Easy method to predict OC capabilities in current Nvidia GPU's

[saaya]

enthusiast sharing useful knowledge, i thought it just doesnt work like this anymore these days?
good to see its not true... big thumbs up to you petri

interesting that lower leakage is better... i would have thought higher leakage is better under ln2...

[[XC] Synthetickiller]

This is perfect benchmark for stability.
http://hwbot.org/rankings/benchmark/...28dx11%29/info

Download full version and run the extreme preset. If it can pass, it is game stable almost for sure.

Thanks for this info. I'm downloading it as I type this.

I'll do benching tomorrow with Heaven to see how far I can get with this card.

I was still able to bench with STALKER and keep 800/1600/1750 stable right now on stock volts (1.050). I'll just see how far I can push this before moving the voltage up.

What is safe voltage for 24/7 on these cards? Some are binned low while mine is on the upper end.

[SF3D]

enthusiast sharing useful knowledge, i thought it just doesnt work like this anymore these days?
good to see its not true... big thumbs up to you petri

interesting that lower leakage is better... i would have thought higher leakage is better under ln2...

Hehe Sascha! If I have not made any agreement with some third party, I am glad that I can share my information. No one can limit our freedom of speech on this one

AMD have been using low leakage chips long time. 5970 would not have been possible with normal Cypress GPU's. AMD also sold these low leak chips to vendors with better price.

Leakage and LN2 is interesting story. There is many other variables, which affect overclockability on LN2, so it is almost impossible to say, what some chips should do. On reference GTX 480 cards there seems to be almost perfect scaling, but with custom design like MSI lightning, it is more complicated. Power supply part is in major role on these cards, so high leakage and better PWM can produce very good results.

Thanks for this info. I'm downloading it as I type this.

I'll do benching tomorrow with Heaven to see how far I can get with this card.

I was still able to bench with STALKER and keep 800/1600/1750 stable right now on stock volts (1.050). I'll just see how far I can push this before moving the voltage up.

What is safe voltage for 24/7 on these cards? Some are binned low while mine is on the upper end.

That is the problem with these high VID cards. You are already near the heat and voltage limits and you don't have much headroom. You propably need to raise voltage and then you will get too hot very soon. It is sad, that we have this type of GPU's in the cards which cost so much.

Try with fixed voltage, when you test Heaven. (1.07V and 100% fan) Then we can compare the results with others. Then you can of course try how high it can go with more voltage.

I've notice this too on my 3 480 i,ve had. 1=1038mv, 2=980mv 3=1075mv

on water
first one max @ 930
2nd one max @ 950
3rd one max @ 888

you don't really need nvidiainspector, just check stock voltage in afterburner
i wish that 2nd card didn't died it looked like a monster!!!

Yes, afterburner show it as well.

That second card would have been very good on LN2 as well. Why all the good ones need to die I have one similar card and that is so far the lowest VID value I have seen. (0.980V or 0.988V)

3rd one is exactly like bad cards are.

[saaya]

On reference GTX 480 cards there seems to be almost perfect scaling, but with custom design like MSI lightning, it is more complicated. Power supply part is in major role on these cards, so high leakage and better PWM can produce very good results.

that makes perfect sense...

hey, when are you coming back for gooc?
send me a text, you got my new number now

[cowie]

hmmm idk
2 460's tested for this method of "binning"

1. 1.00v air 920 ln2 1090 1.21v

2. 1.027 air 920 ln2 1120. 1.21v

I swear thuo some of my 4xx cards should say POS in that box

[SF3D]

Hmm..it might be that P0 VID can not make any difference in some GTX 460, cause some other things are limiting. In GTX 460 case it is most likely the PWM part and temperature.

