High Voltage as long as you keep the temperatures low?

c627627 · Apr 17, 2004

TC wrote on 04-16-04 11:45 PM:
Yes voltage - electrical components are rated for a specified range of voltage, amperage, and temperature. Exceed any one of those for long enough and you will destroy the component(s). I don't know why someone would say that as long as you keep a cpu cool nothing else matters.

The below listed conclusions were made for as much as a 50% increase in Vcore voltage.

It was said that CPU weak spots not discovered during manufacturing stress tests may lead to permanent failure sooner than average, not the fact that you’re running too much voltage through your adequately cooled system. If this is not true, conclusion number 4. below needs to be changed.

1. Increase of 60 MHz at the expense of increasing 0.2 volts is not recommended.

2. Assuming expected CPU life time is 10 years, 30% overvolt statistically reduces it to six years.

3. 30% overvolt for DLT3C's is 1.95 volts and for Bartons and DUT3C T-Breds, about 2.1 volts. (Remember to keep the on die CPU temperature under 65 C.)

4. Highest attainable stable Overclock cannot be reduced within months provided on die CPU temperature is kept under 65 C, however:

5. CPU weak spots not discovered during manufacturing stress tests may lead to permanent failure sooner than average:

hitechjb1 said:
A 10% increase in Vcore, would shorten the failure time to 83% of nominal failure time.
A 20% increase in Vcore, would shorten the failure time to 69% of nominal failure time.
A 30% increase in Vcore, would shorten the failure time to 59% of nominal failure time.
A 50% increase in Vcore, would shorten the failure time to 44% of nominal failure time.

Nominal failure time is assumed to be 10 years at default voltage.

hitechjb1 said:
These are the links to the posts discussing the effect of voltage, temperature, .... on CPU due to electromigration, ....

Effect of high Vcore and electromigration on CPU failure time (page 15)

Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton (page 15)

What could damage a chip/CPU permanently? (page 15)

They have to be looked at in detailed technical context.

dead_man311 · Apr 17, 2004

Very good info there c627627 , you always have some of the best post to read. with great info that you need to know when messing with parts and overclocking

TruckyJ · Apr 17, 2004

Bump for a great breakdown. Good info.

hitechjb1 · Apr 17, 2004

These are the links to the posts discussing the effect of voltage, temperature, .... on CPU due to electromigration, ....

Effect of high Vcore and electromigration on CPU failure time (page 15)

Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton (page 15)

What could damage a chip/CPU permanently? (page 15)

They have to be looked at in detailed technical context. If you use your own interpretation and take some quote from it, the conclusion may be quite different than what it was originally meant, .... (regardless whether what I said is correct or incomplete).

I understand you intend to make a complicated technical subject, relationship between many concepts as simple as possible into few sentences to help others, but things may not be as oversimplified as you have been quoting my original posts in many threads, ...., that may lead to questions and misunderstandings if someone would like to know more details (since links to original posts are missing), ....

We have to be careful trying to simplify a complex problem into a few rules. Someone would look at a few numbers of voltage and temperature, ...., there is NO simple few numbers that can satisfy for all situations, ...., it is statistical and involves more complex engineering concepts if one wants to get a more complete picture of overclocking and the underlying theory, ....

If you read the above posts (in the links), I don't think what you quote in yellow is a complete and correct representation of them.

Also the simplified 10%, 20%, 30%, 50% quote you have been quoting in many thread without linking to the original post may lead to misunderstand, since all the original assumptions, the context, .... are missing.

If you read the posts in details and also the Black equation on electromigration, both voltage and temperature are two key important factors that shorten the failure time significantly, one would NOT come to statement 4 that you suggested.

I suggest that if you want to quote, please quote them in more detailed context and give a link to the original posts.

hitechjb1 · Apr 17, 2004

hitechjb1 said:
Effect of high Vcore and electromigration on CPU failure time

It is known that high temperature and high current (density) have adverse effect on chip behavior due to electromigration, could lead to complete chip failure (not just performance degradation per se). Electromigration may increase the resistance of metal wires and contacts inside a chip (max overclocking degradation before functional failure), and may even lead to open connections and resulted in complete chip failure.

