Power Consumption

[chaosfx]

something like a turbobutton didn't exist

with the button disconnected the cpu would just go fullspeed so it technically underclocked.

 

[Monk]

something like a turbobutton didn't exist

with the button disconnected the cpu would just go fullspeed so it technically underclocked.

You are quite, QUITE correct. The "Turbo" button was so labelled to avoid connotations of incompatability and provide a feeling of raw SPEED (which, in the terms of the day, the faster CPU did provide). The REAL use of the Turbo button was not to make the computer go "faster", but to make it go slower so as to be compatible with software which crashed at higher clock rates (of which there was far too much).

If you go and buy a CPU for your computer, you can't change its max clock speed with the flip of a switch while it's running.

Yes, I can. I have been able to since my first 286 motherboard, where I had a "Turbo" button on the case to let me switch between 8MHz ans 12MHz, I can on my current quad-core Intel beastie which also has a "Turbo" button capability built into the power switch (quick-press = Turbo Mode if I want it configured that way).

I think I came up with a counter argument. Tell me if this works.

The first thing I can tell you is that if it takes a bright chap like you something like 1.5-2 days to come up with a counter argument, then you're stretching. The second thing I can say is "which bit are you countering?"

If that 286 were advertised as being a 12Mhz computer, but came from the factory with the turbo button off but the manual saying "to get a speed boost, push the turbo button", would you complain to the company about false or at least faulty advertising, or have they still done right by supplying you a 12Mhz processor, even if you have to hit a button to make it active first?

Let's look at it this way - then and now. Then the Turbo button was, as indicated above, not REALLY a "Turbo" button - it was just labelled "Turbo". Everyone knew that you ran at 12MHz but pushed the Turbo button to slow down if you used something like Lotus 123 or one of the other programs that tended to "get stuck" if you ran it at over 8MHz.

If it came with the Turbo button in "slow" mode, you probably wouldn't know because you'd hit the thing 3 times in getting the PC out of the box. Once set to "Turbo" mode though - the acct of a fraction of a second, literally, the PC remained at "maximum speed". Changing the max. clock speed was intuitive, easy, and could be done with the PC tuirned on or off.

Right now, my current quad core PC was sold at a given clock rate. It's the clock rate that Intel spec. the CPU to, but there is a "Turbo Button" feature of the motherboard that allows me to overclock the CPU (take it above the standard clock rate) at the flick of a switch. I can, alternatively, use software provided with the motherboard to set a variety of other overclocking scenarios, including "always overclock to THIS speed". I suspect that the CPU would start up at default though, and only truly overclock once the relevant software had been loaded by the Operating System.

If they had avertised it as running at 3.4 GHz and when I opened the box I found I had to run it at 2.4GHz at certain times, "because", or even found that it ran at 2.4 GHz but could be overclocked to 3.3GHz "or possibly higher" via some arcane behaviour, or that they recommended running it at 2.4GHz not at 3.3GHz then yes, I rather think that I WOULD complain that it wasn't a 3.3GHz PC.

Note that the system vendor did not sell the unit to me as being able to reach a specific overclocking speed - overclocking is OVER clocking. They sold it as being the clock rate that Intel spec. the CPU at. Overclocking is an ability that is provided, and a bonus if you can achieve much with it. I wouldn't blast them for it not achieving a speed they haven't mentioned anywhere that I am aware of.

AFAICR OpenPandora's web site doesn't say anything about "overclocking" or "default/non-default" or "underclocking" in the spec. It just sdvertises the Pandora as using:

ARM® Cortex™-A8 600Mhz+ CPU running Linux

I'm a very simple man, and it's a very simple line. It's either true, or it isn't. If, as you say, the CPUand therefore the Pandora is actually a 500MHz device which can be overclocked, then IMHO the advertising should say it's a 500MHz device. "Guaranteed to overclock to at least 600MHz" sounds good enough to me.

 

[Creature XL]

Note that the system vendor did not sell the unit to me as being able to reach a specific overclocking speed - overclocking is OVER clocking. They sold it as being the clock rate that Intel spec. the CPU at. Overclocking is an ability that is provided, and a bonus if you can achieve much with it. I wouldn't blast them for it not achieving a speed they haven't mentioned anywhere that I am aware of.

