FluffyZack
Derp. I'm stupid. I clicked a link that Sajin put on a different thread and thought I was on the same one. Ignore this:
 
Oh just saw Sajin's post about ASIC. So the higher the ASIC the better? I swear the other dudes were just saying it's better for it to be lower? Unless I read it wrong

On air cool, higher is better, on water cool / LN2, lower is better

Muromets
_Nite_ I see  in GPU-Z confirmation of your statement regarding overclocking:
 
But I am curious why this is the case / what's the "physics" of it. Can you please elaborate why there is a difference in "polarity" of the ASIC score between air and water/LN2 cooling? Thanks
 

That I can't answer sorry

Muromets
But I am curious why this is the case / what's the "physics" of it.

It is actually very easy to understand what is the physics, it is explained in Wikipedia, see http://en.wikipedia.org/w..._switching_and_leakage
Circuits inside GPU dissipate power - in wiki there are described all various circuits and mechanics - e.g. subthreshold leakage (that could make 50% or total power consumption in some cases).
 
High Asic quality means smaller power dissipation thus less heat generated, smaller power consumption = possibility of better overclock
Small Asic quality is suitable only for nitrogen/subzero  cooling (not water cooling as people try to persuade here), because smaller asic quality means lot of heat generated so GPU is not frozen (=stopped working) by nitrogen as it would be in case of high Asic quality.
It is exceptional that one has nice overclock with low Asic quality, if she overclocked similar card with high Asic quality overclock would be better.
 

TrekCZ

Muromets
But I am curious why this is the case / what's the "physics" of it.

It is actually very easy to understand what is the physics, it is explained in Wikipedia, see http://en.wikipedia.org/w..._switching_and_leakage
Circuits inside GPU dissipate power - in wiki there are described all various circuits and mechanics - e.g. subthreshold leakage (that could make 50% or total power consumption in some cases).
 
High Asic quality means smaller power dissipation thus less heat generated, smaller power consumption = possibility of better overclock
Small Asic quality is suitable only for nitrogen/subzero  cooling (not water cooling as people try to persuade here), because smaller asic quality means lot of heat generated so GPU is not frozen (=stopped working) by nitrogen as it would be in case of high Asic quality.
It is exceptional that one has nice overclock with low Asic quality, if she overclocked similar card with high Asic quality overclock would be better.
 

That doesn't describe the whole physics of it. What you missed is that the lower ASIC quality causes voltage leak to ground which tends to "drain" fuzzy logic errors (overshoot/undershoot). Since the cooling is effective enough to keep temperatures in check and keep the circuitry at ideal operating temperature, the leakage doesn't negatively affect the circuit's performance. Effectively, the leakage helps stabilize the logic signal and prevent logic errors.

A higher ASIC quality tends to keep the excess harmonics and overshoot / undershoot when logic states change; which is bad. At higher and higher frequencies, harmonics build up more and more on top of the logic signal and can create errors.

Take a look at square wave harmonics if you are interested. The logic signal is effectively a random version of a square wave. The low ASIC quality leakage effectively acts like a filter which filters out some of the harmonics.

Asic quality seems to matter. Mine's 63% and my overclock is weaksauce.

Everyone seems to be dead set on asic % meaning quality, when really that is not what it is. The ASIC % refers to the amount of power leakage the ASIC chip has. The VBIOS has values inside it to set a certain set of voltages at a given range of ASIC %'s so that the entire video card can stay within the stated TDP (TDP is more or less for the purpose of designing a proper cooling solution so the product can stay within spec).
 
A higher % = Higher leakage = Higher heat and power consumption at a given core speed and voltage which, with the new GPU Boost technology, can limit your maximum overclock due to power constraints. They will also generally have a lower stock voltage since a higher stock voltage can cause the GPU to go beyond the stock TDP.
 
On the other end of the spectrum, a lower % = lower leakage = lower heat and power consumption at a given core speed and voltage, which can increase your maximum possible overclock within a certain power envelope since it will be using less power at the same speed and volts as a higher ASIC % card. They will also have a higher stock voltage simply for the purpose of having a uniform power draw across different cards with different ASIC %'s.
 
I tested this with a number of GPU's (About 20) a little before the time GPU Boost started becoming standard. They all seemed to follow this pattern. Setting maximum allowable Power %, a fixed core frequency, and a fixed core volts, the power difference was definitely measurable. A 63% ASIC card with a stock voltage of 1.175v and core set to 1GHz consumed about 100 watts less than an identical model card with an 85% ASIC (stock volts of 1.05 but set to 1.175v) set to the same volts and core. In fact, it consumed so much power that it shut off the system.
 
I've always had excellent experiences using low ASIC % cards if I can ever find them (65% or less) whereas 80% or higher cards I've always seemed to have higher temperatures and a harder working fan.
 
As far as actual overclockability of the chips, it is still luck of the draw. A chip with lower power consumption may have more headroom as far as power constraints go compared to one with higher power consumption, but having a magic chip doesn't seem to correlate with the thermal leakage (AKA ASIC %) of it.

What, is this ssateneth an EVGA marketing guy? All my EVGA cards lately have been lower ASIC %... the only one to say it is good on air is ssateneth.... 

my ssc is 72.5% and stock clock is supposed to be 1190/1342(boost) but when I run at stock settings it goes up to 1405 (rarely just shows this as max on graph) and usually stays around 1380/1367 (goes to 1393/1354 aswell) but its super random... when I o/c the card I can get it up to a max of 1505 thats rare and will usually stay around 1480/1455 I guess my card isnt as good as 1500+ G1 because of the power limit (g1 has 300w via hardware limited to 280w via bios/software) the evga card only has 200w I believe (read this on OC.net in the g1 vs msi vs strix thread)

Fennicillin
Going to have to retract my verdict on ASIC quality a bit, I found out Precision X has been causing my crashes and wonky benchmark scores while overclocked by bugging out and halving my total clock speed. Currently at 1106/1172 and 1800 on the memory and stable using Afterburner. Going to push further later.

did you have a any driver resets? precision going back to default settings when that happens is normal.

_Nite_

Fennicillin
Going to have to retract my verdict on ASIC quality a bit, I found out Precision X has been causing my crashes and wonky benchmark scores while overclocked by bugging out and halving my total clock speed. Currently at 1106/1172 and 1800 on the memory and stable using Afterburner. Going to push further later.

did you have a any driver resets? precision going back to default settings when that happens is normal.

It's not resetting to defaults, the OC profiles save. My card keeps locking itself to 549Mhz in game after about 5-10 minutes. This only started happening since I got Precision. It freezes up Metro:LL and Arkham City constantly. I don't know about other games cause those are the ones I've been grinding lately but I'll assume it will do that others. It's actually happening with Afterburner now too. Clean reinstalling drivers doesn't fix it and the clocks stay locked at half in game until I restart.

koroem
549mhz lock is a usually a symptom and sign of a a driver reset. Check your event logs for the driver crash I bet its listed there.

Ran DDU and it showed 340 drivers and 344.11 which is weird since I always check clean installs. Ran it and hoping the problem is just that conflict and not overclock instability.

I got 76.3% on my EVGA GTX465 SC video card

My 780 Ti Classified is 66.9%... 

My Evga GTX 670 SC Acx 2.0 is 68%

Are you guys all using custom BIOSes when you overclock? Or just using the stock bios?