Voltage, Temperature and Cooling

[Gophermofur]

When people mention not to go above a specific voltage for a chip is it because:

(1) the temperatures will increase too much, and will subsequently make it difficult to adequately cool your chip, thus shortening the life span.

(2) The high voltage could kill your chip and the increase in temperature is just a by-product of the increased voltage.

(3) Neither?

I ask because if your chip is running quite cool, say 45 degrees under full load (hypothetical) and you are at the max range for your chips vcore... Is it safe to continue increasing the voltage as long as you can maintain a decent temp?

Btw, this is for a Core2Duo.

Thanks,
Mo

 

[aigomorla]

Originally posted by: Gophermofur
When people mention not to go above a specific voltage for a chip is it because:

(1) the temperatures will increase too much, and will subsequently make it difficult to adequately cool your chip, thus shortening the life span.

(2) The high voltage could kill your chip and the increase in temperature is just a by-product of the increased voltage.

(3) Neither?

I ask because if your chip is running quite cool, say 45 degrees under full load (hypothetical) and you are at the max range for your chips vcore... Is it safe to continue increasing the voltage as long as you can maintain a decent temp?

Btw, this is for a Core2Duo.

Thanks,
Mo

C2D or C2Q?

C2D's can go as high as 70's on full load with a max voltage on AIR with TR120 of around 1.45Vcore This is where i would keep it at MOST, during winter. During summer i would decrease the voltage a bit to compensate for the high ambients.

If your running h2o, then 1.5-1.525 would be my ideal MAX. I wouldnt push it that far tho, but just telling you where its the max you should push it at.

Now if its a quadcore thats another list of variables in which you can go.


To answer your first part, both 1 and 2 are correct. The heat will kill the cpu, and also too much voltage will kill the cpu. How much of a decrease will your cpu have by overclocking it with voltage? An intel engineer once said for every 10C you drop your CPU temp, you effectively almost double the life of it.

 

[dmens]

2 is the correct answer.

 

[Navid]

A CMOS transistor contains a very thin insulator. This thin insulator can tolerate only so much voltage across it before it breaks. This can happen at any temperature.

 

[bryanW1995]

obviously temps come into play at some level b/c we all hear about staying at certain voltage levels on air, a little bit higher levels on water, and even higher on more exotic solutions. The voltage levels will do just as much damage to your mobo/cpu/etc on water as they will on air, the higher voltages just usually push the temps up much higher much more quickly on air than water. So, technically, you should not exceed a certain voltage/temp threshold regardless of your cooling method. If you have great airflow in your case and a very strong hsf then you can keep temps in line just as well as some low/mid end water solutions, and 1.55 v will do just as much damage to your system at 50c on water as it will at 50c on air.

 

[Gophermofur]

Thanks for all responses, it clarified quite a bit.

From my last test I was able to get the chip stable at 265, with the vcore somewhere below 1.4 (4-5) notches below on the DS3 mobo. The chips were still reaching some pretty high temps, close to 70.

At 275 it ran for a solid 2 hours, but wasn't stable overnight (about 6-8 hour test). My computer rebooted, and I'm guessing it's a heat issue since the temps were pushing 70 degrees. I think that's my limit for the time being until I get the TR-120 Ex.

For the time being I will try and push the RAM a bit further and see how far it's willing to go.

Is it typical for a reboot to be caused by a lack of voltage or high temps? Or does it really go on a case by case basis?

