Propose your solution
Attachments
No attachments.
Duplicates
Comments
|
JaiBee
wrote on the 26 Apr 10 at 15:37
|
|
|
|
Would be awesome!
|
|
Auzy
wrote on the 27 Apr 10 at 05:29
|
|
|
zomg zomg. You totally forgot to mention the stability advantages.. OH WAIT.. I mean DISADVANTAGES. Yep, underclocking similar to overclocking can lead to a decrease in stability, and many CPU's can dynamically underclock themselves anyway.
To what extent do you undervolt, and wouldn't it lead to an increase in crash reports which are caused by CPU dodginess?
|
|
|
Azuy,
we are talking about underVOLTING not CLOCKING.
Once you figured out how much you can reduce the voltage you leave a "security gap" and the system runs stable.
To what extent? That's the main point of undervolting: you have to check it out. The more the better. Once you get a crash while testing it out you know you reached the limit and raise the voltage.
We're talking about an option to implement it. This should not be linked on the desktop but it should at least be possible without compiling an entire new kernel..
I don't know enough about the technical side but it should be possible to introduce this feature in a better/easier way.
|
KIAaze
(Idea reviewer)
wrote on the 27 Apr 10 at 09:41
|
|
|
Sounds like a good idea to me, but why create a new idea instead of adding it as a solution to the idea you linked too?
It's not a direct duplicate, but almost...
|
|
Auzy
wrote on the 27 Apr 10 at 11:33
|
|
|
@Splashote. Have you actually got any benchmarks which show a drastic effect on battery life?
And would there be a net gain by leaving your computer running for a few hours to determine if it's running stable.
It isn't just a case of a single crash and boom, its considered unstable either. A crash may happen in only 2 months time...
I think more information is needed, and stats which show the overall effect on power usage and speed..
|
|
|
@Auzy:
I don't have any benchmarks. My experience is based on 4 different cases. In two cases, the fans were working all-time before the undervolting, and after the reduction of the voltage were only starting every 30minutes.
It's not about letting it run for a few hours. Usually you run a programme that creates full load and see whether the PC crashes. If it does, raise the voltage. If not, lower it even a step more.
So once you have a stable configuration, there won't be a single crash because all the frequencies (multiplicators) were tested at full load.
There is no effect on speed. As already mentioned, there is a significant difference between changing the voltage and over-/underclocking the CPU.
For further information check out all the info at linux-phc.org
Here a short quote that explains the background of undervolting:
"This is possible because we utilize the production tolerance of a CPU.
CPUs have different production qualities so the vendor defines voltages every CPU -even those with low quality- will work with.
If your CPU is of higher quality you can easily run it with lower voltages without getting unstable. "
I don't know anything about stats or long-term surveys, but check out the forum: http://www.linux-phc.org/forum/viewforum.php?f=22&sid=6a7cf0caa6d19623c3af54d87 415365c
Just to show that Ubuntu & undervolting once was easier:
Quote from the-fallen:
"With the release of Ubuntu 9.04 the modules powernow-k8, speedstep-centrino and acpi-cpufreq are not longer available as module but compiled into the kernel.
This takes away the possibility of the easy way to get in touch with PHC: patching and recompiling the module only."
more on the topic:
http://www.linux-phc.org/forum/viewtopic.php?f=22&t=67
|
|
|
@KIAaze:
In my opinion, undervolting deserves an individual idea dedicated to it ;)
I thought it might be worth a new try as the other idea is a bit old.
|
|
Akerbos
wrote on the 27 Apr 10 at 17:20
|
|
|
|
It is not an act without danger, so it should not be simple. I.e., you cannot set the "best" voltage without provoking a crash. Or is there a benchmark for that, maybe similar to memcheck?
|
|
Auzy
wrote on the 27 Apr 10 at 18:16
|
|
|
That's my point... And it is incorrectly assumed that a crash will definitely occur under full-load. It may only be triggered in certain conditions (which makes debugging more difficult).
