Please do not use arm cortex-a15 for pandora2!!!!!!!!


H

homejogger

Still Fresh
Joined
Oct 11, 2013
Messages
6
Please do not even touch the arm cortex-a15.

It is the worst design that arm has ever done!

Let's take tegra4 as an example.

This powerful chip is made by 28nm Lp, it has quard-core arm a15 cpu, and a Gefore ULP gpu.

It seems to be nice to have such an powerful soc, but its power consumption is more than 13W !(11W caused by cpu, 2.6W produced by gpu).

That's why Nvidia Shield has to use a fan to cool down the system. The powerful nvidia shield could only run 3D games(tegra4 optimized version) for around 5 hours with 7350mAh battery!

The arm cortex-a15 would kill the battery dramatically, unless it has been made by 20nm (or advance).

Please take it seriously by deciding to use arm a15 for pandora2 or not.

From my point of view, the Intel bay trail could be an optional choice if we could get the gpu driver in Ubuntu system.

The TDP of bay trail chips are mainly under 3.5W compare with SDP is only 2.5W. And, as it is a x86 based cpu, there will be much easier for us to get softwares for pandora.

Why we chose pandora ?

1. open source

2. powerful chip

3. 10hours battery life

4. multisystem

5. affordable price

As a soc could satisfy the above conditions, we will not care if the cpu is arm based or not, right ?

Thanks for your view.
 
And where do you get the 11W for the CPU from? Various measurements showed that the perf/W is a lot better than with our current one (which means less power usage for the same speed we have right now), and even fully using the SoC (which is GPU and CPU) peaks at 8W with all cores (100 CPU usage to my knowledge)

And you usually don't need the full processing powers (with all CPU cores at once), so most of the time the power usage should be 1-2 W or even less.

I'd love to see where you got these measurements from, could you post a lin?
 
Heh, interesting enough that the Shield has a heat sink and fan. The 3 videos and 2 articles I recently consumed on the console didn't mention that at all.
 
kuru said:
Heh, interesting enough that the Shield has a heat sink and fan. The 3 videos and 2 articles I recently consumed on the console didn't mention that at all.
Click to expand...
Heat sink is suggested for all A15 SoCs, the fan is only needed to get out the highest possible clock speed, it's not needed for tablets or smartphones.

Actually, 7350mAh for impressive 3D games with the Shield is not that bad.

The Shield has impressive speakers (the amp and speakers probably need a lot of power), AND a fan running.

Both these components need quite a bit of power (especially a fan), with the screen needing quite a bit of power as well.
 
Paging Dr. Exophase. Paging Dr. Exophase. Please report to thread.
 
homejogger said:
Please do not even touch the arm cortex-a15.

It is the worst design that arm has ever done!

Let's take tegra4 as an example.

This powerful chip is made by 28nm Lp, it has quard-core arm a15 cpu, and a Gefore ULP gpu.

It seems to be nice to have such an powerful soc, but its power consumption is more than 13W !(11W caused by cpu, 2.6W produced by gpu).

That's why Nvidia Shield has to use a fan to cool down the system. The powerful nvidia shield could only run 3D games(tegra4 optimized version) for around 5 hours with 7350mAh battery!
Click to expand...
... Your figures contradict.  Lithium Polymer batteries tend to be rated at around 3-4Volts.  You're implying voltages that'd probably make a Li-Poly battery explode in a rather pyrotechnic fashion.

Engadget and the Newegg spec sheets places the battery at 28.8W-hr, and Anandtech found battery life to by roughly 6 hours running a GL benchmark.  That puts power consumption for the entire unit, including the screen, fan, etc., at 4.8 Watts when running said benchmark.  The TDP associated with the entire SoC can't be set at over 4 Watts based on that.

During 720p playback, based on Anandtech's numbers, the entire system consumes 1.55 Watts with a life of 18.63 Hours, which speaks to the screen not exactly being a power consumption lightweight.  I expect 5 hours would include wireless antenna and speaker power consumption, assuming you obtained that from a legitimate source.  Personally I don't own a Shield, so I don't know what's normal for it.

The arm cortex-a15 would kill the battery dramatically, unless it has been made by 20nm (or advance).

Please take it seriously by deciding to use arm a15 for pandora2 or not.

From my point of view, the Intel bay trail could be an optional choice if we could get the gpu driver in Ubuntu system.

The TDP of bay trail chips are mainly under 3.5W compare with SDP is only 2.5W. And, as it is a x86 based cpu, there will be much easier for us to get softwares for pandora.
Click to expand...
Uh-huh. I think you need to question whatever crowd has been placing those thoughts in your head at little more critically.
 
Last edited by a moderator:
Jebe

A GL-bench doesn't have to be CPU intensive.

Hardware accelerated video playback definitely isn't.

X86_64 is top lel on scaling down to the idle levels of ARM, over time your problem is that drain, not what you do in access of it.

homejogger

You can scale down the processor and still have it be more power efficient, the need to stress-test the system is merely an option, not something you run into on a regular use-case.

If you really want to you could pair it with a low-end alternative core for real power saving.

Implementing an active cooling design for 13W is just not clever.
 
Last edited by a moderator:
I dont think amy of you need be concerned. Its well known that ED has 100% committed to a 4K screen, Mini SD and an x64 processor running Windows 8.1
 
Let's not put the pandora and shield in the same paragraph let alone same sentence when talking specs, ok paco? And I don't even own either handheld. Just being objective here. One is made by a multi billion dollar tech company, the other by dedicated individuals.

Also:

homejogger said:
5. affordable price
Click to expand...
Yeah, not really.
 
Maize16 said:
homejogger said:
5. affordable price
Click to expand...
Yeah, not really.
Click to expand...
Since everyone is buying the Ghz-Pandora, people started to underestimate the classic and rebirth edition.

Both are fast enough for most of our applications/games and if you overclock a bit, the difference in performance is not that big.

I had a classic and now have a ghz-edition.

The classics are affordable by now and way better than people think,

rebirth even got more RAM and so on.
 
Last edited by a moderator:
Klumpen said:
Maize16 said:
homejogger said:
5. affordable price
Click to expand...
Yeah, not really.
Click to expand...
Since everyone is buying the Ghz-Pandora, people started to underestimate the classic and rebirth edition.

Both are fast enough for most of our applications/games and if you overclock a bit, the difference in performance is not that big.

I had a classic and now have a ghz-edition.

The classics are affordable by now and way better than people think,

rebirth even got more RAM and so on.
Click to expand...
I don't fault Pandora proprietors for the high price tag. They are not a mega corporation that can price to sell. Pandora commands a premium and most of its charm comes from the fact that it was made by small business. If it was slightly less expensive I think I would have one by now.

Sorry if this is OT.
 
I got my first fix because of a used classic-edition (for over 300€ back in 02/2012, should be way less expensive by now).

Maybe you get a cheap classic somewhere here, good luck!

Regarding the topic:

DaMummy said:
Paging Dr. Exophase. Paging Dr. Exophase. Please report to thread.
Click to expand...
If he thinks it is just silly, he may drop this word here. ^^
 
Last edited by a moderator:
Klumpen said:
I got my first fix because of a used classic-edition (for over 300€ back in 02/2012, should be way less expensive by now).

Maybe you get a cheap classic somewhere here, good luck!

Regarding the topic:

DaMummy said:
Paging Dr. Exophase. Paging Dr. Exophase. Please report to thread.
Click to expand...
If he thinks it is just silly, he may drop this word here. ^^
Click to expand...
 I just looked at Ed's site and judging by words, he is still putting them together? I sent him a pm. Most of the emu I like is gba and earlier. I also prefer clam shell to open faced so other handhelds do not interest me.
 
Last edited by a moderator:
Quad core a15 might not be the answer but neither is a dual core or quad core a7. Both on opposite ends of the power/performance spectrum. The quad core big.little a7/a15 looks to be the best option when you can have power when you need it and long life when you don't. Obviously in the smallest manufacturing process available at the time.
 
Last edited by a moderator:
If I'm going to say things it'll just be things I've said in other threads multiple times, but whatever.

I don't think the perf/W characteristics of a plain Cortex-A15 solution are ideal either, but people are blowing it way out of proportion. Outside of the big "citation needed" for that 11W for the CPU cores alone in Tegra 4, even if that number is representative it doesn't give a full picture. Maybe you have to bump the core voltage tremendously to hit the last couple hundred MHz. Is it still so awful if you only need half as much power for 1.7GHz? What if you only need a fourth as much power for 1.5GHz? And you need to look at it under realistic conditions because there are few real world scenarios where a user of a Pandora like-device will want to max out all four cores. You need a full perf/W curve to evaluate this. It's a face that even Tegra 4 is showing up in phones, and while the performance levels are lower than Shield they're not dramatically lower.

BayTrail-T will have better perf/W for sure, probably much better. I've argued that too. But that doesn't mean a Cortex-A15 solution is totally unusable.

Oh and Tegra 4 isn't made on TSMC's 28LP process, it's made on 28HPM. And Shield only ships with a 10W power supply so if the CPU cores alone are allowed to draw 11W nVidia really botched things..
 
Last edited by a moderator:
monstercameron said:
tegra 4 with 
Click to expand...
25WHr battery and 2h34m runtime equates to ~9.75W power high load and ~4.4W under light load [including platform power]. Also for an ice storm score of [total, graphics, physics] 10400, 10200, 11600 [NB: the icestorm numbers seem low compared to the shield but I found a similar result here http://www.laptopmag.com/reviews/tab...tebook-x2.aspx ]http://www.notebookcheck.net/Review-....103282.0.html

Quotes from your links:

For minimum runtime, we set the display to maximum brightness with active wireless and ran Stability Test 2.5 until automatic shutdown after about 2.5 hours.
Click to expand...
That's the minimum runtime with a 10,1" FullHD LCD with full brightness AND wireless active and turned on.

You are aware that this screen probably needs A LOT more power than the Tegra itself, do you?

From your other link:

The SlateBook x2 comes with two power sources, a 3,375 mAh battery in the tablet portion of the device and a separate 2,960 mAh battery in the base. Between the two batteries, the x2 lasted 9 hours and 16 minutes. That's pretty good endurance, but the slate itself lasted 6 hours and 4 minutes.
Click to expand...
That's 9 hours with normal usage (probably Wifi switched on as well), so far away from the 2 hours you mention...

Anyways, the battery life with such devices doesn't say anything about the power usage of the Tegra4.

All it says is that the device needs a bit of power, but that doesn't mean much.

You need to compare the SoCs, not the full devices.
 
You must log in or register to reply here.

Similar threads

_wb_
Overview of SoC options
16 17 18
Replies
348
Views
40K
urisma
U
Mr_Loon
What would the world and his dog think if P2 used an x86 SOC? (but not mini SD) ;)
9 10 11
Replies
200
Views
17K
ingoreis
ingoreis
D
I.mx51 Arm Cortex Soc
Replies
3
Views
4K
Exophase
E
Orz
My Thoughts on the Pyra
3 4 5
Replies
91
Views
11K
levi
levi
C
Another Linux on ARM user (not OpenPandora)
Replies
19
Views
4K
notaz
notaz
Back
Top