[Official] NVIDIA GTX 780 Ti Owner's Club - Page 1387 - Overclock.net - An Overclocking Community

Forum Jump: 

[Official] NVIDIA GTX 780 Ti Owner's Club

Reply
 
Thread Tools
post #13861 of 16563 (permalink) Old 10-25-2014, 02:08 AM
 
vulcan78's Avatar
 
Join Date: May 2014
Posts: 824
Holy crap guys, I don't want to bash all of the hard work on Skyn3t's behalf but if youre on a reference PCB youre better off sticking with the default vbios that came on the card. I flashed to the Skyn3t's revised SC ACX.rom as I was having stability issues at 1241 core / 1851 memory (+100 / +200) in Max Payne 3 with 3D Vision at 1440p, the additional voltage DID allow me to remain stable at 1254 core but no higher. The temps weren't an issue with that game. As far as stability goes, 1241 core on the default vbios is mostly stable in all other demanding games, and I'm talking Crysis 3, Metro LL, AC4: Black Flag etc. But I've since reflashed to the default vbios as I'm waiting for my Kraken G10's to arrrive (to be used in conjunction with Corsair H55's and the existing ACX mid-plate to address VRM cooling) and not wanting to put undue wear on my GPU's until then. I've revisited a few benchmarks to see what kind of temp difference and WOW I am really surprised.

Firestrike with Skyn3t SC ACX vbios:

1254 core / 1850 memory (yes 1281 core is stable IN FIRESTRIKE with this vbios but the highest I can reliably attain in games at 1.212v is 1254 core)

24,494 GPU

Temps:
GPU0: 73 C
GPU1: 66 C

Unigine Valley:

DX11
Ultra
3DVision
2560x1440
4xAA

Temps by end of bench:
GPU0: 87C
GPU1: 78C


Default SC vbios:

Firestrike

1241 core / 1850 memory

24,098 GPU

Temps:
GPU0: 66C (-7C)
GPU1: 62C (-4C)

Unigine Valley

Temps by end of bench:
GPU0: 74C (-13C)
GPU1: 66C (-12C)

Things were getting so out of hand with the Skyn3t vbios (primary was hitting 85C in AC4: Black Flag at 1440p - 3D Vision, removing the side-panel brought that down to about 80 C) that I felt the need to dial back the temp target from 95 to 75C, and even "linking" the temp and power target until I realized that doing so would reduce the power target to 80% with a temp target of 75C. The temps were still outrageous with the temp target reduced. This vbios is GREAT if youre under water, but even then, without the ability to increase the voltage over 1.212v on cards with reference PCB's, what's the point? I gotta hand it to the engineers and programmers at Nvidia and EVGA, they truly optimized the default vbios', at least as far as the SC ACX variant is concerned. Undoing their expertise feels like a fools errand.

To be clear, these temps were in an Air 540, second only to a test-bench in terms of airflow, with aftermarket fan's pushing significantly more CFM AND an aggressive, and I mean aggressive fan algorithm on the ACX coolers (40% RPM at 30C increasing linearly to 100% at 65C):

https://www.youtube.com/watch?v=XGtlksj-Fdc

I will probably stick with the default vbios when I get my G10's going, as it stands I'm still seeing 75C-78C primary in very demanding games.

An old adage comes to mind: "If it aint broke, don't fix it."

Edit:

If youre interested in the G10 AIO cooling solution, "Faceman" in the following thread has the most comprehensive mini-guide out there:

http://linustechtips.com/main/topic/232654-kraken-g10-help/

Also check out the G10 forum here on OCN:

https://www.overclock.net/t/1487012/official-nzxt-kraken-g10-owners-club

For $180 (what I paid for SLI, including everything "Faceman" recommends, G55's on sale for $50 shipped through Newegg on ebay and I already have 120mm fans for push-pull. Expect to pay about $100 per card) this is an excellent option for those with VRM cooling mid-plate cards such as Asus DCUII, EVGA ACX and MSI Lightning as testing has actually shown a 10-20C reduction in VRM temps with the G10's 92mm fan blowing on these mid-plates, an improvement to be sure. Peak temps are typically 50-55C, down from 75-80C AND the heat is expelled from your case AND unlike GPU specific water-blocks, the G10 brackets are forward compatible with GTX 970, 980 and a host of AMD cards.

https://www.youtube.com/watch?v=oU6Kd2VwYag
vulcan78 is offline  
Sponsored Links
Advertisement
 
post #13862 of 16563 (permalink) Old 10-25-2014, 02:16 AM
 
vulcan78's Avatar
 
Join Date: May 2014
Posts: 824
Quote:
Originally Posted by maneil99 View Post

Theres no harm sliding the power slider all the way to 200% right?

