Underclocking = Unsafe, depending on voltage

[Evilonigiri]

I decided to bring in a new thread to bring my side of the argument against iluvgillgill. 😉

I googled about underclocking, and there's not much to it. Go google overclocking and they list all the benefits and dangers.

So here's my thinking process:

The NB and FSB has a "normal" voltage of 1.3V, or at least assume this for now. As you know, 1.3V is plenty for 333MHz, perhaps even 400MHz, so we can say it's a little too much voltage at 266MHz, but it's fine nonetheless and hardly dangerous. However, if we cold lower the voltages to, say 1.2V or so, it's probably better.

Now we all know overvolting is undesirable, perhaps damaging. Take a Q6600 clocked at stock speeds for example. Setting the vcore to 1.5V when really only requiring 1.2V at 2.4GHz is damaging to the cpu, even though it can handle it. Thus, is it not damaging if you set it to 800MHz at 1.2V, when it really requires .8V?

As for damaging, whether it be truly damaging or only slightly damaging, it's "dangerous" according to the manufacturer, since it's out of the specs of the motherboard.

I have gotten quite interested in this, so does anybody have any links that proves anything? Preferably from someone who understands this field of study, and not some hardware enthusiast.

 

[spuddyt]

you can test... 😛

 

[Evilonigiri]

How?

 

[yadge]

I thnk that logic is off.

It would be definitely better if you lowered the voltage, but I don't think it would be damaging to leave it at stock if you underclock.

In my opinion, voltage and speeda re unrelated. Well, not really. But what I mean is that if something is able to run at a certain voltage, it shouldn't matter what speed it is going. it's just that it's able to handle that voltage.

The damage of the voltage is not related to the speed you are running. With your logic, making a cpu run faster will let you raise the voltage higher without problems, as long as it is proportional to the speed of the processor.

I don't know, it's kind of hard to explain. But I guess I mainly think that while lowering voltage is good for a component, it is still rated for a certain voltage, and will run fine within whatever voltage the manufacturer says it can, regardless of how slow the component is.

 

[Evilonigiri]

That I understand. If the manufacturer rates it at a certain range of voltages, it's fine, but only for that certain range of frequency, am I right?

So going lower than the recommended frequency is damaging, at least according to the manufacturer. Not that it is truly damaging.

 

[yadge]

No, because what I am saying is that the voltage doesn't have to do with the speed. It only does if you try to make it go faster. It is definitely helpful to undervolt if you lower frequency, but I don't think it is damaging to leave it normal.

But really, I don't have any evidence or proof. So I suppose you could be right. I'm just basing this on what I think.

 

[Evilonigiri]

Nor do I have any evidence or proof. I could just very well say there's no harm done, after all, not many people underclocks (not counting speedstep).

Besides, in a OCing community, I'll bet everyone will say it's safe to UC.

 

[4745454b]

That I understand. If the manufacturer rates it at a certain range of voltages, it's fine, but only for that certain range of frequency, am I right?

Not that I'm aware of. The chips are rated for voltage X, they are then tested to see what frequency they can get. Lets say AMTel has 100 chips roll off the line. They are all made at the same time, and have the same insides. When AMTel designed the chips, they know that 1V should be the Vcore, so they are set for 1V. (AFAIK, the voltage would be known when the chip is designed.) Lets say that 10 of the chips are able to run at 4GHz with a Vcore of 1V, these will be sold as their Uber chip. Another 15 can run at 3.6GHz. This is the practice of binning.

Those that were running at 3.6GHz might be able to run at 4GHz, but they would require a voltage bump to do so. (1.2V perhaps...) This is where my knowledge isn't as strong, but I'll attempt to explain. The problem with increasing voltage, other then increasing heat, is that it increases internal damage to the CPU. To much voltage can damage the transistors.

I believe the only problem with to little voltage is the transistors won't be able to work. They won't have the energy needed operate. Just like in my example how 1V wasn't enough for some chips to run at 4GHz. If you underclock, there isn't any harm (that I'm aware of) if you decrease the voltage.

 

[Shadow703793]

According to Intel specs undervolting and underclocking becomes a problem only when you go go below the recommended voltage. For the Q6600 it is .85v.

See Intel Specs:
http://processorfinder.intel.com/Details.aspx?sSpec=SL9UM

 

[jaguarskx]

Underclocking is not dangerous, I have done it before with my old and new HTPC. If you set the voltage too low you will simply get a BSOD or the system will restart. If that happens simply increase the voltage to the CPU, GPU, or whatever you have underclocked.

For example, with my old HTPC which has an Athlon XP-M, I set the voltage to 1.35v instead of the default 1.45v and the system worked fine.