But these are long term effect and would not happen in days or even months (see arguments below). Also the failure is statistical in nature, measured over a large sample of chips, a particular one or few chips may behave very differently within certain statistical deviation. It would not be accurate to simple put a number of max voltage on a particular chip and try to predict its life. Same like trying to put a number of max overclocking frequency on a CPU, ...

Assuming statistical failure is due to electromigration, the statistical failure rate usually measured in terms of the failure time of 50% of the population in a large sample of same chips is given by

T_failure = A exp (E / k T) / (J^2)

where A, k are constants, and E is activation energy of the material, J is current density, and T is temperature (in K). The equation is call Black's equtation and is based on empirical results. There may be some deviations from it (mainly on the exponent of J) for different material in wires and contacts, e.g. Al, Al/Cu alloy, Cu.

So put in simple terms, a rough estimtation, we can say that if

- Vcore is increased by 10%, on the average, the current density in the wires inside a chip would be increased by 10% (assume uniformly distributed current density), keeping temperature the same, so the failure time would be shorten by 17%. So

A 10% increase in Vcore, would shorten the failure time to 83% of nominal failure time.
A 20% increase in Vcore, would shorten the failure time to 69% of nominal failure time.
A 30% increase in Vcore, would shorten the failure time to 59% of nominal failure time.
A 50% increase in Vcore, would shorten the failure time to 44% of nominal failure time.

- An increase of temperature by 20 C over nominal max temperature would roughly result in doubling the electromigration rate, hence shortening the failure time to 50%. But that is measured by the temperature over the max temperature specification (above 85 C), which is not likely to happen in daily overclocking.

E.g. for a TBred B DLT3C, from the estimate, increase Vcore from 1.5 V to 1.95 V (a 30% increase) would shorten the statistical chip failure time defined above to 59%. This number is still in the ranges of 5+ years, assuming the nominal life expectancy of the chip is 10+ years.

Based on the above, if the numbers stand, one can pick and chose the max voltage and max overclocking frequency to trade with the life expectancy of a CPU.

The above only discussed the long term failure rate or life expectancy issues related to electromigration, but the question about short term degradation related to high voltage (if there is any) remains open.

hitechjb1 · Apr 17, 2004

hitechjb1 said:
Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton

In the last post, I estimated and related the CPU over voltage with expected 50% sample failure time (life expectancy) based on electromigration emprical results (Black's equation). Here put them into numbers for CPU voltages.

Statistically, for the same level of temperature specification,
A 10% increase in Vcore, would shorten the failure time to 83% of nominal failure time.
A 20% increase in Vcore, would shorten the failure time to 69% of nominal failure time.
A 30% increase in Vcore, would shorten the failure time to 59% of nominal failure time.
A 50% increase in Vcore, would shorten the failure time to 44% of nominal failure time.

So a 30% increase of Vcore reduces the 50% sample failure time to 59%. 30% over stock voltage for Tbred B/Barton are
- 1.95 V for DLT3C, such as the famous Tbred B 1700+/1800+
- 2.08 V for DUT3C, such as the popular 2100+
- 2.15 V for DKT3C, such as the Barton 2500+ or higher.

E.g. If the nominal CPU life expectancy is 10 years, for Tbred B DLT3C
- 30% overvolt to 1.95 V, the number would be down to about 6 years (59%).
- 20% overvolt to 1.80 V, the number would be down to about 7 years (69%).
- 10% overvolt to 1.65 V, the number would be down to about 8.3 years (83%).

They seem to fit nicely w/ the AMD absolute rating 2.05, 2.15 and 2.20 V on Vcore for the DLT3C, DUT3C and DKT3C respectively. Max Vcore for Tbred B and Barton (page 5)

Based on the analysis, we can rule out the guessing numbers of 1.8 - 2.0 - 2.2 V for max Vcore flowing around and also the concern of failure within weeks or months.