 

[mrlithium]

random reboots are caused by instability, narrowing down what causes what is really hard. usually I would say temperatures cause it, but voltages can too. If your CPU doesnt get enough voltage it can either hang, or BSOD/restart. theres a phenomenon called vdroop common to all processors that lowers the voltage a smidgen when the CPU sees a load. hardcore OC'ers mod their boards so the chip doesnt decide lower the voltage whenever it wants and enter an unstable zone.but this doesnt sound like its become a problem for you yet.
If the system ran at 275 for a solid 2 hours the only thing that changed after those 2 hours was heat. either your ram heated up too much or your CPU did, or possibly even the PWM. if the PWM heats up too much, like in the 65-70C range, the voltage regulation in it will be thrown off. If i had to guess I'd guess it was your ram heating up, people neglect ram cooling a lot, but ram tends to just get hotter and hotter, slowly by slowly, since the ram can not dissipate as much heat as it is accumulating especially with a lot of bargain ram sticks coming without even heatsinks. plus heat from the CPU contaminates to the ram, and heat from the video card contaminates everything above it, CPU,ram,PSU.... so if you have bad airflow, your system is poisoning itself slowly with heat. do the touch test, if you can't touch your ram for more than 5 sec without burning , its too hot, and you should ziptie an 80mm fan somewhere around that area, that will help tremendously. also, you said you were going to get the Thermaltake 120 Extreme, let me just say right now that its probably one of the best investments you can make. this heatsink kicks so much a** and will take the CPU heat out of the equation.

 

[Gophermofur]

I didn't think it was the ram intially. I half-tested the ram all the way up to 390 where it stopped. Half-tested meansI only let it run about 2 hours or so everytime I ran memtest. I didn't think 275 would be much of an issue for it, especially since it is running at 1.9v (manufacturer recommended voltage).

However, I did run Orthos on blend instead of small fft (woops). So it very well may be a RAM issue. I will try again today with small FFT test and see if I get any further. I'll definately do the touch test a fair bit into the stress test to see how hot it's getting.

I've read that for my particular motherboard, FSB speeds below 333 don't get the MCH too hot. Beyond that point, cooling is recommended. But I will check into how hot that is getting as well.

Thanks for the info. Will try playing around with the computer when I get home.

 

[bryanW1995]

You should be running ram at 1:1 while stability testing the cpu so you can eliminate ram as a potential roadblock. It is almost definitely a heat issue for you. How is the airflow in your case? If you have an ultra 120-x on the way then I would strongly recommend that you hold off on any more oc adventures until it arrives b/c you'll have to start over anyway. Do you have your x1900 set to 2d/3d clocks? That can drop idle temps a ton on one of those.

 

[jeffw2767602]

hey guys this is my first post. ive just built my first system and i have a similar question. i have a q6600 overclocked to 3.33Ghz (9x370) on air with an ultra 120 extreme. 1. the voltage to the cpu in the bios is set at 1.35v but in speedfan and cpu-z it reads at 1.31 (stable). i was wondering which one was correct and why there is a discrepancy. 2. i was hoping to get this cpu up to 3.6GHz on air. i am running very cool at 3.33GHz (around 35C idle and 50-55load) and was wondering why i couldnt get this bad boy up to 3.6GHz stable on manual settings (1.45v in the bios or 1.41v in cpuz and speedfan). Auto settings run fine but i would like to minimize my heat output and i doubt the auto settings did that. so basically, my question is... how high can the voltage of a quad core go safely? when i tested it at 3.6GHz & 1.45v bios or 1.41v speedfan the load temps never exceeded 65C and i havent lapped my CPUs IHS yet (prime95 failed after 23 minutes so im fairly close to a stable voltage).

my system:
q6600 g0 @ 3.33GHz
gigabyte ga-p35-dq6
thermalright ultra-120 extreme lapped
crucial ballistix pc2 6400 ddr2 800 2 gb
seagate 7200.10 500gb
evga geforce 8880gts 640

 

[Gophermofur]

Originally posted by: bryanW1995
You should be running ram at 1:1 while stability testing the cpu so you can eliminate ram as a potential roadblock. It is almost definitely a heat issue for you. How is the airflow in your case? If you have an ultra 120-x on the way then I would strongly recommend that you hold off on any more oc adventures until it arrives b/c you'll have to start over anyway. Do you have your x1900 set to 2d/3d clocks? That can drop idle temps a ton on one of those.