[joger]

I'm testing my two cards now in 3DMark Vantage GT1&2

Gainward Ref.(Green PCB) 1.063v stock vid: 850 core at 1.075v (MSI AB, can't take real measurements)

Asus Ref. (Orange PCB 1.013v stock vid: 870 core at at 1.075v (MSI AB)

Hope this helps

[dinos22]

gotta check my cards

[Riska]

is this P=0 voltage something in the bios or on the card because i loaded my Asus gtx480 with the Hipro bios and didnt save the original bios an my card have 1075mv in P0

[SF3D]

is this P=0 voltage something in the bios or on the card because i loaded my Asus gtx480 with the Hipro bios and didnt save the original bios an my card have 1075mv in P0

I am not sure can it be overflashed, but all my cards still have original P0 values after bios flash. So this value is stored in the GPU like VID values are in CPU.

1.075V seems to be highest stock P0 I have seen. How does the card overclock?

[Riska]

it is weird when i set 1070mv in afterburner og nvidiainspector it is way higher measured and i have to set it at 1056mv to measure 1069mv but af shows it at 1029mv in AF wierd but at 1069 id did 855mhz on core at 100% fan in vantage

[Riska]

In nivdiainspector P0 is level3 for my card is that right?

[rusty]

In nivdiainspector P0 is level3 for my card is that right?

I've got the same p0 level 3. can someone explain it ?

[mk-ultra]

it is weird when i set 1070mv in afterburner og nvidiainspector it is way higher measured and i have to set it at 1056mv to measure 1069mv but af shows it at 1029mv in AF wierd but at 1069 id did 855mhz on core at 100% fan in vantage

don't worry about that AF is not that precise at reading voltage, from what i've seen the value you set is pretty much the voltage you are running, forget about what AB shows after applying.

[Funky]

I got my GTX 470 Twin Frozr II today.

Check this out.





havent started ocing yet. Room is at 32°C at the moment. My A/C has gone to refilling of gas, should come in few hours. After that its oc time

[Mauro]

I also have 0.925V Asus GTX470 but I never actually tested how far it can go.

Furthest I went was simply putting voltage to 1,087V, fan on 100% and than tried 860 on core. Didn't actually put much effort in finding core clock, 860 was just random number I chosen.
Passed Heaven and Vantage without problem.
I should really once take time and see how far it can go (currently working on stock )

[luke997]

#1 Asus GTX 480 - P0 = 1.012v - 920 @1.213v - 860 @ 1.025
#2 Asus GTX 480 - P0 = 1.012v - 920 @1.213v - 860 @ 1.025
#3 Asus GTX 480 - P0 = 1.025v - 915 @1.213v - 860 @ 1.025

[Olivon]

Interesting thread ! Thanks Petri

[[XC] Synthetickiller]

I got my GTX 470 Twin Frozr II today.

havent started ocing yet. Room is at 32°C at the moment. My A/C has gone to refilling of gas, should come in few hours. After that its oc time

Lucky! Much lower vid than mine. I'm jealous!

Please post what you can get out of it.


I'm a little nervous to bump th voltage on my card as its already high and this is a gaming card, not just a benching card I can throw money away on.

[Funky]

Lucky! Much lower vid than mine. I'm jealous!

Please post what you can get out of it.


I'm a little nervous to bump th voltage on my card as its already high and this is a gaming card, not just a benching card I can throw money away on.

http://www.xtremesystems.org/forums/...&postcount=709


afterburner is broken at the moment. I want the damn triple voltage control card :P
MSI still havent released it.

[Mauro]

Asus GTX470 VID 0.925

Managed Heaven on highest settings with core on 870. Voltage on 1.087V, fan on 100%.
Memory not overclocked.

http://img4.imageshack.us/img4/3203/heavencore870.jpg

Still, need some time to see maximum it could go and to install and bench Vantage.

[Cookiesowns]

Man. Both my GTX480s SC suck. One of them is 1.063V the other is 1.075 >:|

[PiLsY]

MSI 470 VID 1050mv max oc air 766mhz/1866mhz 94c load 1087mv // water 850mhz/2066mhz 1137mv 49c load. More volts causes grey/pink screens. Need ocp mod?
POV 470 VID 1037mv max oc air 788mhz/1840mhz 90c load 1087mv // water 850mhz/1990mhz 1150mv 48c load. More volts doesnt help much, could maybe hit 866mhz.
PNY 470 VID 1000mv max oc air 833mhz/1866mhz 72c load 1087mv // new card - block tomorrow.

Holding true here so far... New PNY card looks to be impressive. POV card went in a friends pc. MSI is mine as is the PNY. May have to try and sell the MSI to buy another one to match the PNY - ie not a dog. The PNY card was an import, the POV card and MSI card are both from different UK retailers. I suspect some sorting is going on in order to be able to uphold the longer warranties on certain card models (and outside europe - US warranties are way better in general). Bought the PNY on that hunch. Glad I did, but can't say im right on that basis. I havent seen a good UK card yet outside of EVGA (now lifetime warranty here too on SC/SSC/FTW parts co-incidentally). Probably just paranoia .

[Timoleros]

There is my Sparkle GTX480
P0=0.987V / MAX OC on air = 914MHz@1113v in 3DMark Vantage

[saaya]

P0 - air - water - ln2 - card - user

1.075 - NA - 888 - NA - GTX480 - mk-ultra
1.063 - 775 - NA - 1100 - GTX480 - SF3D
1.063 - 850 - NA - NA - Gainward 480 - joger
1.038 - NA - 930 - NA - GTX480 - mk-ultra
1.025 - 915 - NA - NA - Asus 480 - luke997
1.013 - 910 - NA - NA - GTX480 - STEvil
1.013 - 870 - NA - NA - Asus 480 - joger
1.012 - 920 - NA - NA - Asus 480 - luke997
1.012 - 920 - NA - NA - Asus 480 - luke997
0.980 - NA - 950 - NA - GTX480 - mk-ultra
0.987 - 914 - NA - NA - Sparkle 480 - Timoleros
0.985 - 900 - NA - 1300 - GTX480 - SF3D

1.027 - 920 - NA - 1120 - GTX460 - cowie
1.000 - 920 - NA - 1090 - GTX460 - cowie
0.987 - NA - NA - NA - MSI 460 Cyclone - dafour
0.950 - 902 - NA - NA - GTX460 - ButtSprinkles

1.050 - 800 - NA - NA - MSI 470 TFII - [XC] Synthetickiller
1.050 - 766 - 850 - NA - MSI 470 - PiLsY
1.037 - 788 - 850 - NA - MSI 470 - PiLsY
1.000 - 833 - NA - NA - MSI 470 - PiLsY
0.925 - 870 - NA - NA - Asus 470 - Mauro

there seems to be a higher chance to have a good clocking card if it has a low vid, but the differences seem to be rather small, at least on air and from the numbers you guys posted so far...

thesis: lower vid = higher oc
the 3 480s with the highest voltages all clock notably below average
one of the worst reported 470 cards has the highest reported vid
the 2 470s with the lowest vid are notably better than the rest

counter thesis:
rank 4, 5 and 6 of the 480s clock almost the same as the 480s with the lowest vids, at least on air and water
the card with the lowest reported vid for a 480 does great on ln2 but is slightly below average on air
even among cards with the exact same vid there are fluctuations of 33-75mhz, while the worst to best clocks are only 125mhz apart

from these results there is definitely a trend id say, but its not linear scaling...
lower vid = higher chance to have a good clocker, but the most obvious results are that high vid = almost certainly a bad clocker

for 470s the trend seems to be stronger than for 480s from these numbers, which makes sense as there is more binning going on for 470s than 480s...
theres a much wider range of gpus with different specifications that can end up as a 470 than there are for a 480 candidate.

for 460s there is no clear trend from these few numbers...
the lowest vid clocks the worst and cowie tried two cards and found the higher vid card to clock slightly higher...