As far as temperature to not having additional adverse effect on chip behavior from electromigration on top of voltage, it should be below the max temperature rating of 85/90 C (for TBred B/Barton). So using a temperature cap of 65-70 C is reasonable, since above which most CPU would be overclocked above the break-even point of 10 MHz/C for Tbred B and Barton. Further increase in voltage and temperature beyond 30% and 65 C, even if it is stable, one would get very little return in MHz, but greatly shortening the expected failure time. (Besides temperature is kept under 65-70 C, HSF, motherboard FSB, memory, PSU, ... are assumed not to be limiting the stablity of the system.)

In conjuction with the MHz gain from 10%, 20%, 30% over voltage, one can pick and chose the tradeoff between MHz gain and the reduction of statistical expectancy of CPU failure time.

If one plans to use the CPU for 20 years, or if one is not comfortable of using a 30% higher voltage at which CPU is working above the break-even point (10MHz/C) of frequency and temperature (on air), one would not lose too much MHz even the Vcore is lowered by 10% (~150 mV) at that level, estimated by about 100 MHz. For practical reason, apart from short term benchmarking and fun, trading 100 MHz for 150 mV lower in Vcore is justified.

hitechjb1 · Apr 17, 2004

hitechjb1 said:
CPU voltage: from stock to max absolute, from efficient overclocking to diminishing return

1. For Tbred B/Barton, the default voltage ratings (stock voltage) are
- For mobile Barton, 1.45 V
- DLT3C 1.5 V
- DUT3C 1.6 V
- DKT3C 1.65 V
This is the default voltage rating AMD recommends to use.

2. The max absolute voltages that AMD put up are:
Quoted from AMD:
"The AMD Athlon XP processor model 8 should not be subjected to conditions exceeding the absolute ratings, as such conditions can adversely affect long-term reliability or result in functional damage."

- For DLT3C, e.g. 1700+ DLT3C
Vcc_core_dc_max = 1.5 + 0.05 = 1.55 V
The absolute rating for Vcore = 1.55 + 0.5 = 2.05 V

- For DUT3C, e.g. 1700+ DUT3C, 2100+
Vcc_core_dc_max = 1.6 + 0.05 = 1.65 V
The absolute rating for Vcore = 1.65 + 0.5 = 2.15 V

- For DKT3C and Barton, e.g. 2500+, 3200+
Vcc_core_dc_max = 1.65 + 0.05 = 1.70 V
The absolute rating for Vcore = 1.70 + 0.5 = 2.20 V
Ref:
Max Vcore for Tbred B and Barton (page 5)
How much voltage can be applied to a CPU (page 5)

3. For overclocking, the "efficient overclocking voltage" that gives the most overclocking frequency and keeps temperature below diminishing return is
- between 1.5 to 1.85 V for DLT3C and mobile Barton,
- between 1.6 to 1.95 V for DKT3C and desktop Barton
getting about 100 - 130 MHz per 100 mV.
Ref:
General rules on voltage and temperature for CPU overclocking (page 16)

4. If one needs to get the last MHz (last stable 100 MHz) from the CPU, then the CPU has to operate above the "efficient overclocking voltage" and below the "max absolute voltage" . The CPU would have to operate in the diminishing return regime in which every mV of voltage added to speed up the CPU frequency would be counter-acted by the heat increase which in turn slow down the CPU. The return of MHz from voltage is small (< 30 MHz per 100 mV, < 10 MHz / C) and is costly in term of cooling, power supply in this operating range.

This voltage range is recommended for benchmark testing and competition, and not necessary for 24/7 usage. If one has only a CPU to rely on, don't operate it constantly in this voltage range.
Ref:
Some numbers to determine max CPU overclocking frequency - Vcore vs temperature,
When do the CPU's slow down? (page 13)
Explanation (page 13)

5. The effect of high voltage on CPU life expectancy is discussed in:

How to determine "highest" voltage and temperature for CPU overclocking (page 16)
Effect of high Vcore and electromigration on CPU failure time (page 15)
Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton (page 15)
What is gate break-down voltage (page 16)

Related links:

Relationship of clock, die temperature and voltage (update)
- What is the active power of a CPU at frequency f and voltage V
- How to estimate CPU static and active power
- Effect of die temperature on CPU clock frequency at a given Vcore
(page 13)

Vcore vs processor frequency and cycle time (page 19)

What is CPU stability (page 19)

Why frequency and voltage are important for overclocking performance (page 19)

c627627 · Apr 17, 2004

As you know, other posts have "Big thanks to hitechjb1 whose collection of posts should probably be made into a sticky" on top.