Oh i'm not overclocking the ram. I wanted to rule out the ram, so I ran memtest for 18 hours and there were no errors. I presume the CPU just got too hot. The computer is sitting open on my desk, well under my desk, but airflow should still be way better than when in a case. My X1900 is in a different system, I'm testing this machine with a spare video card (Asus X800).

But I am confused about what you mean when you ask if I have my X1900 set to 2d/3d clocks? I've heard of lowering the clock speed on the 2d graphics component to reduce heat... is that what you mean?

 

[joejccva]

Yea Jeff I'm with ya on this one.

I was o/c'ing last night on my E6750, and I just went to 3.2ghz and manually set my voltage to 1.30, however in CPU-Z and Speedfan it shows it at 1.26 for some reason. Little confused about that. Did something lower my voltage when I booted into Windows?

My temps are about the same as yours, 34c idle and 54c load or so. I manually set my ram to 4-4-4-12 timings, but when I set my voltage to auto for it the bios said (under pc-health) that it failed. So I +.01 on the DDR voltage and it was fine. Booted into windows and it's showing the DDR voltage at 1.98. I guess this is ok, I ran a blend test in Orthos for about 10 hours last night and all tests passed. Checked my temps this morning at it was reading 54c under the Orthos test load.

I wonder if I could manage to get 3.4ghz. Think I would need to up any voltages from where they are? Also to achieve this I'd have to change my settings from 8x400 to 8x425 to get 3.4, but what about the ram settings? It's set at 2 in the bios right now (1:1 at 800mhz). Do I need to change the frequency multiplier for my ram? Or just keep it at 2?

Thanks.

 

[Gophermofur]

Originally posted by: joejccva
Yea Jeff I'm with ya on this one.

I was o/c'ing last night on my E6750, and I just went to 3.2ghz and manually set my voltage to 1.30, however in CPU-Z and Speedfan it shows it at 1.26 for some reason. Little confused about that. Did something lower my voltage when I booted into Windows?

That's called Vdroop. It's common with all boards, but some boards tend to do it more than others. I know my DS3 does it a fair bit. There is actually another thread on this forum about who gets the highest vdroop. From that thread you can see what people are setting their Vcore to in bios and what is being reported in Windows. Vdroop Post

I wonder if I could manage to get 3.4ghz. Think I would need to up any voltages from where they are? Also to achieve this I'd have to change my settings from 8x400 to 8x425 to get 3.4, but what about the ram settings? It's set at 2 in the bios right now (1:1 at 800mhz). Do I need to change the frequency multiplier for my ram? Or just keep it at 2?

IIRC, you can't drop your RAM multiplier below 1:1. So if you want to run at 425, then your RAM will have to run overclocked a little. Most RAM can do 50MHz, but in any case it would be a good idea to:
1. Lower your CPU multiplier to 6x (or as low as it will go)
2. Increase your FSB to 425
3. Set your RAM speed to 4-4-4-12
4. Run Memtest

You may need to add voltage here and there. Find out what your manufacturer recommends/upper limit for your RAM. If it's stable in memtest, then it shouldn't be the limiting factor in your overclock.

Those are just my suggestions, but when the GURU's wake up/log on, I'm sure they can help you out a bit more

 

[joejccva]

Thanks for your help.

But won't the x6 cpu multiplier at 425 bring my cpu speed down to 2.5ghz?

 

[Gophermofur]

If you are testing your RAM, then you don't want your CPU to be the limiting factor. So you take the CPU out of the equation.

Your CPU can easily run at 1.5GHz, so during the testing of your RAM, you don't have to worry about any funny business from your CPU.

Once you've found the ceiling, or in your case if your RAM can run at 425, then you can increase your CPU back to it's maximum speed. At least this way you'll know that if your computer is unstable at 425, the problem isn't the RAM since you've already stress tested it.