Furthermore, as GPU's are integrated into CPU cores, crashes may become less predictable too.
The other thing, is that there are still plenty of power saving options we haven't explored (Windows can put devices to sleep, and has an incredible power over its power options). I'd rather we explore supported methods first, rather than go the dodgy route.
|
|
|
The danger lies in the crashes. No hardware will be harmed.
It's not about making it "easy as changing the wallpaper" but making it easer than compiling a whole new kernel.
Check the work of PHC-Linux. Until 9.04 there was a better way to enable it, now certain modules are already integrated in the kernel. Reverting this step (that had a doubtable effect on boot-up time) might be a solution.
|
|
|
And it's not only about power saving but improving the user experience, too.
Did you ever sit in a library with you fan roaring non-stop? It just bothers when the notebooks/PCs in general are noisy and undervolting is, in my experience and that of many others, an effective way to reduce the noise-generation.
|
|
Auzy
wrote on the 28 Apr 10 at 21:06
|
|
|
The question is, is there actually benchmarks out there which show a noticeable power saving?
I don't doubt that it's possible. But is there any proof that they can be undervolted enough to make any kind of difference? And is there any proof that any kind of undervolting doesn't lead to random crashes later on (it's no good cutting down the fan noise if the computer crashes whilst typing a 2000word essay). For it to make any sense to implement this, we must be sure it wont decrease productivity.
|
|
|
Azuy,
it's good to be sceptical, but there are a lot of tools/programmes that offer undervolting and a lot of people interested in it. I'd rather recognize it as a proof of conecpt. They wouldn't be there if it didn't work.
About crashing: Once you have found the right voltage, your computer won't crash. It's simple like that. And adjusting the voltage requires about an hour or so in my experience. I just enabled it yesterday on another Win XP machine.
I recommend to enable autosave anyway when working with OOo. ;)
UPDATE:
I found this tutorial about an undervolting script:
http://ubuntuforums.org/showthread.php?t=786402
This is not possible anymore due to the kernel changes already named. But it might serve as a concept to implement it.
This script makes the whole work on finding out about a stable voltage by itself.
Nice one!
|
|
Auzy
wrote on the 29 Apr 10 at 07:39
|
|
|
That's all great.. But you keep repeating that it works over and over again, but nobody has actually demonstrated the practical benefits.
Btw, there are plenty of technologies that had little practical benefits but were damned popular. Three examples of duds are RAM defraggers, registry cleaners and the geniuses who thought that manually setting the swap space size led to a performance increase.
Show us a benchmark which shows:
1) Effect on power (even with a laptop battery maybe)
2) Effect on noise
So please, seeing as you have done it, prove the benefits
Just a reminder, nobody should vote for ideas which don't have proof.
How about registry cleaners? Those don't work, yet, they
|
|
mradelet
wrote on the 29 Apr 10 at 18:31
|
|
|
I think its worth noting that all modern processors (within the past 4+ years) do this automatically without user intervention. My current installation of Ubuntu 9.10 is capable of reporting my current CPU load and other metrics to reflect this. My CPU will fluctuate between 1.1 and 1.4 volts by itself.
What happens when processor load increases? Like a car, it will need more fuel... How do you propose this is handled? Does the user have to set limits for when the processor is allowed more voltage?
What about all the BIOS settings that automatically handle processor sleep states, voltage regulation, and cpu/thermal throttling? Again, this has been around for years.
The fact that this still takes you an hour to complete after several times through, and that you ask a user to intentionally crash their system makes me vote NO.
|
|
Auzy
wrote on the 30 Apr 10 at 04:55
|
|
|
|
That's one thing I was trying to say mradelet..
|
|
Auzy
wrote on the 30 Apr 10 at 07:59
|
|
|
|
Oops. I should add that's why I was so insistent about benchmark, because as mentioned, there are plenty of "fake" optimisations.
|
|
|
Ok, guys.