From now on they'll have the following quote added:

hitechjb1 said:
These are the links to the posts discussing the effect of voltage, temperature, .... on CPU due to electromigration, ....

Effect of high Vcore and electromigration on CPU failure time (page 15)

Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton (page 15)

What could damage a chip/CPU permanently? (page 15)

They have to be looked at in detailed technical context.

Of course, if these three links are insufficient to put simplified conclusions into a broader context, all other ones will be added.

TV guide had to adapt and shorten its articles recently because they were deemed too complicated. We can stick our heads in the sand, but that’s what we’re dealing with. So yes, it is a sad sad fact of life that people do not get half way through the full and complete answers to voltage questions being repeatedly asked at the forums. They're looking for simplified answers instead. It's that or nothing for them.

You know how difficult it was for me to get simplified statements that were as close to general facts as possible and how reluctant you were to break things down into simplified conclusions because they inevitably detract from the big and complete picture that I unsuccessfully and repeatedly lobbied to be placed as a sticky in the AMD section. I continue to urge people to get mods to sticky a collection of your posts.

This is about “Highest attainable stable Overclock cannot be reduced within months provided on die CPU temperature is kept under 65 C.”

If you have a DUT3C T-Bred or a desktop Barton, 2.1 volts will not kill your CPU provided on die CPU temperature is kept under 65 C. That was the conclusion reached after extensive debates in many threads.

Why did 2.1 volts kill some CPUs if their temperatures were under control? “Because CPU weak spots not discovered during manufacturing stress tests may lead to permanent failure sooner than average.” That was another conclusion.

Are these conclusions fundamentally wrong?

hitechjb1 · Apr 17, 2004

I think you are trying to push for getting some rigid numbers and rigid rules, but these things are statistical in nature.

I've been trying to explain in terms of underlying technical concept and trend, rather than giving exact numbers (which I don't think I am able to do).

Failure time, failure rate are statistical in nature, if you quote numbers so black and white in terms of many much volts, temperature number in C, what % in CPU life expectancy, ..., someone may pick up these numbers and misuse them, ...

One chip can withhold certain voltage, but another chip may fail at a lower voltage even with same stepping and from the same batch. We do not know what exactly the circuit inside a chip is, there are many factors that can affect the voltage, current and temperature, the gate oxide thickness, transistor threshold voltage, transistor leakage characteristics, ... can affect the maximum voltage one can put into a CPU. That is why P4 and Tbred B/Barton, both from 130 nm, their max voltages are very different.

As for overclocking, it is already working way above most manufacturers' specification on voltage and temperature, and component life expectancy are greatly reduced due to high voltage and high temperature. Black's equation is an attempt to answer these questions in a general setting and trends, not exactly in voltage, deg C, or how many years that you may be looking for, ..., unless we have measurements on the sample data.

We are individual overclockers and most of us are not using large enough samples and working in a well controlled environment like labs in AMD, Intel, IBM, ..... We cannot answer those questions in exact terms and numbers, due to lack of engineering samples, data, times, equipments, ....

hitechjb1 · Apr 17, 2004

For example, within a technology (such as 130 nm), transistors can be of different gate oxide thickness, say 50 A, 35 A (angstom), 25 A, .... 1 angstom (A) = 10^-10 m.

In 90 nm, transistors with even thinner oxide are used.

Usually, a mixture of different transistor oxide thickness can be implemented in a chip. Since we do not know what types of transistors are used, if there are transistors with thin enough oxide and a high enough voltage is applied to those transistor, the chip can be damaged permanently due to gate oxide breakdown.

hitechjb1 said:
What is gate break-down voltage

MOS FET’s which makes up most of the CPU and CMOS chips have very high gate impedance, since the gate is insulated from the underlying source drain channel of a FET, and the gate itself would NOT conduct current (except for the order of magnitude smaller tunneling current due to quantum effect) even when the gate voltage is above its threshold voltage (about 0.2 – 0.3 V for 0.18, 0.13 and even 0.09 micron regular FET’s). So applying 1-2 V to the gate per se (at least up to 0.13 micron chips) should not be a concern.

