GIGABYTE GTX 9xx H2O/AIR BIOS Tweaking ?(ô?ô)? - Page 97 - Overclock.net - An Overclocking Community

Forum Jump: 

GIGABYTE GTX 9xx H2O/AIR BIOS Tweaking ?(ô?ô)?

Reply
 
Thread Tools
post #961 of 7857 (permalink) Old 07-01-2015, 02:36 PM
New to Overclock.net
 
Edkiefer's Avatar
 
Join Date: Feb 2013
Posts: 1,441
Rep: 115 (Unique: 62)
Quote:
Originally Posted by Fantasy View Post

my 970 idles at 1037/4059 @1018v because I have three monitors connected to it. If I only connect one or two monitors my idle speed will drop to 135/300 @ 0.856v.

Are you sure setting the card at 1.325v will be over 1.4v? I have no way of measuring my card actual voltage. I could by a multimeter but I have no idea how to use it and I have no idea how to measure my card voltage with it.

I can only speak for my sample, but I know others who have measured it also got way higher than what GPU-Z reports .
What GPU-Z and other monitors report, is the target voltage drivers says, not actual voltage .

Test run with Heven ultra setting, 1024res, 4x AA , this gives high fps and load .

stock voltage, GPU-Z 1.218 , DVM 1.297-1.317 (start vary a bit more )
voltage +21 = GPU-Z = 1.243 ,DVM=1.337-1.34v

voltage +41 = GPU-Z = 1.262 ,DVM=1.36v

So all these are with stock voltage table , you can easy see if I raised more it would go over 1.4v mark .
You can see a +offset of 0.082-0.095v over what driver reported .

I personally wouldn't go over 1.275v, thats just me, I don't find performance scaling well once you go over 1450-1500+ , that range seems to be sweet spot .
Now this is 970, your results may vary .

Edkiefer is offline  
Sponsored Links
Advertisement
 
post #962 of 7857 (permalink) Old 07-01-2015, 03:52 PM - Thread Starter
Maximum_Unleashed
 
Laithan's Avatar
 
Join Date: Mar 2015
Location: United States
Posts: 3,712
Rep: 500 (Unique: 265)
Quote:
Originally Posted by Edkiefer View Post

I can only speak for my sample, but I know others who have measured it also got way higher than what GPU-Z reports .
What GPU-Z and other monitors report, is the target voltage drivers says, not actual voltage .

Test run with Heven ultra setting, 1024res, 4x AA , this gives high fps and load .

stock voltage, GPU-Z 1.218 , DVM 1.297-1.317 (start vary a bit more )
voltage +21 = GPU-Z = 1.243 ,DVM=1.337-1.34v

voltage +41 = GPU-Z = 1.262 ,DVM=1.36v

So all these are with stock voltage table , you can easy see if I raised more it would go over 1.4v mark .
You can see a +offset of 0.082-0.095v over what driver reported .

I personally wouldn't go over 1.275v, thats just me, I don't find performance scaling well once you go over 1450-1500+ , that range seems to be sweet spot .
Now this is 970, your results may vary .

I asked about this a while back and I think what it came down to was where the reading was taken. The extreme hard modders talk about having to remove inline resistors to increase voltages which doesn't prove anything other than it is possible to take the reading before, or after the resistors.

I think one thing is for sure, that if you are truly pushing 1.4v, SO IS EVERYONE ELSE smile.gif and after what I've done to my 980's during the process of helping develop the Max Unleashed BIOS' - They aren't dead yet!

All kidding aside, we've all got voltage protection on our cards, we practically CAN'T fry them with too much voltage unless we somehow disable overvolt protection. I've never even heard of someone frying any Maxwell to be honest. One of the key reasons is due to the power efficiency. Older GPUs needed more voltage and got MUCH hotter. These Maxwell GPUs are sipping less than one and a half volts FOR ALL THESE CORES and compute performance! My GPUs need LESS voltage to overclock than my CPU does and arguably outperforms it smile.gif.