I'm sorry to repeat this once more:
Undervolting does not effect the speed of your CPU. NOT AT ALL.
It's onyl about adjusting the voltage.
If you don't believe me, consider to seek information on the named sites or google it.
The fact that this still takes you an hour to complete after several times through, and that you ask a user to intentionally crash their system makes me vote NO.
It takes me about 5min to install the software on Windows. I said it takes an hour because the software offers a test modus where the new voltages are tested for every multiplicator before enabling it. To avoid any crash this takes a bit of time. But once your through without a crash you'll not face any misbehaviour.
About benchmarks: I can only refer once more to the successfull implementation I did myself. On a Vaio of a friend of mine, I was able to reduce the voltage for every multiplicator to the minimum voltage with a Dual Core 2. You don't have to study phisics to conclude that a reduction of voltage results in lower heat and power consumption.
Ok, I googled a bit more to convince you.
Check out this charts:
http://www.tomshardware.co.uk/undervolt-cpu-phenom,review-31631-7.html
Here's another interesting article: http://dottech.org/tipsntricks/4932
These are the most relevant parts:
"*Advantages of Undervolting
- Cooler CPU (5 to 20c cooler)
- More battery life (15-30mins more)
- Less Fan noise & activity
- Longer overall notebook life
- Eco-Friendly
By reducing the voltage fed to the CPU, it will run much cooler. It will also draw less power from the battery which can give you an extra 10-30+mins battery life depending on the degree of the undervolt. Also a cooler CPU means less fan activity which again points to more battery life and less fan noise. A Cooler CPU could also lead to a cooler GPU if they share the same heatsink/fan. It is common sense that a cooler running notebook will last longer than a warmer/hotter counterpart. Heat is the killer of electronic parts.
*Disadvantages/Risks
- BSOD (Blue Screen of Death)
You will get a BSOD at the very most if the voltages are too low. This is why we do a stability test to make sure we get no BSOD's. If the voltages are setup properly you shouldn’t get any at all. A BSOD seems to be the only risk for undervolting in my experience. So make sure everything is saved before you do anything. I dont wanna be responsible for any loss or hoo-haa's."
I found a test of a mayor german games magazine. They tested undervolting with a Core 2 Duo P8400 and achieved a 20°C CPU-temperature reduction under stress (full load) and gained 20 min runtime in idle mode and 40 min under load.
http://www.pcgameshardware.de/aid,706480/How-To-CPU-Undervolting-fuer-laengere- Akkulaufzeiten-und-niedrigere-Temperaturen/CPU/News/
This is a very nive how-to for windows:
http://www.atomicmpc.com.au/Tools/Print.aspx?CIID=135596
If you have a windows-machine flying around, test it. ;)
Ok, I think I was able to present some proove. Now I'm waiting for tests that show that undervolting DOES NOT work.
;)
|
|
|
@mradelet:
Just one more thing: Even new CPU's can be undervolted successfully. (Please quote a proove that confirms your argument if you're sure I'm wrong and they do all the work by themselves.)
I suppose you're talking about the governors anyway. Please read the whole discussion to understand what undervolting is about.
|
|
Zorak
wrote on the 6 Nov 10 at 13:29
|
|
|
Hello
I'm rather new to linux and tried all that kernel compiling stuff to get a way to undervolt, but somehow didn't succeed.
Under Windows with Notebook Hardware Control it was totally easy. Before I undervolted, the fan of my notebook was rather loud and annoying. After I lowered the voltage about 0.3 volts (i tried lower, but went up for safety margin) it nearly stopped running at all. I accidentally blocked the fan, when i cleaned it shortly after. It took me about a year to find out, because there where no thermal or stability issues without fan. The last two years I had no crashes at all (even with windows ;-) ) and would like to have a quiet notebook again. I have no numbers on power consumption exept about 25% reduction in core voltage, but the noise under linux makes me consider going back to quiet windows again.
|
Post your comment
|