The guy I ordered my 780 Ti from had issues with getting his 780 Ti voltage to stick in PX so he used Afterburner. He has his boost clock set to 1137 but it often spikes into the 1350+ area in games and benches. Is this card a dud or is this because AB doesn't support the 780 ti properly?

If you want to unnecessarily cook your VRM's go ahead and keep Power Target at 200%. I saw NO correlation between performance and stability at 160% power target vs. 106% with my SC ACX on the Sky3t vbios, only about 5C higher temperatures and an increase, yes and increase, in artifacts in Unigine Valley. See post above.

Edit: Zero difference in Firestrike bench figures as well.
vulcan78 is offline  
post #13863 of 16563 (permalink) Old 10-25-2014, 02:18 AM
New to Overclock.net
 
maneil99's Avatar
 
Join Date: Nov 2011
Posts: 656
Rep: 4 (Unique: 4)
Quote:
Originally Posted by vulcan78 View Post

If you want to unnecessarily cook your VRM's go ahead and keep Power Target at 200%. I saw NO correlation between performance and stability at 160% power target vs. 106% with my SC ACX on the Sky3t vbios, only about 5C higher temperatures and an increase, yes and increase, in artifacts in Unigine Valley. See post above.

Edit: Zero difference in Firestrike bench figures as well.
I was under the impression that the GPU will only take as much as it needs. Hence why go to 200% so it can take all it wants without throttling.
maneil99 is offline  
Sponsored Links
Advertisement
 
post #13864 of 16563 (permalink) Old 10-25-2014, 02:20 AM
New to Overclock.net
 
maneil99's Avatar
 
Join Date: Nov 2011
Posts: 656
Rep: 4 (Unique: 4)
Quote:
Originally Posted by vulcan78 View Post

Holy crap guys, I don't want to bash all of the hard work on Skyn3t's behalf but if youre on a reference PCB youre better off sticking with the default vbios that came on the card. I flashed to the Skyn3t's revised SC ACX.rom as I was having stability issues at 1241 core / 1851 memory (+100 / +200) in Max Payne 3 with 3D Vision at 1440p, the additional voltage DID allow me to remain stable at 1254 core but no higher. The temps weren't an issue with that game. As far as stability goes, 1241 core on the default vbios is mostly stable in all other demanding games, and I'm talking Crysis 3, Metro LL, AC4: Black Flag etc. But I've since reflashed to the default vbios as I'm waiting for my Kraken G10's to arrrive (to be used in conjunction with Corsair H55's and the existing ACX mid-plate to address VRM cooling) and not wanting to put undue wear on my GPU's until then. I've revisited a few benchmarks to see what kind of temp difference and WOW I am really surprised.

Firestrike with Skyn3t SC ACX vbios:

1254 core / 1850 memory (yes 1281 core is stable IN FIRESTRIKE with this vbios but the highest I can reliably attain in games at 1.212v is 1254 core)

24,494 GPU

Temps:
GPU0: 73 C
GPU1: 66 C

Unigine Valley:

DX11
Ultra
3DVision
2560x1440
4xAA

Temps by end of bench:
GPU0: 87C
GPU1: 78C


Default SC vbios:

Firestrike

1241 core / 1850 memory

24,098 GPU

Temps:
GPU0: 66C (-7C)
GPU1: 62C (-4C)

Unigine Valley

Temps by end of bench:
GPU0: 74C (-13C)
GPU1: 66C (-12C)

Things were getting so out of hand with the Skyn3t vbios (primary was hitting 85C in AC4: Black Flag at 1440p - 3D Vision, removing the side-panel brought that down to about 80 C) that I felt the need to dial back the temp target from 95 to 75C, and even "linking" the temp and power target until I realized that doing so would reduce the power target to 80% with a temp target of 75C. The temps were still outrageous with the temp target reduced. This vbios is GREAT if youre under water, but even then, without the ability to increase the voltage over 1.212v on cards with reference PCB's, what's the point? I gotta hand it to the engineers and programmers at Nvidia and EVGA, they truly optimized the default vbios', at least as far as the SC ACX variant is concerned. Undoing their expertise feels like a fools errand.

