- Joined
- May 8, 2016
- Messages
- 1,741 (0.60/day)
System Name | BOX |
---|---|
Processor | Core i7 6950X @ 4,26GHz (1,28V) |
Motherboard | X99 SOC Champion (BIOS F23c + bifurcation mod) |
Cooling | Thermalright Venomous-X + 2x Delta 38mm PWM (Push-Pull) |
Memory | Patriot Viper Steel 4000MHz CL16 4x8GB (@3240MHz CL12.12.12.24 CR2T @ 1,48V) |
Video Card(s) | Titan V (~1650MHz @ 0.77V, HBM2 1GHz, Forced P2 state [OFF]) |
Storage | WD SN850X 2TB + Samsung EVO 2TB (SATA) + Seagate Exos X20 20TB (4Kn mode) |
Display(s) | LG 27GP950-B |
Case | Fractal Design Meshify 2 XL |
Audio Device(s) | Motu M4 (audio interface) + ATH-A900Z + Behringer C-1 |
Power Supply | Seasonic X-760 (760W) |
Mouse | Logitech RX-250 |
Keyboard | HP KB-9970 |
Software | Windows 10 Pro x64 |
Hello everyone.
I hope this topic helps users increase longevity of those chips, while maximising their performance.
Informations contained here are mainly for anyone that wants to use GK110(b) class GPUs on air or for users of such cards that have throttling issues (it's also highly recommended if card uses reference "NVTTM" cooling).
Long story short : I got my Titan Black this friday, and I wanted to do some testing on it.
But after first test ended, I saw "1,187V" as default GPU voltage in GPU-z...
^"This can't be right..." - I thought,
Since my GTX 780 Ti Reference had only 1,0325V as stock (granted, for ~1020MHz clock - but still too high jump).
Why it's bad to both You (user) and your card to have high stock GPU voltage :
1) GPU Boost 2.0 will throttle card A LOT harder on higher voltage (which means max. boost speed isn't possible for any meaningfull lenght of time), because it works within Power/TDP and temp limits.
2) Higher VRM load (in my case, 1,187V is just two notches below NVidia max. allowed limit of 1,212V), and everything that it means. Here's video on blowing up VRAM VRM from buildzoid : LINK.
Because of that, I decided to so something with this situation using KeplerBiosTweaker.
Max. boost frequency for my card was 1058MHz (seen for first few seconds of test... but it was there), keeping in mind my Titan Black uses stock NVTTM cooler I knew that frequency may be beyond reach (since stability of it depends on die quality and how it scales with voltage), but regadless I gave it a try using software only method first (MSI Afterburner). After few tests on lower and lower voltage, I finished with 1V and that was fine for 1058MHz speed I wanted to set (can't be greedy here, maxing frequency/Voltage is silicon quality and cooling limited).
I opted for skyn3t's Titan Black vBIOS, which has power limit unlocked/GPU boost off, instead of stock one (preference).
Since skyn3t's has 1,212V max. voltage, I had to increase max. table frequency under "Boost table" tab, which in turn would decrease default GPU voltages used for lower frequencies (basic Boost Table to Voltage Table comparison should tell you where you SHOULD end up Clock/VDDC wise).
Going from default "1202MHz" Max. Table Clock (boost table) to "1359MHz" in MY CASE meant I got my 1V GPU voltage on 1058,5MHz frequency.
After I set 1058,5MHz as both base and max. boost frequency for my card, I checked all settings second time and flashed my card...
Result :
Pretty good
(grey bar in "PerfCap reason" means utilisation)
But then I thought : What if I decreased GPU Voltage some more ?
Well, only one way to find out...
Checking voltage table and boost tables again, I set "0,987V" after tweaking Max. Table Clock again, and ended up here :
Of course Fan profile remained the same for both tests.
Conclusion (probably not REALLY accurate, but "good enough") :
+0,0125V = +6W in power (according on GPU-z), and that's on ~1V level, where leakage isn't that big of a deal yet.
However I decided to not use this setting, since 0,975V wasn't stable at all in 3D tests (and it's always better to have some headroom for stability "just in case").
Now, why I decided to created this topic :
Since many first users of highest end cards liked to "maxed" them out on water/air, some users that buy those cards used may end up with cards that have relative high stock GPU voltage (like me).
For anyone that doesn't have skills required to do this, you don't have worry... just yet.
vBIOS should automaticly downclock (and downvolt), overheating card the moment it reaches temnperature threshold (assuming it's not set too high). Because of this, your card won't die fast... in theory.
I hope this topic helps users increase longevity of those chips, while maximising their performance.
Informations contained here are mainly for anyone that wants to use GK110(b) class GPUs on air or for users of such cards that have throttling issues (it's also highly recommended if card uses reference "NVTTM" cooling).
Long story short : I got my Titan Black this friday, and I wanted to do some testing on it.
But after first test ended, I saw "1,187V" as default GPU voltage in GPU-z...
Since my GTX 780 Ti Reference had only 1,0325V as stock (granted, for ~1020MHz clock - but still too high jump).
Why it's bad to both You (user) and your card to have high stock GPU voltage :
1) GPU Boost 2.0 will throttle card A LOT harder on higher voltage (which means max. boost speed isn't possible for any meaningfull lenght of time), because it works within Power/TDP and temp limits.
2) Higher VRM load (in my case, 1,187V is just two notches below NVidia max. allowed limit of 1,212V), and everything that it means. Here's video on blowing up VRAM VRM from buildzoid : LINK.
Because of that, I decided to so something with this situation using KeplerBiosTweaker.
Max. boost frequency for my card was 1058MHz (seen for first few seconds of test... but it was there), keeping in mind my Titan Black uses stock NVTTM cooler I knew that frequency may be beyond reach (since stability of it depends on die quality and how it scales with voltage), but regadless I gave it a try using software only method first (MSI Afterburner). After few tests on lower and lower voltage, I finished with 1V and that was fine for 1058MHz speed I wanted to set (can't be greedy here, maxing frequency/Voltage is silicon quality and cooling limited).
I opted for skyn3t's Titan Black vBIOS, which has power limit unlocked/GPU boost off, instead of stock one (preference).
Since skyn3t's has 1,212V max. voltage, I had to increase max. table frequency under "Boost table" tab, which in turn would decrease default GPU voltages used for lower frequencies (basic Boost Table to Voltage Table comparison should tell you where you SHOULD end up Clock/VDDC wise).
Going from default "1202MHz" Max. Table Clock (boost table) to "1359MHz" in MY CASE meant I got my 1V GPU voltage on 1058,5MHz frequency.
After I set 1058,5MHz as both base and max. boost frequency for my card, I checked all settings second time and flashed my card...
Result :
Pretty good
(grey bar in "PerfCap reason" means utilisation)
But then I thought : What if I decreased GPU Voltage some more ?
Well, only one way to find out...
Checking voltage table and boost tables again, I set "0,987V" after tweaking Max. Table Clock again, and ended up here :
Of course Fan profile remained the same for both tests.
Conclusion (probably not REALLY accurate, but "good enough") :
+0,0125V = +6W in power (according on GPU-z), and that's on ~1V level, where leakage isn't that big of a deal yet.
However I decided to not use this setting, since 0,975V wasn't stable at all in 3D tests (and it's always better to have some headroom for stability "just in case").
Now, why I decided to created this topic :
Since many first users of highest end cards liked to "maxed" them out on water/air, some users that buy those cards used may end up with cards that have relative high stock GPU voltage (like me).
For anyone that doesn't have skills required to do this, you don't have worry... just yet.
vBIOS should automaticly downclock (and downvolt), overheating card the moment it reaches temnperature threshold (assuming it's not set too high). Because of this, your card won't die fast... in theory.
Last edited: