Intel 45nm vs. 65nm, does the Hi-K metal gate not able to withstand voltage as well?

[GundamF91]

I heard that 45nm is not able to do as well in terms of voltage increase (proportionally) compared to 65nm? So they're more vulnerable to electromigration when more voltage is pushed on them, ie. Q9450 v. Q6700

 

[BadRobot]

the rumor is the 45nm fries easier/faster so you shouldn't be using 1.6v on the chip /shrug

 

[GundamF91]

Yeah, but I do'nt think 65nm could handle 1.6v either, at least not for long. So proportionally though, I'm wondering if it's 45nm does worse, ie. +20% vCore for 45nm vs. 60nm.

 

[aigomorla]

Originally posted by: GundamF91
Yeah, but I do'nt think 65nm could handle 1.6v either, at least not for long. So proportionally though, I'm wondering if it's 45nm does worse, ie. +20% vCore for 45nm vs. 60nm.

the conroe is one durable chip.

Ive tortured the living hell out of a E2180 b4 retiring it. Gave it as much as 1.65V which came out to 1.63Vcore under load.

This isnt safe, but im just showing you guys its durable as hell. Personally i was trying to kill it and it just wouldnt die.

The Wolfdale. BAH... that hurts still today, considering this one toped at 4.2ghz @ 1.41Vcore. :\

Anyhow i killed it. 45nm + 1.5V+ = kiss that baby good bye.

My QX is set at 1.425V. Under load it goes to ~1.41 - 1.42

I wouldnt dream of setting it any higher then that. If i was on air tho, i wouldnt push more then 1.4V on it.

 

[BadRobot]

it seems to me that 45nm hits its ceiling faster for ghz AND volt

So proportionally it is not any more susceptible to frying.
However, 1.45 or 1.5 might fry a 45nm where it wouldn't for a 65nm according to some people.

 

[aigomorla]

Originally posted by: BadRobot
it seems to me that 45nm hits its ceiling faster for ghz AND volt

So proportionally it is not any more susceptible to frying.
However, 1.45 or 1.5 might fry a 45nm where it wouldn't for a 65nm according to some people.

45nm + voltage = Crack to an addict.

Thats the simplist way to put it.

The more you push, the faster the cpu will go, but will also eat its life away NOS on race car.

 

[error8]

Well it seems logical, that the smaller the electric circuits are, the more affected they should be at higher voltages. But these wolfdales are pretty sensible really. 🙂

 

[aigomorla]

Originally posted by: error8
Well it seems logical, that the smaller the electric circuits are, the more affected they should be at higher voltages. But these wolfdales are pretty sensible really. 🙂

if im lucky i'll have 2 more 45nm chips to play with. :T

No i dont intend to kill these guys either. There refreshing my kentfields. :T


Oh man, my cousins are gonna bust out world war 3 now. Where is that lottery wheel.

 

[mancunian]

Originally posted by: aigomorla
45nm + voltage = Crack to an addict.

Thats the simplist way to put it.

The more you push, the faster the cpu will go, but will also eat its life away NOS on race car.

So I'm assuming that this E7200 at 3.6Ghz and 1.216v in CPU-Z should be safe enough? It stays cool, but voltage is something I know much less about with on the Intel chips, had AMD for years.

Given it took you 1.5v to kill a 45nm chip, I should be at a safe threshold? Right?

Just wanting opinions of those more in the know than me.


Cheers 🙂


Edit: Apologies if this is seen as hijacking the thread, but the original post seemed to be one wanting a bit of discusson. I can start a new one if necessary.

 

[Extelleron]

Intel's 45nm chips are definitely much more sensitive to voltage than the 65nm Conroe chips. Part of this is of course that operating voltages decrease with every smaller process. If you look at Pentium II's @ 250nm/350nm, you were talking about 2-3V. Now you see operating voltages of between 1-1.3V with Intel's 45nm chips.

It does seem like the 45nm chips are extremely sensitive to voltage though, more than is usual with a new process. Since the transistor materials are completely different, and this is the first time these new materials have been used, this makes sense. We should expect some changes in how chips react to voltage changes and the like with such a big change in what the chips are made of.

Just don't go beyond 1.4V on air cooling with a 45nm chip and you should be fine. If you are running 24/7, I would try to stick with 1.35V or less. These E8xxx are capable of so much with so little voltage, you do not need to pump 1.4V+ into them. It is not worth it for an extra 100MHz. I'd rather be @ 4GHz w/ 1.3V than 4.2GHz w/ 1.4V.

 

[zach0624]

Originally posted by: aigomorla

Originally posted by: error8
Well it seems logical, that the smaller the electric circuits are, the more affected they should be at higher voltages. But these wolfdales are pretty sensible really. 🙂

if im lucky i'll have 2 more 45nm chips to play with. :T

No i dont intend to kill these guys either. There refreshing my kentfields. :T


Oh man, my cousins are gonna bust out world war 3 now. Where is that lottery wheel.

can I be your cousin.... Please

 

[mancunian]

Originally posted by: Extelleron
Intel's 45nm chips are definitely much more sensitive to voltage than the 65nm Conroe chips. Part of this is of course that operating voltages decrease with every smaller process. If you look at Pentium II's @ 250nm/350nm, you were talking about 2-3V. Now you see operating voltages of between 1-1.3V with Intel's 45nm chips.

It does seem like the 45nm chips are extremely sensitive to voltage though, more than is usual with a new process. Since the transistor materials are completely different, and this is the first time these new materials have been used, this makes sense. We should expect some changes in how chips react to voltage changes and the like with such a big change in what the chips are made of.

Just don't go beyond 1.4V on air cooling with a 45nm chip and you should be fine. If you are running 24/7, I would try to stick with 1.35V or less. These E8xxx are capable of so much with so little voltage, you do not need to pump 1.4V+ into them. It is not worth it for an extra 100MHz. I'd rather be @ 4GHz w/ 1.3V than 4.2GHz w/ 1.4V.

Thanks Extelleron, that puts my mind at rest. So it seems some are getting 4+ Ghz with this chip, but are also perhaps putting in about 1.4v and upwards to get it.

Think I'll stick with 3.6 and 1.216v, plenty fast.


Thanks again.

🙂

 

[error8]

Originally posted by: mancunian

Originally posted by: Extelleron
Intel's 45nm chips are definitely much more sensitive to voltage than the 65nm Conroe chips. Part of this is of course that operating voltages decrease with every smaller process. If you look at Pentium II's @ 250nm/350nm, you were talking about 2-3V. Now you see operating voltages of between 1-1.3V with Intel's 45nm chips.

It does seem like the 45nm chips are extremely sensitive to voltage though, more than is usual with a new process. Since the transistor materials are completely different, and this is the first time these new materials have been used, this makes sense. We should expect some changes in how chips react to voltage changes and the like with such a big change in what the chips are made of.

Just don't go beyond 1.4V on air cooling with a 45nm chip and you should be fine. If you are running 24/7, I would try to stick with 1.35V or less. These E8xxx are capable of so much with so little voltage, you do not need to pump 1.4V+ into them. It is not worth it for an extra 100MHz. I'd rather be @ 4GHz w/ 1.3V than 4.2GHz w/ 1.4V.

Thanks Extelleron, that puts my mind at rest. So it seems some are getting 4+ Ghz with this chip, but are also perhaps putting in about 1.4v and upwards to get it.

Think I'll stick with 3.6 and 1.216v, plenty fast.


Thanks again.

🙂

With my E7200 I went as far as 1,41 V in bios ( 1,38 V load cpu-z ), on air for 3,9 ghz.
Right now, even though we're in the middle of a heat wave and I don't have any AC, I keep it at 1,36 V in bios ( 1,32 load cpu-z ) at 3,8 ghz.