This wasn't always the case though, GTX 590's were getting fried back in the day when BIOS mods got in the hands of the dangerous and overbrave rolleyes.gif

thumb.gif

PS. I'm throwing out my prediction now.. way ahead of anything.. O/S' will be ported to CUDA, x86 will become outdated (we're there now), Intel will become Cyrix and Motherboards will become giant GPU PCBs thumb.gif

Wow that's a mouthful tongue.gif

GIGABYTE GTX 9xx G1 GAMING BIOS Tweaking
  ̿̿ ̿̿ ̿̿ ̿̿̿'̿'\̵͇̿̿\з=༼ຈل͜ຈ༽=ε/̵͇̿̿/'̿'̿ ̿ ̿̿ ̿̿ ̿̿  
░▒▓│ FORKAY‼ «  » WΘΘT ‼ │▓▒░
Xeon E5-1680 V2 (IVY-E) Inside


Laithan is offline  
post #963 of 7857 (permalink) Old 07-01-2015, 05:04 PM
New to Overclock.net
 
Edkiefer's Avatar
 
Join Date: Feb 2013
Posts: 1,441
Rep: 115 (Unique: 62)
It could very well be the GPU gets slightly lower voltage than what comes out of the regulators/chokes .

I can not easy probe much closer to GPU w/o disassemble system on testbed ,even then it is risky IMO .

Also, the voltage is not flat steady like you see in GPU-Z , it mimic's more what you see with Vcore on CPU .

My main point is/was if you can get 1500+ on stock, does it pay to run with 100+mv more for minor speed bump .

Again each chip is unique so YMWV .

Edkiefer is offline  
Sponsored Links
Advertisement
 
post #964 of 7857 (permalink) Old 07-01-2015, 07:17 PM
New to Overclock.net
 
Oggaaaa's Avatar
 
Join Date: Nov 2013
Posts: 35
Rep: 3 (Unique: 2)
Guys. Im thinking about flashing my 980Ti G1. I can get a somewhat stable boostclock of 1479 (crashes in chrome and witcher 3 sometimes shuts down), but i want to reach 1500++ as alot of cards seems to reach.
I dont know if i suck at overclocking or if i should be able to reach higher with stock bios.

Anyways my card is also behaving rather strange. If i lets say overclock my card +10mhz on the coreclock and then max out the voltage +87mv it will crash in just a minute. But if dont apply any voltage it will be fine. It seems like i have to have the "perfect" voltage for a certain coreclock speed. Right now im running +45 on the voltage and +100 on the core, but if i run the voltage on +87 it crashes, despite the volt staying the same on 1218mv...

One more question. Is the bios on the first page "980F3DP-MAX-UNLEASHED-REL1.0.zip" for the 980ti? Cuz it seems like its for the 980 and i cant find one for the 980ti.
Oggaaaa is offline  
post #965 of 7857 (permalink) Old 07-01-2015, 07:32 PM - Thread Starter
Maximum_Unleashed
 
Laithan's Avatar
 
Join Date: Mar 2015
Location: United States
Posts: 3,712
Rep: 500 (Unique: 265)
Ok so while we are all looking around at each other with our heads still but eyes moving from side to side, all saying "Not it" wondering who is going hack/update MBT, I had some more testing to do... just because I can and apparently I have no life tongue.gif

So check this out...

* Disclaimer: The following results are mine. There are others like them but these ones are mine. tongue.gif I can't say that this will be the case for everyone.

Back to some 980 love wubsmiley.gif I did some comparison tests between Max Unleashed 980 1.0 vs. Max Unleashed Special version 1312 1.0

Max Unleashed 980 1.0
http://www.3dmark.com/fs/5275486
MaxUnleashed9801.0Firestrike.txt 437k .txt file



Max Unleashed Special version 1312 1.0
http://www.3dmark.com/fs/5275308
MaxUnleashed98013121.0Firestrike.txt 93k .txt file



Umm... more than margin of error I think.. Same clocks on both, only difference was the max voltage in the BIOS.


----

How do these scores compare to a 980Ti?

Max Unleashed 980 1.0 Firestrike EXTREME
http://www.3dmark.com/fs/5275428
MaxUnleashed9801.0FirestrikeExtreme.txt 96k .txt file



Max Unleashed 980 1.0 Firestrike ULTRA
http://www.3dmark.com/fs/5275455
MaxUnleashed9801.0FirestrikeUltra.txt 76k .txt file
Attached Files
File Type: txt MaxUnleashed9801.0Firestrike.txt (437.2 KB, 8 views)
File Type: txt MaxUnleashed98013121.0Firestrike.txt (92.6 KB, 2 views)
File Type: txt MaxUnleashed9801.0FirestrikeExtreme.txt (96.3 KB, 5 views)
File Type: txt MaxUnleashed9801.0FirestrikeUltra.txt (76.3 KB, 2 views)

GIGABYTE GTX 9xx G1 GAMING BIOS Tweaking
  ̿̿ ̿̿ ̿̿ ̿̿̿'̿'\̵͇̿̿\з=༼ຈل͜ຈ༽=ε/̵͇̿̿/'̿'̿ ̿ ̿̿ ̿̿ ̿̿  
░▒▓│ FORKAY‼ «  » WΘΘT ‼ │▓▒░
Xeon E5-1680 V2 (IVY-E) Inside


Laithan is offline  
post #966 of 7857 (permalink) Old 07-01-2015, 07:39 PM - Thread Starter
Maximum_Unleashed
 
Laithan's Avatar
 
Join Date: Mar 2015
Location: United States
Posts: 3,712
Rep: 500 (Unique: 265)
↑ Clean runs ( ͡° ͜ʖ ͡°)

Quote:
Originally Posted by Oggaaaa View Post

Guys. Im thinking about flashing my 980Ti G1. I can get a somewhat stable boostclock of 1479 (crashes in chrome and witcher 3 sometimes shuts down), but i want to reach 1500++ as alot of cards seems to reach.
I dont know if i suck at overclocking or if i should be able to reach higher with stock bios.

Anyways my card is also behaving rather strange. If i lets say overclock my card +10mhz on the coreclock and then max out the voltage +87mv it will crash in just a minute. But if dont apply any voltage it will be fine. It seems like i have to have the "perfect" voltage for a certain coreclock speed. Right now im running +45 on the voltage and +100 on the core, but if i run the voltage on +87 it crashes, despite the volt staying the same on 1218mv...

One more question. Is the bios on the first page "980F3DP-MAX-UNLEASHED-REL1.0.zip" for the 980ti? Cuz it seems like its for the 980 and i cant find one for the 980ti.

Hi there.

No that is for a 980 ONLY

BUT

Check out this post https://www.overclock.net/t/1544574/gigabyte-gtx-9xx-g1-gaming-h2o-air-bios-tweaking/920#post_24111927
thumb.gif

Are you saying that with your stock BIOS, raising your default core speed by +10Mhz and maxxing out the voltage slider is causing a crash? Can you please elaborate as much as possible as to what all your settings are, how you re-produce a crash and if you can include a GPU-z log of when the problem occurs, that would be awesome. biggrin.gif

We'll see what we can do to help. smile.gif

GIGABYTE GTX 9xx G1 GAMING BIOS Tweaking
  ̿̿ ̿̿ ̿̿ ̿̿̿'̿'\̵͇̿̿\з=༼ຈل͜ຈ༽=ε/̵͇̿̿/'̿'̿ ̿ ̿̿ ̿̿ ̿̿  
░▒▓│ FORKAY‼ «  » WΘΘT ‼ │▓▒░
Xeon E5-1680 V2 (IVY-E) Inside


Laithan is offline  
post #967 of 7857 (permalink) Old 07-01-2015, 11:04 PM
New to Overclock.net
 
Join Date: May 2015
Posts: 120
Rep: 42 (Unique: 32)
I have analyzed the file structure of the roms a bit.

I found out, that all voltages are defined in 34-byte long blocks. First 4 bytes are an integer value, defining min-voltage. This value is in milivolts times 1000. Which would make them microvolts, I suppose. Next 4 bytes are the max voltage for that definition. The rest of the values, I have no idea about. The block is not 4-byte aligned, therefore some of the values have to be something other than 4-byte integers (obviously, 4-byte alignment does not imply, that all the values would be 4-byte integers. Could be 2x 2-byte integers instead of one 4-byte integer). Also, it appears, that all the voltage definitions are right next to each other, which suggests, that they are ordered into a typed array.

I have compared the MaxUnleashed bios from the OP (for the 980), and my own modded bios of the 980 ti (MSI 6G). It would appear, that they are largely identical in terms of voltage definition structure, though values vary, obviously. Also, they start at different offsets in the file.

Here is a spreadsheet analyzing the voltage blocks up to the clock voltage table

https://docs.google.com/spreadsheets/d/1fRJoQTnzutjLt6LsJq2ytyJ8g_b4Laf5Um1X89i_Owo/pubhtml

Here is a 2010 macro-enabled Excel

ReverseEngineering.zip 116k .zip file

From what I can gather, the purpose of the voltage blocks, in relation to their position in the array, is identical. Theoretically, since the MBT does not allow tweaking the two sliders for external voltage adjustment and temp throttle, you could just replace the values directly in the file, with a hex editor like HxD. The thing about voltage values, they seem to be multiples of some number. Also, the value you see in MBT is kinda rounded up, so in the file, there are values like 1281,25 instead of 1281.3. Oh, and of course, the file has a checksum in it, which I have no idea how is calculated. It is only 1 byte long though. Maybe just sum across all bytes mod 255?

What baffles me, is why the MBT does not display those 2 specific sliders. I'm not sure if it is possible, but since the voltage array is on a different offset, the MBT could just be rewriting random pieces of the file, and we just kinda got lucky that it did not corrupt it entirely. This would kind of explain, why messing around with the voltages and max voltage slider has no effect, and maybe why moving the p05 slider actually has an effect, following the broken clock principle.

Thoughts?

EDIT: I just noticed google excel likes to be funny, and put spaces instead of decimal separators. So there is that. 797 435 actually means 797,435

Also, just checked, the MBT does edit the "correct" position in the 980TI file, that is supposed to correspond to the max voltage slider (which I edited in MBT), and does not overwrite random chunks of file. It also overwrites the "correct" position of the p00 slider. Which again begs the question, why does MBT not display the 2 sliders, since it obviously has the correct offsets? Is there a dependency in the voltage blocks, that is unclear, and as a safety measure, it refrains from editing them?

EDIT2: Ok, so this happened. I rewrote the value for external regulation and temp throttle in my GM200 bios to that of the GM204-MaxUnleashed. Obviously, when I opened the file in MBT, the green filed with checksum was red, and a difference between calculated checksum and checksum in file was shown. However, now an additional slider appeard in the voltage table. It seems to be the temp throttle slider. MBT can now rewrite the correct position in the file, and as an added bonus, it calculates the correct checksum when it saves. Great success! Here is my bios

These are demo BIOSes for MSI 6G! Do not flash on G1! They have different offsets!

Base.zip 152k .zip file

EDIT3: Even greater success! I changed both min and max values of the external adjustment entry to match that of the MaxUnleashed, and now it appears as well! Here is that bios (Still MSI 6G):

Base2.zip 152k .zip file

The root of all evil rests with the base values for those sliders. In stock GM200 bios, those are 1150,675 - 1198,47 for the external regulation and 1190,349 - 1239,838 for the temp throttle. It would appear, that along with other "unkown" voltage table entries, the MBT discards values, that is doesn't consider "neat" (multiples of 10, I would assume). Therefore, changing the values to multiples 100 (what gm204 bioses have) caused them to appear in the GUI. I am assuming, this would work for the other unknown values as well, though no reason to poke that bear.

TL;DR:
Here is the 980Ti_F3DP interim pre-BETA1 BIOS (for G1, posted earlier), now with 20% more voltage sliders!
G1_Unlocked.zip 152k .zip file

It would warrant verifying, if the only values replaced by the MBT, are the first 2 4-byte integers on start of each voltage entry, That is, if the rest of the values in the voltage entry are somehow dependent on these first two values! Here is where the source code would come in handy...
Also, buyer beware: Maybe there is a reason the values are not neat? I am not an expert in inner workings of the maxwell architecture or GM200-350. Flashing a corrupted bios has serious risk of rendering the card unusable! This would be alleviated a bit with dual-bios cards, like the EVGA classified...
Attached Files
File Type: zip ReverseEngineering.zip (115.6 KB, 58 views)
File Type: zip Base.zip (152.2 KB, 42 views)
File Type: zip Base2.zip (152.2 KB, 94 views)
File Type: zip G1_Unlocked.zip (151.8 KB, 62 views)
WerePug is offline  
post #968 of 7857 (permalink) Old 07-02-2015, 01:29 AM
New to Overclock.net
 
xonare's Avatar
 
Join Date: Mar 2013
Location: Warsaw, Poland
Posts: 270
Rep: 17 (Unique: 11)
Quote:
Originally Posted by WerePug View Post

EDIT3: Even greater success! I changed both min and max values of the external adjustment slider to match that of the MaxUnleashed, and now it appears as well! Here is that bios:

Base2.zip 152k .zip file

The root of all evil rests with the base values for those sliders. In stock GM200 bios, those are 1150,675 - 1198,47 for the external regulation and 1190,349 - 1239,838 for the temp throttle. It would appear, that along with other "unkown" voltage table entries, the MBT discards values, that is doesn't consider "neat" (multiples of 100, I would assume). Therefore, changing the values to multiples 100 (what gm204 bioses have) caused them to appear in the GUI. I am assuming, this would work for the other unknown values as well, though no reason to poke that bear.

And so here is the 980Ti_F3DP interim pre-BETA1 (for G1), now with 20% more voltage sliders!
G1_Unlocked.zip 152k .zip file
Very interesting did you try to flash with this modded file?
Quote:
Originally Posted by Oggaaaa View Post

Guys. Im thinking about flashing my 980Ti G1. I can get a somewhat stable boostclock of 1479 (crashes in chrome and witcher 3 sometimes shuts down), but i want to reach 1500++ as alot of cards seems to reach.
I dont know if i suck at overclocking or if i should be able to reach higher with stock bios.

Anyways my card is also behaving rather strange. If i lets say overclock my card +10mhz on the coreclock and then max out the voltage +87mv it will crash in just a minute. But if dont apply any voltage it will be fine. It seems like i have to have the "perfect" voltage for a certain coreclock speed. Right now im running +45 on the voltage and +100 on the core, but if i run the voltage on +87 it crashes, despite the volt staying the same on 1218mv...

One more question. Is the bios on the first page "980F3DP-MAX-UNLEASHED-REL1.0.zip" for the 980ti? Cuz it seems like its for the 980 and i cant find one for the 980ti.
Have you tried to find max OC without touching voltage slider? In my case it's 1483Mhz on core.
If you have G1 gaming default load voltage should be 1.180V and with slider maxed out 1.224V so getting 1.218V is strange smile.gif
Applying overvoltage gives you different boost value that is based on boost table entries. Generally I noticed overclocking it works better when you aim at +13Mhz ranges which are next boost clocks in table. For example I get 1354Mhz with no OC in boost no overvoltage and when I apply overvoltage I get 1379Mhz (ommiting decimal values) which is around +26Mhz meaning shift in boost table of 2 positions, after that when I add in Precision X some +Mhz to core I generally target other entries from boost table i.e. 1483Mhz, 1497Mhz, 1506Mhz, 1516Mhz, 1529Mhz (it is not always +13Mhz, check boost table in MBT for exact values smile.gif )

On the voltage side, there are also "set values" you can get like 1.180V, (didn't check values between), 1.224V, 1.237V, 1.250V, (it's possible here is also 1 more value), 1.274V

The Precision X allows smaller differences than values I wrote about, maybe these smaller increments are not good for card thus crashes?
xonare is offline  
post #969 of 7857 (permalink) Old 07-02-2015, 01:31 AM
New to Overclock.net
 
xonare's Avatar
 
Join Date: Mar 2013
Location: Warsaw, Poland
Posts: 270
Rep: 17 (Unique: 11)
Quote:
Originally Posted by Laithan View Post

How do these scores compare to a 980Ti?
14.5k FireStrike on my PCIE 2.0 and old i5-2500K smile.gif
People with PCIE 3.0 and new i7 get around 18k in Firestrike.
Quote:
Originally Posted by Laithan View Post

PS. I'm throwing out my prediction now.. way ahead of anything.. O/S' will be ported to CUDA, x86 will become outdated (we're there now), Intel will become Cyrix and Motherboards will become giant GPU PCBs thumb.gif

Wow that's a mouthful tongue.gif
I would love that biggrin.gif and expect the same, look at Tegra X1 core from Nvidia smile.gif
xonare is offline  
post #970 of 7857 (permalink) Old 07-02-2015, 01:42 AM
New to Overclock.net
 
Join Date: May 2015
Posts: 120
Rep: 42 (Unique: 32)
Quote:
Originally Posted by xonare View Post

Very interesting did you try to flash with this modded file?

Not yet, I am at work and I can't flash anything through RDP.
WerePug 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