MOS FET’s have a so called gate break down voltage. The gate of a FET transistor is insulated from the underlying source drain channel by a thin layer of silicon dioxide (SiO2) forming a MOS capacitor. The thickness is very thin (of the order of 20-30 A for 90 and 130 nm, and is getting thinner for next generations, approaching 10 atoms or so thickness). So when a high enough voltage is applied to the FET gate, the intense electric field (electric_field = voltage/oxide_thickness) may damage the gate dielectric (dielectric breakdown). I estimated that voltage is somewhere between 2 and 3 V, depending on the oxide thickness. So if a DMM or analog multimeter has high open circuit voltage, they indeed can potentially damage FET.

For commercial chips, the internal FET’s are well insulated from the external pins from package. The package pin should not be directly connected to the internal, smaller transistors, without going through some stage of I/O buffers which also can stand a higher voltage (e.g. thicker oxide) than the internal small and fast transistors. There are also protective diode at each package pin to minimize damage due to electrostatic discharge, ….

c627627 · Apr 17, 2004

You made the point abundantly clear.

The use of the justifying term "statistically reduces..." is not enough to avoid the "It depends..." answer of "T_failure = A exp (E / k T) / (J^2)"

I thought I couldn't reply with talk of the gate oxide thickness, transistor threshold voltage, transistor leakage characteristics because there'd be nothing but blank stares if I do.

On the other hand there's no excuse for the over-generalization of what was said in many of the previous posts, so there'll be no more of those orange posts.

hitechjb1 said:
I think you are trying to push for getting some rigid numbers and rigid rules, but these things are statistical in nature.

I've been trying to explain in terms of underlying technical concept and trend, rather than giving exact numbers (which I don't think I am able to do).

makaka · Apr 17, 2004

just question so it is safe to run with 1.95 v barton 2500+ at 2500mhz with 60cfull load(prime95) i have read all thread and found that i cpu temp should not go over 65c but people in the forum recommand 50c what is the estimation of the life time of this chip ? maybe 1, 2 , 5 years?
what do u think?

Steven4563 · Apr 17, 2004

makaka said:
just question so it is safe to run with 1.95 v barton 2500+ at 2500mhz with 60cfull load(prime95) i have read all thread and found that i cpu temp should not go over 65c but people in the forum recommand 50c what is the estimation of the life time of this chip ? maybe 1, 2 , 5 years?
what do u think?

i would guess around 7 years maybe it will last longer u wont know because as like overclocking some overclock better than others with the SAME stepping.

for example not overclocked cpu's if u have 2 cpu ones got a weak spot and lasts 8 years the other one is strong it could last up to around 16 or more years depends on if the the CHIP u buy has a weak spot and u know know intill it dies

i still wouldnt go above 2.0V tho haha

hitechjb1 · Apr 28, 2004

c627627 said:
TC wrote on 04-16-04 11:45 PM:
Yes voltage - electrical components are rated for a specified range of voltage, amperage, and temperature. Exceed any one of those for long enough and you will destroy the component(s). I don't know why someone would say that as long as you keep a cpu cool nothing else matters.

Click to expand...

The below listed conclusions were made for as much as a 50% increase in Vcore voltage.

It was said that CPU weak spots not discovered during manufacturing stress tests may lead to permanent failure sooner than average, not the fact that you�re running too much voltage through your adequately cooled system. If this is not true, conclusion number 4. below needs to be changed.

1. Increase of 60 MHz at the expense of increasing 0.2 volts is not recommended.

2. Assuming expected CPU life time is 10 years, 30% overvolt statistically reduces it to six years.

3. 30% overvolt for DLT3C's is 1.95 volts and for Bartons and DUT3C T-Breds, about 2.1 volts. (Remember to keep the on die CPU temperature under 65 C.)

4. Highest attainable stable Overclock cannot be reduced within months provided on die CPU temperature is kept under 65 C, however:

5. CPU weak spots not discovered during manufacturing stress tests may lead to permanent failure sooner than average:

Originally posted by hitechjb1
A 10% increase in Vcore, would shorten the failure time to 83% of nominal failure time.
A 20% increase in Vcore, would shorten the failure time to 69% of nominal failure time.
A 30% increase in Vcore, would shorten the failure time to 59% of nominal failure time.
A 50% increase in Vcore, would shorten the failure time to 44% of nominal failure time.

Click to expand...

Nominal failure time is assumed to be 10 years at default voltage.

I relook at the thread, ....

Even these "rules" are not quoted in the context of the detailed analysis, as long as rule 4 is not used in isolation without rule 2 (and the other rules), both of which spell that both voltage and temperature play a role in affecting chip failure time, due to electromigration. It does not say voltage is not an issue if chip temperature is kept low, but was stated to the contrary that voltage plays a major rule.
Quote:

"Vcore is increased by 10%, on the average, the current density in the wires inside a chip would be increased by 10% (assume uniformly distributed current density), keeping temperature the same, so the failure time would be shorten by 17%. So

A 10% increase in Vcore, would shorten the failure time to 83% of nominal failure time.
A 20% increase in Vcore, would shorten the failure time to 69% of nominal failure time.
A 30% increase in Vcore, would shorten the failure time to 59% of nominal failure time.
A 50% increase in Vcore, would shorten the failure time to 44% of nominal failure time."

The 65 C number is based on 20 C below a max temperature specificiation. One can use a lower relative number instead of 65 C. The key point is temperature increase affects failure time, quantitatively.
Quote:

"An increase of temperature by 20 C over nominal max temperature would roughly result in doubling the electromigration rate, hence shortening the failure time to 50%. But that is measured by the temperature over the max temperature specification (above 85 C), which is not likely to happen in daily overclocking."

So the simplified rules are basically correct but do not reflect a complete picture of the subject especially when a more detailed technical matter is sought.

Effect of high Vcore and electromigration on CPU failure time (page 15)

Effect of high Vcore and electromigration on expected failure time for Tbred B/Barton (page 15)

What could damage a chip/CPU permanently? (page 15)

hitechjb1 · Apr 28, 2004

makaka said:
just question so it is safe to run with 1.95 v barton 2500+ at 2500mhz with 60cfull load(prime95) i have read all thread and found that i cpu temp should not go over 65c but people in the forum recommand 50c what is the estimation of the life time of this chip ? maybe 1, 2 , 5 years?
what do u think?

From the AMD tech doc for model 8 and model 10, the max die temperature is 85 C.
QUOTE from AMD doc,
"Thermal design power represents the maximum sustained power dissipated while executing publicly-available software or instruction sequences under normal system operation at nominal VCC_CORE . Thermal solutions must monitor the temperature of the processor to prevent the processor from exceeding its maximum die temperature."

Since we don't have the data about how long the life expectancy (an average number) of a certain CPU at certain voltage and temperature, so making some ASSUMPTIONS, say it is 10 years at nominal voltage 1.5 V and nominal temperature of 65 C (20 C lower than the max die temperature specification).

Now if the CPU is running at 1.8 V and 65 C, compared to the reference life expectancy number, it is 20% higher in voltage and temperature is within the reference temperature, so the life expectancy would decrease to 69%.

But if the 10 years assumption is for 1.5 V and 45 C, then it is 20% higher in voltage and 20 C higher in temperature, then the life expectancy would decrease to 34.5% (69% due to voltage and then 50% due to temperature).

These numebrs should be used as a guideline and not be taken as absolute as there are other factors that can affect the outcome. They are statistically reflecting the behaviour of a large sample, a particular CPU can deviate from that.

High Voltage as long as you keep the temperatures low?

c627627

c(n*199780) Senior Member

dead_man311

Member

TruckyJ

Member

hitechjb1

Senior Member

hitechjb1

Senior Member

hitechjb1

Senior Member

hitechjb1

Senior Member

c627627

c(n*199780) Senior Member

hitechjb1

Senior Member

hitechjb1

Senior Member

c627627

c(n*199780) Senior Member

makaka

Member

Steven4563

Member

hitechjb1

Senior Member

hitechjb1

Senior Member

Similar threads