So, I think that 1,4 V it's pretty safe, although nobody knows if the cpu's life is affected somehow ( probably is). E7200 seems a bit more tough over the E8XXX series. You can search the whole internet and you're not going to find someone that managed to kill or degrade this chip because of the voltage and there are some weirdos that pumped a lot of voltages into them , on air.

If you're on 3,6 ghz with 1,2 V, you'll probably touch 4 ghz with 1,3 V or something. I envy you. 😉

 

[wwswimming]

i don't know how it works for Hi-K and other semiconductor materials.

for many materials, voltage standoff is roughly proportional to thickness.

for Kapton (polyimide), voltage standoff for the first mil (.001 inch)
is about 5 KV, then slowly decreasing. after about .005, voltage standoff
is about 3 kV per mil. that's an insulator, not a semi-conductor.

i wonder what the standoff numbers are for semiconductor materials.
also, if voltage 'fatigue' is anything like mechanical stress fatigue, where
you add up numbers of cycles at different stress levels using something
called Miner's Analysis.

 

[Idontcare]

Originally posted by: wwswimming
i don't know how it works for Hi-K and other semiconductor materials.

for many materials, voltage standoff is roughly proportional to thickness.

for Kapton (polyimide), voltage standoff for the first mil (.001 inch)
is about 5 KV, then slowly decreasing. after about .005, voltage standoff
is about 3 kV per mil. that's an insulator, not a semi-conductor.

i wonder what the standoff numbers are for semiconductor materials.
also, if voltage 'fatigue' is anything like mechanical stress fatigue, where
you add up numbers of cycles at different stress levels using something
called Miner's Analysis.

The Hi-K (for insulating the drain from the gate) as well as low-k (for insulating line-to-line) dielectrics serve the exact same purpose and the concepts are identical.

The vocabulary differs of course as is common with all industries, but the physics and mathematics are identical (when properly applied).

To ascertain whether Intel's specific implementation of their 45nm materials, integration, and design rules render the CPU's of said node any more or less susceptible to electric field induced defect mechanisms than their 65nm node one must first account for the obvious differences in design rules first and formost.

A 65nm node with critical metal pitch of 180nm (90nm dielectric spacer, 90nm metal line) versus a 45nm node with critical metal pitch of 140nm (70nm dielectric spacer) simply cannot support the same absolute voltage or changes in voltage because of the obvious dielectric spacer change...unless the dielectric material is improved commensurate with the electric field increase.

This is but one of the many many reason operating voltage is reduced node-on-node from prior nodes.

No one can answer the question of whether Intel's 45nm process tech is more/less susceptible to electric field induced degradation (once the voltage has been normalized so as to normalize the electric field) compared to their 65nm process tech except Intel...and you can be sure they do know the answer as this is routinely and robustly characterized and optimized during the node development cycle well before production starts.

For us hacks, the typical enthusiast consumer who doesn't know jack about process technology let alone standard electric field effects, the least we could do is use the rule of thumb that operating voltage tolerance ought to be scaled by the skrink factor...45nm/65nm = 69%...so we should not expect our 45nm Intel chips to survive beyond a 69% voltage increase relative to the kind of voltage increase we expected our 65nm Intel chips to survive (this is just simple electrical engineering, no rocket science here).

So if your 65nm Q6600 with a VID of 1.3V was expected to survive a 0.3V overvolt (i.e. overvolted to 1.3+0.3 = 1.6V) you should not expect your 45nm Q9450 with a VID of 1.1V to survive a 0.3V*0.69 = 0.2V over-volt.

In other words, putting your Q9450 at Vcore of 1.1V+0.2V = 1.3V ought to be considered just as much of an "over-voltage" as putting your Q6600 at a Vcore of 1.6V.

Go above 1.3V on your Q9450 and expecting it to survive is about the same as putting more than 1.6V on your Q6600 and expecting it to survive...and this is just scaling the voltage for electric field induced degradation issues, no mystery here.