AFAICR OpenPandora's web site doesn't say anything about "overclocking" or "default/non-default" or "underclocking" in the spec. It just sdvertises the Pandora as using:

ARM® Cortex™-A8 600Mhz+ CPU running Linux

I'm a very simple man, and it's a very simple line. It's either true, or it isn't. If, as you say, the CPUand therefore the Pandora is actually a 500MHz device which can be overclocked, then IMHO the advertising should say it's a 500MHz device. "Guaranteed to overclock to at least 600MHz" sounds good enough to me.

There is one thing you don't want to understand.
600MHz is NO OVERCLOCKING. It is within the specs. This fact I learned from this thread. I wonder why you didn't learn it. Some month ago i was i loved you fro a brilliant arguing with Shaun i think. And now, you didn't read the other posts correctly

The cortex has different operating modes called "Operating Points"(OPP). OPP1 thru OPP6 (I think, the answer is somewhere in this thread). In OPP3 500MHz is the spec'ed speed and in OPP5 600MHz is the spec'ed speed. You can additionally overclock the chip. You can set it to 600MHz in OPP-mode that is overclocking. You can set it to 720MHz in OPP5-mode (try to guess it.... this is overclocking). However, switching from OPP3-mode to OPP5-mode is NOT OVERCLOCKING.

Page 5 in this: http://www.bsquare.com/Data/Sites/1/marketing_documents/TI_OMAP/App_Note_PowerManagement.pdf

 

[juniorm33]

Your correct apart from your assumption what OPP it's running at by default.
To change the OPP you will need to install a different firmware.
It's running at OPP3 from what we have been told.
Therefore 600Mhz is overclocking, it's just that it's guaranteed by OPT.
So Monk wasn't incorrect. Neither was you apart from your initial assumption.

 

[Monk]

So Monk wasn't incorrect. Neither was you apart from your initial assumption.

Thanks Na-Noo, although to a certain degree I think that getting tied up in the terminology (overclocking, underclocking, overvolting and so on) could be a mistake. I'm just trying to point out that there seems to be some dissonance on the topic of the Pandora's "correct"/"default" clock rate, and that specific arguments used in the discussion don't seem to gel (e.g. "you can't change the clock rate of an Intel CPU while it's running", to paraphrase). Though until they ship I suspect everything we say about Pandora could be deemed "hearsay" as it's subject to delays. I mean, subject to possible change.

I like the film Crazy People, and the concept that it explores of having truth in advertising. An advert should say what a thing is, the thing should be what the advert says it is. I really am THAT simple, at heart. There's lots more that oculd be said, but I said more than I'd hoped to when I first posted to this thread, having seen most of the arguments before

 

[gp32rich]

So I was wondering how TI market the chip to see if OPT are following their lead.

http://focus.ti.com/paramsearch/docs/parametricsearch.tsp?family=dsp&sectionId=2&tabId=2225&familyId=1526&paramCriteria=no

I don't think OPT have anything to answer for, but you could ask TI about their marketing strategies.

 

[sebt3]

What i mean is ,i would like to have a full always say 750mhz ,is this possible at bootup and throughout use ?

AFAIK all CPUs nowadays do scaling.
Intel only offer two speed rates (half and full) whereas ARM can change the speed on-the-go in MHz without you even noticing.

As said, I might be talking bullshit here, I THINK it's the case, but I'm not really sure

Please read : http://www.mjmwired.net/kernel/Documentation/cpu-freq/governors.txt
Most distribution set this at "ondemand" and that the reason why people think intel can only do 2 speeds. But as wikipedia (http://en.wikipedia.org/wiki/SpeedStep) dont report this limitation, i'm not sure it's real... (but I didn't realy checked, so i'm not that confident)

 

[juniorm33]

So Monk wasn't incorrect. Neither was you apart from your initial assumption.

Thanks Na-Noo, although to a certain degree I think that getting tied up in the terminology (overclocking, underclocking, overvolting and so on) could be a mistake. I'm just trying to point out that there seems to be some dissonance on the topic of the Pandora's "correct"/"default" clock rate, and that specific arguments used in the discussion don't seem to gel (e.g. "you can't change the clock rate of an Intel CPU while it's running", to paraphrase). Though until they ship I suspect everything we say about Pandora could be deemed "hearsay" as it's subject to delays. I mean, subject to possible change.

I agree with you 100% and I said pretty much exactly what you said only a few pages ago, that also included how it could be worded on the site. But I'm trying to stay out of the debate now, although it's interesting reading it

So I was wondering how TI market the chip to see if OPT are following their lead.

http://focus.ti.com/paramsearch/docs/parametricsearch.tsp?family=dsp&sectionId=2&tabId=2225&familyId=1526&paramCriteria=no

I don't think OPT have anything to answer for, but you could ask TI about their marketing strategies.

Their marketing the chip not how it's used (implemented), as explained in the post above 600Mhz is correct at OPP5, so there is a difference.
Therefore that is the default speed at OPP5. So you get the chip run it at OPP5 and that's the speed you get, no tweaking no messing, afaik.
OPT has it set to OPP3 and ships it at the default/standard/normal speed which is 500Mhz.

 

[Creature XL]

Your correct apart from your assumption what OPP it's running at by default.
To change the OPP you will need to install a different firmware.
It's running at OPP3 from what we have been told.
Therefore 600Mhz is overclocking, it's just that it's guaranteed by OPT.
So Monk wasn't incorrect. Neither was you apart from your initial assumption.

I am leaving "attack mode" and enter "Gather information mode".

On this site:
http://elinux.org/OMAP_Power_Management#OPPs_control

I found this text (in the cpufreq paragraph):
"The 'stats' directory has info about the cpufreq transitions and the time spent in the various OPPs"

If I need to install a new firmware, which should be followed by a reboot, how can the kernel ( remember the reboot of the kernel?) have a counter for the ticks spent in the OPPs?

AS said, I am curious. I am will try to read more about it in the meantime...


EDIT:
http://e2e.ti.com/support/dsp/omap_applications_processors/f/43/p/35292/123066.aspx

if you look closely the guy responding with "cpufreq ONLY sets the OPP" is a TI-employee

 

[urjaman]

If I need to install a new firmware, which should be followed by a reboot, how can the kernel ( remember the reboot of the kernel?) have a counter for the ticks spent in the OPPs?

My guess: It doesnt. The counters always start from zero on boot (like uptime, network packet counters, etc...).

 

[Emnasut]

I like the film Crazy People, and the concept that it explores of having truth in advertising.

THANK YOU!! I've been looking for this movie for YEARS but had no idea who was in it, when I saw it in my childhood, what the title was etc. and *never* found it!

 

[WizardStan]

The first thing I can tell you is that if it takes a bright chap like you something like 1.5-2 days to come up with a counter argument, then you're stretching. The second thing I can say is "which bit are you countering?"

I wasn't actually looking to say anything originally. I was just minding my own business when my brain suddenly said "HEY! LISTEN!" so I came back, reread, and tried this different way of explaining the situation.

I'm a very simple man, and it's a very simple line. It's either true, or it isn't. If, as you say, the CPUand therefore the Pandora is actually a 500MHz device which can be overclocked, then IMHO the advertising should say it's a 500MHz device. "Guaranteed to overclock to at least 600MHz" sounds good enough to me.

Except it's not overclocking to run it at 600Mhz anymore than the 286 was overclocked at 12Mhz. It's just a "turbo" button set to low speed; only difference is that it's a software switch that takes a few mouse clicks rather than just a hardware switch. That's why I asked if you would have felt your turbo switch being shipped in slow mode was questionable advertising (regardless of how it actually worked: you said you were simple, simple men don't look at a label that says "turbo" and know that it actually slows it down when active )
Unless OPT really is just using OPP3 for their 600Mhz and are only hoping it works. If they really don't have a mechanism to kick it to OPP5 when it goes to 600Mhz if it's needed, then I withdraw all my statements, because then they really can't guarantee any more than 500Mhz.

 

[Monk]

I wasn't actually looking to say anything originally. I was just minding my own business when my brain suddenly said "HEY! LISTEN!" so I came back, reread, and tried this different way of explaining the situation.

I can see that actually, but this is the internet. People aren't supposed to be reasonable...

Post quacks-visit edit - bug cut because it's just not important enough to me.

 

[juniorm33]

I have to admit I'm not sure about the OPP now, I've asked the question (couple times) before but never got an answer. :blink:

If when you set it to 600Mhz it changes to OPP5 then why does OPT guarantee it to run at that speed, that's just the standard TI specification no more no less.

Although for me this still doesn't change my view of how it's described on the site.

Am I right in believing we need an updated kernel to take advantage of the full power management of the chip?

 

[Monk]

I like the film Crazy People, and the concept that it explores of having truth in advertising.

THANK YOU!! I've been looking for this movie for YEARS but had no idea who was in it, when I saw it in my childhood, what the title was etc. and *never* found it!

No worries. It's a brilliant film. Everyone should watch it and then insist on ads made by the insane

 

[WizardStan]

Although for me this still doesn't change my view of how it's described on the site.

But can you understand now why OPT would think they're describing it well enough?

Am I right in believing we need an updated kernel to take advantage of the full power management of the chip?

I can't say anything about the voltage, but we've already seen videos of them messing with the clock speed. If the clock speed can be configured in the kernel, then the voltage control should also be available, but it might not be as easy to change. Unless they intentionally disabled it in their kernel to prevent people from making their Pandoras explode. Only someone from OPT can answer this, and I can guarantee that if they tried something would be misunderstood. Far better to just wait until we've got our Pandoras and you can just check it out yourself rather than risk misunderstanding another poorly worded statement.

 

[juniorm33]

----Snip----
Far better to just wait until we've got our Pandoras and you can just check it out yourself
----Snip----

This I 100% agree with :lol:
As far as I'm concerned we have more than enough idea of what we can hope for, so let's wait and see.
Only reason for them to get into the whole overclocking (in terms of expected speeds) thing again would be because it's not a couple weeks away from being shipped in my eye's.
And god do I hope that isn't the case.

The other question I had though was any idea why the previous revisions seemed to clock better? (I'm sure someone mentioned that)

 

[gp32rich]

And god do I hope that isn't the case.

No! Don't mention the cases!

 

[Caine]

I'm a very simple man, and it's a very simple line. It's either true, or it isn't. If, as you say, the CPUand therefore the Pandora is actually a 500MHz device which can be overclocked, then IMHO the advertising should say it's a 500MHz device. "Guaranteed to overclock to at least 600MHz" sounds good enough to me.

Except it's not overclocking to run it at 600Mhz anymore than the 286 was overclocked at 12Mhz. It's just a "turbo" button set to low speed; only difference is that it's a software switch that takes a few mouse clicks rather than just a hardware switch. That's why I asked if you would have felt your turbo switch being shipped in slow mode was questionable advertising (regardless of how it actually worked: you said you were simple, simple men don't look at a label that says "turbo" and know that it actually slows it down when active )
Unless OPT really is just using OPP3 for their 600Mhz and are only hoping it works. If they really don't have a mechanism to kick it to OPP5 when it goes to 600Mhz if it's needed, then I withdraw all my statements, because then they really can't guarantee any more than 500Mhz.

Agreed. If OPT cannot guarantee 600Mhz at either OPP3 or OPP5 then they're guilty of false advertising.

In my opinion, the Pandora button should map onto a menu which provides a series of system wide options:
- lcd brightness
- keymap profiles (e.g. dpad and nubs behaviour)
- Toggle turbo mode (Switching this on should clock to 600Mhz and if required (and possible) to OPP5. Throwing a warning for increased power consumption or something similar.)
- Window manager (PMenu, XFCE, enlightenment, etc...)

 

[Monk]

I don't want to risk reinterpreting anything (that way madness lies!) so I'm just going to quote:

{responding to speculation that newer Pandora boards might be slower than older ones, and if so why tha tmight be)

Actually, it's the other way around. Rev1's were the slowest technically. It had nothing to do with the PCB, but the newer batch of chips that went on rev2 and beyond.

Rev1 did 625MHz at OPP3 before losing stability (well crashed at 650 for sure). They did 900MHz at OPP5. Rev2 and later do on average 750MHz at OPP3 but we've never tested OPP5 because my old kernel panics at the UART init due to some changes somewhere and we have never had the voltage scaling in the kernels since. I'm not going to speculate on what OPP5 will do on the newer chips but my fingers are *unofficially* crossed.

from:

http://www.gp32x.de/board/index.php?/topic/53304-100-hand-painted-special-edition-pandoras/page__view__findpost__p__850303

and quickly add "this seems to suggest that OpenPandora are using mostly OPP3, including at 600MHz".