To be clear, these temps were in an Air 540, second only to a test-bench in terms of airflow, with aftermarket fan's pushing significantly more CFM AND an aggressive, and I mean aggressive fan algorithm on the ACX coolers (40% RPM at 30C increasing linearly to 100% at 65C):

https://www.youtube.com/watch?v=XGtlksj-Fdc

I will probably stick with the default vbios when I get my G10's going, as it stands I'm still seeing 75C-78C primary in very demanding games.

An old adage comes to mind: "If it aint broke, don't fix it."
Did you make sure the fan curve was the same.
maneil99 is offline  
post #13865 of 16563 (permalink) Old 10-25-2014, 02:43 AM
 
vulcan78's Avatar
 
Join Date: May 2014
Posts: 824
Quote:
Originally Posted by maneil99 View Post

I was under the impression that the GPU will only take as much as it needs. Hence why go to 200% so it can take all it wants without throttling.

I was under the same impression! No improvement in performance (Firestrike) but about 5C higher temps (Unigine Heaven) and artifacts! Now that I think about it, I believe the artifacts were a side-effect of the primary GPU sitting at 92 C in Heaven! Dialing back Power Target from 160% to 106% brought primary temp down to 87C and hence less artifacts. I hate to think about what kind of temps my VRM's were seeing with the core at 92C; I've heard anything higher than say 85-90C on the VRM spells the death knell for GK110. I'm glad it was for only about 10-15 minutes total while I played around with the Power Target settings and did runs with both the side-panel on and off; I imagine playing Skyrim with 100+ mods and an ENB for hours on end with similar temperatures would just about cook your card in a few months time.
Quote:
Originally Posted by maneil99 View Post

Did you make sure the fan curve was the same.

Absolutely. I added rather lengthy post-script edit to my post you quoted pertaining to the G10 AIO cooling solution. If youre running a card with a VRM cooling plate you'll probably want to have a look at the G10.
vulcan78 is offline  
post #13866 of 16563 (permalink) Old 10-25-2014, 02:49 AM
 
Join Date: Sep 2013
Location: Ocala, FL
Posts: 567
Quote:
Originally Posted by maneil99 View Post

Theres no harm sliding the power slider all the way to 200% right?

The guy I ordered my 780 Ti from had issues with getting his 780 Ti voltage to stick in PX so he used Afterburner. He has his boost clock set to 1137 but it often spikes into the 1350+ area in games and benches. Is this card a dud or is this because AB doesn't support the 780 ti properly?

Nope none whatsoever.
HighTemplar is offline  
post #13867 of 16563 (permalink) Old 10-25-2014, 03:04 AM
New to Overclock.net
 
mongomunken's Avatar
 
Join Date: Sep 2013
Posts: 19
Rep: 0
Quote:
Originally Posted by OccamRazor View Post

Not that buggy Precision 15 version, use the old one: http://forums.videocardz.com/topic/532-evga-precision-x-421/
I would be a little more conservative on the clocks if i were you... 1500mhz is a bit high don't you think? Average high 780Ti OCs are in the 1300mhz!
Set the voltage to 1,212V and increase the clocks every time a benchmark or a game loop is successful!


Cheers

Occamrazor

Thanks, now I got K-Boost working. Seems slightly more stable.

Do you mean 1500mhz is dangerous or something? Should I stop at 1300mhz?
I set offset to +200 (1245MHz) and increase by 20 after running about half of Unigine Valley.
I ended up at a stable +440 (1485MHz). I was amazed as I hadn't been close to this with afterburner. Verified the clock in gpu-z.

Then I restart my computer. Suddenly Valley is instantly crashing, and I have to go down to +240 (1285MHz) to be stable.
mongomunken is offline  
post #13868 of 16563 (permalink) Old 10-25-2014, 04:15 AM
What is old is new
 
OccamRazor's Avatar
 
Join Date: May 2012
Location: Germantown MD
Posts: 4,447
Rep: 631 (Unique: 322)
Quote:
Originally Posted by maneil99 View Post

Theres no harm sliding the power slider all the way to 200% right?
The guy I ordered my 780 Ti from had issues with getting his 780 Ti voltage to stick in PX so he used Afterburner. He has his boost clock set to 1137 but it often spikes into the 1350+ area in games and benches. Is this card a dud or is this because AB doesn't support the 780 ti properly?