 

[joejccva]

Originally posted by: Gophermofur
If you are testing your RAM, then you don't want your CPU to be the limiting factor. So you take the CPU out of the equation.

Your CPU can easily run at 1.5GHz, so during the testing of your RAM, you don't have to worry about any funny business from your CPU.

Once you've found the ceiling, or in your case if your RAM can run at 425, then you can increase your CPU back to it's maximum speed. At least this way you'll know that if your computer is unstable at 425, the problem isn't the RAM since you've already stress tested it.

Ahhh now I understand. I have the Gigabyte P35C-DS3R motherboard, and in the bios when I hit Ctrl-F1 to bring up the Advanced settings so I can manually set the timings, I didn't see anywhere where I can set my memory to 1T. This may be a shot in the dark, but do you know where this setting is? I noticed you had a Gigabyte board as well even though it's a different model. I just thought I'd ask.

Oh and as far as ram voltage or even cpu voltage is concerned, I understand vdroop but which numbers should I always go by? The bios voltage set or what it reads in cpu-z (after vdroop), etc etc?

Thanks again for your help.

 

[Gophermofur]

This is just an educated guess, so someone correct me if I'm wrong.

I'd assume that you would go by the numbers reported in Windows. The reason being that Vdroop is a real problem and not just a cosmetic problem. By real problem i mean the voltage is actually dropping, whereas a cosmetic problem would be if it were appearing to drop, but really stayed the same.

Of course this is assuming that the application in Windows (I assume you are using CPU-Z) is reporting the voltages correctly. In the other thread i posted above, there was some talk about another application that does a much better job of tracking vcore.

I believe some people on this board mentioned that it is also ok to go 10% above the max recommended vcore (as per Intel's specifications). 6750 Specs

Edit: That better 'program' I thought they were referring to, was really a Digital Multi-Meter. So unless you have one of those you'll be limited to apps like CPU-Z.

 

[jeffw2767602]

thanks you very much gopher for the vdroop info.

also in my previous post i mentioned that if i set the voltages to auto i can run stable at 3.6Ghz on a q6600. on auto the voltage for the cpu is 1.39v but when i manually set it to 1.4v+ with everything else on normal it wont run stable. the FSB was doing fine without any voltage adjustment at 400x8. the problem is i dont know enough about the relationship between the voltages of the nb sb fsb and cpu to be able to tell which one needs to be raised should an error occur. are there any programs that will tell me the voltages of my fsb cpu nb sb and ram so i can see what the auto settings on my motherboard are doing so that when in manual mode i can just work backwards from those numbers?


i should probably mention that my bios(on a gigabyte p35-dq6) doesnt show you the actual voltages, it just lets you overvolt it by certain increments.

thanks

 

[joejccva]

Thanks again Gopher. You've been extremely patient and helpful with some of my stupid questions.

Could you answer the question about the 1T/2T settings in the bios for the Gigabyte boards? I couldn't find where to make it so the ram runs at 1T. I can setting the timings to 4-4-4-12 but I didn't see anything related to 1T/2T in bios.

 

[Gophermofur]

Your questions aren't stupid and we all have to start somewhere! Not to mention that I started this thread because I had so very many questions of my own

I'm not really sure about the 1T/2T thing. My board doesn't support it, but apparently the P35's do. Although it would seem like it is difficult getting the ram to run in 1T.

This is all based on the thread linked below
Linky

 

[SickBeast]

I'm pretty sure that the 'vdroop' that everyone seems so concerned about is normal and is actually good.

The CPU will use more voltage when it's under load, and less when it's at idle. Perhaps they quote the idle voltage in the specs. I usually run my CPUs .3v or so over stock, and that sometimes goes even higher under load. I've never fried a chip.

As to the voltage alone frying the chip, at a certain point it will, but really when you increase the voltage, your CPU will pump out way more heat, which will then kill the core. The thing is, modern CPUs have overheating protection built in and will shut down if they get too hot. :light: