still believes that the ATM7029 could be an A9 with a unique VFPv4 upgrade: forget it.
It'd be ridiculous to think that Actions downgraded the pipeline to a slower in-order design.

The Cortex-A5 on the other hand does have an in-order pipeline.

 

[Tzul]

Best test is to open it & see for yourself....Its very easy, opening from sides. I did & found my has 7025, I have AINOL novo 7 Venus...The question is does A9 means its should have 7029? & is 7025 a A5 chip? A5 & A9 are more of architecture specifications...

No, 7029 does not mean A9 just because it ends with a 9. The ATM7025 is not listed on Actions' website. According to all we know, it appears to be a identical to the ATM7029 except that it doesn't support higher CPU frequencies. So you could say that ATM7025 = ATM7029 lite. Both have a Cortex-A5 quad core CPU. See this thread.

 

[Tzul]

Very good question. Apparently not too many people care. Chinese tablet manufacturers have always exaggerated their specs, but this is a new low.

Note that Actions doesn't call the CPU a Cortex-A9, but "A9 family" (if they manage to spell it properly). That way, they might be on the safe side legally and still be able to fool customers.

 

[Tzul]

Is this really a case of false advertising? If so, I will demand for a refund. If they won't allow, I will take legal action. What do you guys think?

I'd say it is false advertising, but I'm not a lawyer.
Is this the Cherry Mobile packaging? The prominent "A9 Quad core" logo is really audacious.
Most dealers probably copy & paste the manufacturer's info without thinking about it, so they're not entirely to blame. But I also wouldn't call them innocent - they should care about what exactly it is they're selling.

 

[Tzul]

Maybe Actions Semi. hid CPU part for ATM7029?

Yes, they modified their kernel sources. Hard to believe it's been a year already...

I'd be surprised to see any ATM7029 based tablet that still shows CPU part data in /proc/cpuinfo.