Read my articles about it:

Warning: Spoiler! (Click to show)
"You can always choose with the slider what ever power target your card will have: Min 300W - Max 600W, my advice has always been to leave PT at default 300W = 100% and only increase it if you see stutters or frame drops!
As soon as you increase the slider and your card is power hungry, the voltage will allow more current into the card (THE AMOUNT ALWAYS DEPENDING ON THE SLIDER) and usually with stock air coolers (ACX as well) that are not capable of handling more than 350W of continued heat but to a much lesser extent a split second heat spike in the mosfets (VRM's)! on top of that i see lots of people using kombustor, mining, oc scanner, occt etc without knowing exactly how those programs should be run,( If anyone interested in mining with 780Ti, PM Gordan for the safest settings) loading the card to an extreme generating more heat that the cooler can handle!



In RED the Power Mosfets (Actually DrMos Modules for High current DC-DC conversion) for core and Memory (U8-U13 = core, U98/U99 Memory)
(The problem with Mosfets is their tiny size, they generate huge amounts of heat and only have a very small size making it very difficult to dissipate all that heat effectively, if there is a heat spike, even with LN2 they just "blow"...)
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

In Yellow the Inductors (Current) [R22] for the core and [R33] for the memory

In BLUE the Capacitors (Voltage)
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

(An capacitor and inductor are similar in the way that a capacitor resists a change of a voltage and an inductor resists a change in current. The way how 'strong' they can resist depends on their value.)

In GREEN more Mosfets (4) and the NCP4206 Voltage controller 6 Phases

CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

Unseen in the pic are: Memory voltage controller and the monitoring chip INA3221




Now:

ON AIR COOLING (STOCK COOLER, ACX, COOLER WITH LESS DISSIPATION THAN 350W

DON'T GO OVER:

With newer bios revisions: 300W x 120% PT = 360W
With older bios revisions: 330W x 110% PT= 363W


WATERCOOLING:

ANY BIOS REVISION: be extremely careful with anything over 450W!"

Warning: Spoiler! (Click to show)
The PT is the increase of TDP (thermal design power) which is determined by the chips maker (GK110 =250W) but this is not a fixed value, refers to the maximum amount of power the cooling system, in this case a chip, is required to dissipate. The TDP is typically not the most power the chip could ever draw, but the maximum power that it would draw when running "real applications". This ensures the chip will be able to handle essentially all applications without exceeding its thermal envelope, or requiring a cooling system for the maximum theoretical power.
"TDP is meant to be the wattage of the processor at load. I say "wattage" because it is unclear if this is meant to correspond most immediately to how much power is consumed in watts, or how much heat is produced in watts, but as near as I can tell the TDP is pretty much meant to indicate both" GL



(where C is capacitance, f is frequency and V is voltage)

Now, you dont have to make complicated calculus or anything like that because you have this chip here:



It monitors real-time voltage and power draw and its where AB/PrecisionX gets its hardware monitor readings from! wink.gif

Stock bios come with 250W TDP (AKA PT) so when its at 100% you will have 250W of power draw, if you increase it to the max stock 106% youll get: 250x106%=265W
The same is with modded 300/400/500W bios what you see in AB or precisionX is the percentage above what you set!
Ex: with a 500W bios (Slider set to maximum of course) you see 60% usage, this equals to: 500x60%=300W
YOU CAN DIRECTLY CONTROL TDP WITH THE SLIDER! thumb.gif

Now, why has AB a 300% slider while PrecisionX uses 200% for the same bios with the same PT?
Well, AB and precision have different interfaces so the readings are different for the same thing, just keep in mind the base TDP value and make your calculations from there
It doesnt matter what the % slider is in any program, just increase it if you having stutters or frame drops and when making calculations always make them from the base TDP with my formula:

aW x b% = cW (a= bios base TDP, b= OSD TDP, c= aproximate power draw)