In my new HTPC, I underclock my X1900XT 512mb down to 196MHz Core and 297MHz Memory using ATITool 0.27. It allows profiles to be set up and also when it detects a 3D application (game) is running it will revert back to default speed. Once the application is ended the GPU drops back down to the low speed I set.

There is also an option to set voltages, but ATITool 0.27 doesn't allow you to save it to a profile, so it must be manually adjusted. I set the VCore to 1.2v (down from 1.45v) and all other voltages to minimum. This drops the GPU's temp from 57C when idle down to 50C with the fan speed set to 22%.

Experiment with voltages by decreasing them incrementally, then test by play a game (for the video card), or running some intensive program like CPUBurn (for the CPU); run one instance for each core and it will put a 100% load on the CPU. If everything is okay, then lower the voltage and test again.

-----------------------------------

EDIT: "underclocked" replaces "overclocked" in the first paragraph.

-----------------------------------

 

[Evilonigiri]

If you underclock, there isn't any harm (that I'm aware of) if you decrease the voltage.

And if you don't decrease the voltage? My argument was the underclock will do harm if the voltage isn't decreased as well. Although it may be able to handle it, I think it'll do more damage than stock speeds.

 

[Evilonigiri]

Underclocking is not dangerous, I have done it before with my old and new HTPC. If you set the voltage too low you will simply get a BSOD or the system will restart. If that happens simply increase the voltage to the CPU, GPU, or whatever you have overclocked.

I agree underclocking isn't dangerous, to a certain extent. And by dangerous, I mean more harm than stock speeds. In the case of the motherboard, there is the NB and FSB voltage where you can't go any lower than, say, 1.3v. So, using an extreme example, if I were to use the motherboard at 1MHz FSB, 1.3V will do damage, at least more so than running stock.

 

[lucuis]

I agree underclocking isn't dangerous, to a certain extent. And by dangerous, I mean more harm than stock speeds. In the case of the motherboard, there is the NB and FSB voltage where you can't go any lower than, say, 1.3v. So, using an extreme example, if I were to use the motherboard at 1MHz FSB, 1.3V will do damage, at least more so than running stock.

Clocking a processor lower will not magically make it susceptible to damage from it's rated stock voltage.

There's only two main things that damage a processor. Electron migration due to long term overvoltage. And heat. Neither of those will EVER occur due to underclocking at stock voltage. To clarify overvoltage is when the voltage is set higher then it's rated. 1.6ghz at 1.2v stock is not overvoltage if the processor can run at 2.4ghz at the same voltage.

As far as i know if you increase the clock speed, you increase the power draw. At stock speeds a chip will draw a given amperage. If you overclock it mildly on stock voltage it will draw more amperage. Inversely if you underclock it it will draw less.

In other words, at stock voltage and at stock or lower clockspeeds you will NEVER over/underpower the chip. Thus making it impossible to be damaged by underclocking.

This goes the same for the motherboard.

 

[Evilonigiri]

There's only two main things that damage a processor. Electron migration due to long term overvoltage. And heat. Neither of those will EVER occur due to underclocking at stock voltage. To clarify overvoltage is when the voltage is set higher then it's rated. 1.6ghz at 1.2v stock is not overvoltage if the processor can run at 2.4ghz at the same voltage.

Electro migration occurs in any electrical component. The higher the voltage, the faster the process.

So from what you are saying, a 2.4GHz at 1.3V is okay for 800MHz? What about 500Mhz? 1Mhz? According to my experience, overvolting isn't a good thing, even if the component can handle it.

 

[lucuis]

Very true, i should have been more specific. What i meant is that only when overvolting electron migration can become a problem. Instead of like 30yrs later

So from what you are saying, a 2.4GHz at 1.3V is okay for 800MHz? What about 500Mhz? 1Mhz? According to my experience, overvolting isn't a good thing, even if the component can handle it.

Absolutely. 1.3v at 2.4ghz vs 500mhz. Temps should decrease due to less current required. And electron migration is unaffected. The only bad thing is it would run horribly slow

 

[4745454b]

And if you don't decrease the voltage? My argument was the underclock will do harm if the voltage isn't decreased as well. Although it may be able to handle it, I think it'll do more damage than stock speeds.

Is there a reason why you think this way? Perhaps if we understood your thinking we could understand you better. (I don't know anything about the thread you referenced in your first post, nor do I want to get involved with any bickering that might be going on.)

Go back to the example that I gave. Remember that all chips are being made at the same time. Some of the chips can run at 4GHz, while others can only run at 3.6GHz. Some might even be able to only do 3GHz, or 2.8GHz. None of these "lower" CPUs should have the voltage lowered however, as they are only stable at these lower speeds with the normal 1V.

Clock speed/FSB and voltage are not locked or linked in any "real" way. You can decrease one without touching the other. If the chip calls for 1.3V, and you decrease the FSB down to 100MHz, I don't see why you'd HAVE to decrease the voltage. Most people would simply seeing as there is no point in having it that high if the clock speed is that low.

 

[Evilonigiri]

And if you don't decrease the voltage? My argument was the underclock will do harm if the voltage isn't decreased as well. Although it may be able to handle it, I think it'll do more damage than stock speeds.

Is there a reason why you think this way? Perhaps if we understood your thinking we could understand you better. (I don't know anything about the thread you referenced in your first post, nor do I want to get involved with any bickering that might be going on.)

Go back to the example that I gave. Remember that all chips are being made at the same time. Some of the chips can run at 4GHz, while others can only run at 3.6GHz. Some might even be able to only do 3GHz, or 2.8GHz. None of these "lower" CPUs should have the voltage lowered however, as they are only stable at these lower speeds with the normal 1V.

Clock speed/FSB and voltage are not locked or linked in any "real" way. You can decrease one without touching the other. If the chip calls for 1.3V, and you decrease the FSB down to 100MHz, I don't see why you'd HAVE to decrease the voltage. Most people would simply seeing as there is no point in having it that high if the clock speed is that low.

My thinking is overvolting something will do more damage than not overvolting. Surely overvolting will do more harm then good?

While you explain through using different cpus with different requirements, I'm looking at only one cpu, all relative to itself. If it can do 4Ghz on 1V, doing something extreme such as 1MHz on 1V is damaging, although perhaps not damaging enough to do any permanent damage.

 

[4745454b]

My thinking is overvolting something will do more damage than not overvolting. Surely overvolting will do more harm then good?

Correct, over volting is bad. I don't change the voltage for my CPU, and would never raise it above what its set at.

While you explain through using different cpus with different requirements, I'm looking at only one cpu, all relative to itself. If it can do 4Ghz on 1V, doing something extreme such as 1MHz on 1V is damaging,

Lets ignore the fact that you can't get a modern CPU down to 1MHz. (even the original 8086 ran at 4.something MHz.) Why is 1MHz @ 1V damaging? 1V was fine for running it at 4GHz, why is it bad now? As I said, perhaps if you can explain why you think this is bad I'd have a better idea of where your coming from.

 

[lucuis]

My thinking is overvolting something will do more damage than not overvolting. Surely overvolting will do more harm then good?

While you explain through using different cpus with different requirements, I'm looking at only one cpu, all relative to itself. If it can do 4Ghz on 1V, doing something extreme such as 1MHz on 1V is damaging, although perhaps not damaging enough to do any permanent damage.

But it's not damaging to underclock, even at that extreme. I'm not sure how else to explain it. Perhaps reading over some electrical basics might help.

 

[Shadow703793]

Any one know what Intel/AMDs standing on underclocking? After all they do void your warranty if OCed :lol: (Only if you tell them though will they know)

 

[yadge]

Just because you underclock, and the voltage is higher than needed, that doesn't mean it is overvolted. It just means that it is more than it needs to be. Overvolting is more voltage than the stock voltage.

 

[jaywalker256]

Well, I have 2 arguments about why underclocking will cause no harm.

1. Take a look at old cpus. A pentium MMX @ 200mhz for example, required 2.8v. Thats an awful lot of volts for such a low frequency. But was it damaging? Not at all. This argument isnt really relevant or important, but its just a thought.

2. Voltage is only pushed through a chip during each clock cycle. Ex. A cpu running at 200mhz at 2.8v, will recieve a 2.8v signal, 200 million times a second. Now lets clock it down to 1mhz. Now, it recieves a 2.8v signal, 1 million times a second. The same amount of voltage is being applied less frequently. Thus, it is impossible to do damage to the cpu in this way. Sure, it may not be necessary to have it at 2.8v, but given this example, it is doing no more than 200x less damage to the cpu per second, than running at stock speed.

 

[lcaley]

The voltage isn't directly correlated with the clock speed. If you have 1MHz at 1v (we'll call that stock voltage), it is still running at the 1v that AMTel (lmao, I love that..) specified their chip to run at. Just because it is moving slower doesn't mean that it can handle any less voltage. It just doesn't need it.

I heard this analogy somewhere else, so I by no means claim it, but think of voltage as food to feed a construction worker. Give give the worker too little food, he is unable to do his job, what with the starvation and all. Too much food, he still does his job just fine, he just doesn't require that much food to get the job done. His belly fat is higher temps.

 

[yadge]

I heard this analogy somewhere else, so I by no means claim it, but think of voltage as food to feed a construction worker. Give give the worker too little food, he is unable to do his job, what with the starvation and all. Too much food, he still does his job just fine, he just doesn't require that much food to get the job done. His belly fat is higher temps.

That's a really good analogy. I was trying to think of one earlier, but I couldn't.

 

[yadge]

EDIT: Mistaken double post.