Warning: Spoiler! (Click to show)
"...Voltage is just is the electrical potential for a circuit to do work, Current is the flow of electric charge and wattage is the rate at which energy is transferred by an electrical circuit. Typically wattage is measured by multiplying Amperage by Voltage! V * I = W (“I” is the variable for current, or amperage, in electronics and physics. It stands for “Impetus”.) wink.gif
When you set a higher power limit, you are allowing for more amperage to be drawn, but it doesnt mean ITS drawing that amperage as its just a upper limit you set with the slider!
It depends on the load the card has from the software its running! The more harsher it is the more power it draws, voltage only will allow more amperage to flow.
Conductor materials tend to increase their resistivity with an increase in temperature!
The reasons for these changes in resistivity can be explained by considering the flow of current through the material. The flow of current is actually the movement of electrons from one atom to another under the influence of an electric field. Electrons are very small negatively charged particles and will be repelled by a negative electric charge and attracted by a positive electric charge. Therefore if an electric potential is applied across a conductor (positive at one end, negative at the other) electrons will "migrate" from atom to atom towards the positive terminal.
Only some electrons are free to migrate however. Others within each atom are held so tightly to their particular atom that even an electric field will not dislodge them. The current flowing in the material is therefore due to the movement of "free electrons" and the number of free electrons within any material compared with those tightly bound to their atoms is what governs whether a material is a good conductor (many free electrons) or a good insulator (hardly any free electrons).
The effect of heat on the atomic structure of a material is to make the atoms vibrate, and the higher the temperature the more violently the atoms vibrate.
In a conductor, which already has a large number of free electrons flowing through it, the vibration of the atoms causes many collisions between the free electrons and the captive electrons. Each collision uses up some energy from the free electron and is the basic cause of resistance. The more the atoms jostle around in the material the more collisions are caused and hence the greater the resistance to current flow.
So to sum it up, we want lower temperatures which lead to lower electrical resistance, hence having less heat produced as waste and more power to our Titans/780´s cores to OC higher!
HEAT KILLS!
Exactly what happens depends on how excess the power is. It may be a sustained cooking. In this case, the MOSFET gets hot enough to literally unsolder itself. Much of the MOSFET heating at high currents is in the leads - which can quite easily unsolder themselves without the MOSFET failing! If the heat is generated in the chip, then it will get hot - but its maximum temperature is usually not silicon-restricted, but restricted by the fabrication. The silicon chip is bonded to the substrate by soft solder and it is quite easy to melt this and have it ooze between the epoxy and the metal of the body, forming solder droplets! Excess heat leads to short circuit! Usually, a MOSFET will fail short first. This is because excessive heat will, by diffusion, mix the dopants enough to create a good conductor instead of the p-n or n-p barriers that were there originally. Often, the gate oxide will be taken into the diffusion, too, causing a short between all three terminals.
Only if the short circuit current after this first mode of failure is high enough to blow the bond wires or the entire transistor, there is an open circuit.
The lower the temperature the better! My advice is always go water, that way your VRM's are roughly the same temperature as the core, max VRM's operating temperature is 85C (voltage controller, caps etc) on some parts and others( mosfets) 125C (Absolute MAX), so, hitting 80C on the core means that other VRM components are above spec temp and others and below!
Rule of the thumb in semi-conductors is 10C less doubles the elements life, also leakage current increases exponentially (leakage current doubles for every 8 to 10 °C increase in temperature). This is a very good reason to try to keep the operating temperature as low as possible! wink.gif"


Quote:
Originally Posted by tigertank79 View Post

I use afterburner (for best OSD) + EVGA Voltage Tuner. I saved it from old PrecisionX, It's a standalone module smile.gif
PrecisionX 15 support Ti voltage but OSD isn't good as statistics server.

That works too! thumb.gif

Quote:
Originally Posted by vulcan78 View Post

Holy crap guys, I don't want to bash all of the hard work on Skyn3t's behalf but if youre on a reference PCB youre better off sticking with the default vbios that came on the card. I flashed to the Skyn3t's revised SC ACX.rom as I was having stability issues at 1241 core / 1851 memory (+100 / +200) in Max Payne 3 with 3D Vision at 1440p, the additional voltage DID allow me to remain stable at 1254 core but no higher. The temps weren't an issue with that game. As far as stability goes, 1241 core on the default vbios is mostly stable in all other demanding games, and I'm talking Crysis 3, Metro LL, AC4: Black Flag etc. But I've since reflashed to the default vbios as I'm waiting for my Kraken G10's to arrrive (to be used in conjunction with Corsair H55's and the existing ACX mid-plate to address VRM cooling) and not wanting to put undue wear on my GPU's until then. I've revisited a few benchmarks to see what kind of temp difference and WOW I am really surprised.
Firestrike with Skyn3t SC ACX vbios:
1254 core / 1850 memory (yes 1281 core is stable IN FIRESTRIKE with this vbios but the highest I can reliably attain in games at 1.212v is 1254 core)
24,494 GPU
Temps:
GPU0: 73 C
GPU1: 66 C
Unigine Valley:
DX11
Ultra
3DVision
2560x1440
4xAA
Temps by end of bench:
GPU0: 87C
GPU1: 78C
Default SC vbios:
Firestrike
1241 core / 1850 memory
24,098 GPU
Temps:
GPU0: 66C (-7C)
GPU1: 62C (-4C)
Unigine Valley
Temps by end of bench:
GPU0: 74C (-13C)
GPU1: 66C (-12C)
Things were getting so out of hand with the Skyn3t vbios (primary was hitting 85C in AC4: Black Flag at 1440p - 3D Vision, removing the side-panel brought that down to about 80 C) that I felt the need to dial back the temp target from 95 to 75C, and even "linking" the temp and power target until I realized that doing so would reduce the power target to 80% with a temp target of 75C. The temps were still outrageous with the temp target reduced. This vbios is GREAT if youre under water, but even then, without the ability to increase the voltage over 1.212v on cards with reference PCB's, what's the point? I gotta hand it to the engineers and programmers at Nvidia and EVGA, they truly optimized the default vbios', at least as far as the SC ACX variant is concerned. Undoing their expertise feels like a fools errand.
To be clear, these temps were in an Air 540, second only to a test-bench in terms of airflow, with aftermarket fan's pushing significantly more CFM AND an aggressive, and I mean aggressive fan algorithm on the ACX coolers (40% RPM at 30C increasing linearly to 100% at 65C):
https://www.youtube.com/watch?v=XGtlksj-Fdc
I will probably stick with the default vbios when I get my G10's going, as it stands I'm still seeing 75C-78C primary in very demanding games.
An old adage comes to mind: "If it aint broke, don't fix it."


You are forgetting not all cards are the same! Do not advise people to stay away from our bios when only your card cannot handle it, when hundreds of others that use our bios are fine!
Your card doesn't need it because it has a good batch chip or you are probably using the wrong bios for your card; yes, wrong because there are different 80.80.30/34.00.80 out there due to small differences in memory timings and slight PCB changes that are reflected in the bios by those engineers you praise so much, best is to mod your own original bios when you are having problems!
And to be clear i can also change the ACX bios fan algorithms to match users needs (already did for some users in windows and linux environment), maybe i am not as good (clearly not) as those engineers, so maybe you can ask them to mod your bios to make it better!


Cheers

Occamrazor
OccamRazor is offline  
post #13869 of 16563 (permalink) Old 10-25-2014, 04:25 AM
What is old is new
 
OccamRazor's Avatar
 
Join Date: May 2012
Location: Germantown MD
Posts: 4,447
Rep: 631 (Unique: 322)
Quote:
Originally Posted by mongomunken View Post

Thanks, now I got K-Boost working. Seems slightly more stable.
Do you mean 1500mhz is dangerous or something? Should I stop at 1300mhz?
I set offset to +200 (1245MHz) and increase by 20 after running about half of Unigine Valley.
I ended up at a stable +440 (1485MHz). I was amazed as I hadn't been close to this with afterburner. Verified the clock in gpu-z.

Absolutely not! Some cards are amazingly good and can reach very high clock on air, now, your card is exhibiting instability at high clocks while it can reach them, its might be due to excess heat (lower temps might stabilize the OC) or it just needs more voltage to stabilize the OC when a higher load is imposed by the software!
Stable is a very dynamic word... for me a stable OC is when i can play a certain game at some certain clocks for hours or days without a crash, then another game engine might need more voltage for the same clocks... For other people stable is when everything is working without crashes for a long time but at lower clocks!
I prefer to have different OC settings for different game engines, as all cards are different, your mileage my vary!

Then I restart my computer. Suddenly Valley is instantly crashing, and I have to go down to +240 (1285MHz) to be stable.
Quote:
Originally Posted by HighTemplar View Post

Nope none whatsoever.

Think twice about it for reference cards, read my articles here: https://www.overclock.net/t/1438886/official-nvidia-gtx-780-ti-owners-club/13800_100#post_23051292


Cheers

Occamrazor
OccamRazor is offline  
post #13870 of 16563 (permalink) Old 10-25-2014, 04:28 AM
What is old is new
 
OccamRazor's Avatar
 
Join Date: May 2012
Location: Germantown MD
Posts: 4,447
Rep: 631 (Unique: 322)
Quote:
Originally Posted by vulcan78 View Post

I was under the same impression! No improvement in performance (Firestrike) but about 5C higher temps (Unigine Heaven) and artifacts! Now that I think about it, I believe the artifacts were a side-effect of the primary GPU sitting at 92 C in Heaven! Dialing back Power Target from 160% to 106% brought primary temp down to 87C and hence less artifacts. I hate to think about what kind of temps my VRM's were seeing with the core at 92C; I've heard anything higher than say 85-90C on the VRM spells the death knell for GK110. I'm glad it was for only about 10-15 minutes total while I played around with the Power Target settings and did runs with both the side-panel on and off; I imagine playing Skyrim with 100+ mods and an ENB for hours on end with similar temperatures would just about cook your card in a few months time.
Absolutely. I added rather lengthy post-script edit to my post you quoted pertaining to the G10 AIO cooling solution. If youre running a card with a VRM cooling plate you'll probably want to have a look at the G10.

Read my articles:

"You can always choose with the slider what ever power target your card will have: Min 300W - Max 600W, my advice has always been to leave PT at default 300W = 100% and only increase it if you see stutters or frame drops!
As soon as you increase the slider and your card is power hungry, the voltage will allow more current into the card (THE AMOUNT ALWAYS DEPENDING ON THE SLIDER) and usually with stock air coolers (ACX as well) that are not capable of handling more than 350W of continued heat but to a much lesser extent a split second heat spike in the mosfets (VRM's)! on top of that i see lots of people using kombustor, mining, oc scanner, occt etc without knowing exactly how those programs should be run,( If anyone interested in mining with 780Ti, PM Gordan for the safest settings) loading the card to an extreme generating more heat that the cooler can handle!



In RED the Power Mosfets (Actually DrMos Modules for High current DC-DC conversion) for core and Memory (U8-U13 = core, U98/U99 Memory)
(The problem with Mosfets is their tiny size, they generate huge amounts of heat and only have a very small size making it very difficult to dissipate all that heat effectively, if there is a heat spike, even with LN2 they just "blow"...)
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

In Yellow the Inductors (Current) [R22] for the core and [R33] for the memory

In BLUE the Capacitors (Voltage)
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

(An capacitor and inductor are similar in the way that a capacitor resists a change of a voltage and an inductor resists a change in current. The way how 'strong' they can resist depends on their value.)

In GREEN more Mosfets (4) and the NCP4206 Voltage controller 6 Phases

CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90
CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90

Unseen in the pic are: Memory voltage controller and the monitoring chip INA3221




Now:

ON AIR COOLING (STOCK COOLER, ACX, COOLER WITH LESS DISSIPATION THAN 350W

DON'T GO OVER:

With newer bios revisions: 300W x 120% PT = 360W
With older bios revisions: 330W x 110% PT= 363W


WATERCOOLING:

ANY BIOS REVISION: be extremely careful with anything over 450W!"

The PT is the increase of TDP (thermal design power) which is determined by the chips maker (GK110 =250W) but this is not a fixed value, refers to the maximum amount of power the cooling system, in this case a chip, is required to dissipate. The TDP is typically not the most power the chip could ever draw, but the maximum power that it would draw when running "real applications". This ensures the chip will be able to handle essentially all applications without exceeding its thermal envelope, or requiring a cooling system for the maximum theoretical power.
"TDP is meant to be the wattage of the processor at load. I say "wattage" because it is unclear if this is meant to correspond most immediately to how much power is consumed in watts, or how much heat is produced in watts, but as near as I can tell the TDP is pretty much meant to indicate both" GL



(where C is capacitance, f is frequency and V is voltage)

Now, you dont have to make complicated calculus or anything like that because you have this chip here:



It monitors real-time voltage and power draw and its where AB/PrecisionX gets its hardware monitor readings from! wink.gif

Stock bios come with 250W TDP (AKA PT) so when its at 100% you will have 250W of power draw, if you increase it to the max stock 106% youll get: 250x106%=265W
The same is with modded 300/400/500W bios what you see in AB or precisionX is the percentage above what you set!
Ex: with a 500W bios (Slider set to maximum of course) you see 60% usage, this equals to: 500x60%=300W
YOU CAN DIRECTLY CONTROL TDP WITH THE SLIDER! thumb.gif

Now, why has AB a 300% slider while PrecisionX uses 200% for the same bios with the same PT?
Well, AB and precision have different interfaces so the readings are different for the same thing, just keep in mind the base TDP value and make your calculations from there
It doesnt matter what the % slider is in any program, just increase it if you having stutters or frame drops and when making calculations always make them from the base TDP with my formula:

aW x b% = cW (a= bios base TDP, b= OSD TDP, c= aproximate power draw)



"...Voltage is just is the electrical potential for a circuit to do work, Current is the flow of electric charge and wattage is the rate at which energy is transferred by an electrical circuit. Typically wattage is measured by multiplying Amperage by Voltage! V * I = W (“I” is the variable for current, or amperage, in electronics and physics. It stands for “Impetus”.) wink.gif
When you set a higher power limit, you are allowing for more amperage to be drawn, but it doesnt mean ITS drawing that amperage as its just a upper limit you set with the slider!
It depends on the load the card has from the software its running! The more harsher it is the more power it draws, voltage only will allow more amperage to flow.
Conductor materials tend to increase their resistivity with an increase in temperature!
The reasons for these changes in resistivity can be explained by considering the flow of current through the material. The flow of current is actually the movement of electrons from one atom to another under the influence of an electric field. Electrons are very small negatively charged particles and will be repelled by a negative electric charge and attracted by a positive electric charge. Therefore if an electric potential is applied across a conductor (positive at one end, negative at the other) electrons will "migrate" from atom to atom towards the positive terminal.
Only some electrons are free to migrate however. Others within each atom are held so tightly to their particular atom that even an electric field will not dislodge them. The current flowing in the material is therefore due to the movement of "free electrons" and the number of free electrons within any material compared with those tightly bound to their atoms is what governs whether a material is a good conductor (many free electrons) or a good insulator (hardly any free electrons).
The effect of heat on the atomic structure of a material is to make the atoms vibrate, and the higher the temperature the more violently the atoms vibrate.
In a conductor, which already has a large number of free electrons flowing through it, the vibration of the atoms causes many collisions between the free electrons and the captive electrons. Each collision uses up some energy from the free electron and is the basic cause of resistance. The more the atoms jostle around in the material the more collisions are caused and hence the greater the resistance to current flow.
So to sum it up, we want lower temperatures which lead to lower electrical resistance, hence having less heat produced as waste and more power to our Titans/780´s cores to OC higher!
HEAT KILLS!
Exactly what happens depends on how excess the power is. It may be a sustained cooking. In this case, the MOSFET gets hot enough to literally unsolder itself. Much of the MOSFET heating at high currents is in the leads - which can quite easily unsolder themselves without the MOSFET failing! If the heat is generated in the chip, then it will get hot - but its maximum temperature is usually not silicon-restricted, but restricted by the fabrication. The silicon chip is bonded to the substrate by soft solder and it is quite easy to melt this and have it ooze between the epoxy and the metal of the body, forming solder droplets! Excess heat leads to short circuit! Usually, a MOSFET will fail short first. This is because excessive heat will, by diffusion, mix the dopants enough to create a good conductor instead of the p-n or n-p barriers that were there originally. Often, the gate oxide will be taken into the diffusion, too, causing a short between all three terminals.
Only if the short circuit current after this first mode of failure is high enough to blow the bond wires or the entire transistor, there is an open circuit.
The lower the temperature the better! My advice is always go water, that way your VRM's are roughly the same temperature as the core, max VRM's operating temperature is 85C (voltage controller, caps etc) on some parts and others( mosfets) 125C (Absolute MAX), so, hitting 80C on the core means that other VRM components are above spec temp and others and below!
Rule of the thumb in semi-conductors is 10C less doubles the elements life, also leakage current increases exponentially (leakage current doubles for every 8 to 10 °C increase in temperature). This is a very good reason to try to keep the operating temperature as low as possible! wink.gif"


Cheers


Occamrazor
OccamRazor is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off