Jump to content


Photo
- - - - -

Bios flashing for Kepler cards


  • Please log in to reply
16 replies to this topic

#1 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 29 March 2013 - 10:01 PM

Right, so I've spent the day overclocking my gtx 680 in Precision X and I've found the maximum offsets. Now I'm ready to permanentize it with a bios flash. Except bios flashing for the kepler chip seems iffy at this point as there are only two applications that can modify it and they are both in early stages of development. One is the kgb bios editor and the other one is the KeplerBiosTweaker. 

 

The actual flashing process isn't an issue. I'm mainly looking for a reliable way to modify my bios with the overclocked offsets, without anything else being affected.

 

For example, I don't know how modifying the core/mem clocks will reflect on the dynamic overclock that these cards have as well as the boost clock. Considering that in Precision X everything scales well I'd rather not modify anything except the base clocks since my max core/mem clocks are based on those standard boosts.

 

Another question I have is weather it's okay for my card's power percent to not be reaching 100% or even the 111% that's achievable in Precision X. I only got to like 90% in my final stress tests and I'm wondering if that is causing the boost clocks to throttle. Or is it that if it hasn't reached 100% it means that there's just no need for it and I'm actually getting my max overclock anyway.

 

FYI final overclock I got was 1280/1710 @ ~58C on a Gigabyte Windforce 2GB. I feel like if the card drew more power I could get higher speeds.


Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED


#2 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 06:41 AM

What's wrong with just applying them on boot? Permanent balls-to-the-wall clocks are usually a bad idea.

sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#3 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 30 March 2013 - 06:15 PM

Need to set flashing aside for a moment. 

 

The 680/titan cards seem to have issues with the TDP as I suspected yesterday. They can't reach 100% or higher TDP which is causing stability issues and obviously not being able to achieve nicer overclocks. 

 

For those who are hearing this for the first time, here's a quote from an OCN member describing the case nicely in a few words: 

 

It has nothing to do with overclocking.


The card doesn't seem to be able to maintain its top boost clock at STOCK even when it is under the 100% Power and 80C limits. This behavior is different than what was described by Nvidia which is why people are upset. I have 2x MSI cards set to ship tomorrow and while I still plan on running SLI I am not thrilled about this throttling situation.

 

Anyone have some interesting thoughts on the matter?


Edited by sack_patrol, 30 March 2013 - 06:17 PM.

Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED


#4 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 06:39 PM

Thoughts? It's a meaningless number that people are obsessing about. You can't hit "100% power usage" with most graphics workloads. Even most GPGPU workloads can't. :tongue:

sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#5 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 06:39 PM

Thoughts? It's a meaningless number that people are obsessing about. You can't hit "100% power usage" with most graphics workloads. Even most GPGPU workloads can't. :tongue:

sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#6 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 30 March 2013 - 06:41 PM

So what's the point of being able to increase the power target to beyond 100%? If cards can't get to 100% it would seem like a completely useless slider in afterburner and precision.


Edited by sack_patrol, 30 March 2013 - 06:42 PM.

Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED


#7 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 06:45 PM

Because when you do need that little extra bit the card will go over the limit.

sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#8 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 30 March 2013 - 06:47 PM

But, it never does. Under no circumstance does it ever go over the limit. Not even close.


Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED


#9 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 06:59 PM

Under the ones you've tested. I guarantee I could make a workload that'd do it. :lol:

sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#10 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 30 March 2013 - 07:09 PM

Well, I wouldn't mind seeing that. 

 

I think that it should be standard for the cards to at least reach 100% when under full load from a benchmark especially if there's an overclock involved. Not to mention the overclocking software supports increasing that power limit to over 100%. I can't imagine it being a non functional slider that serves no purpose at all.


Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED


#11 Waco

Waco

    Lab Rat 2

  • Reviewer
  • PipPipPipPipPipPipPipPip
  • 15404 posts
  • Gender:Male
  • Location:Los Alamos, NM

Posted 30 March 2013 - 08:27 PM

It's not nonfunctional.  How often do you really thing you max out your CPU? :lol:


sig.png

GTX 770 replaced with watercooled GTX780...but too lazy to change the sig.


#12 sack_patrol

sack_patrol

    cozza frenzy

  • Members
  • PipPipPipPipPip
  • 2184 posts
  • Gender:Male

Posted 30 March 2013 - 08:38 PM

When I run stress tests that push it to the limit. Just as I have done with my graphics card. Am I missing something? 

 

Many people are reporting throttling issues with their titan cards which are downclocking or just oscillating for no reason at all. This isn't the issue for me as 680 doesn't use the same gpu boost features but I'm still not understanding why the overclocking software are including a power target slider that based on your statements seems to be useless as regular benchmarks and games will not cause the card to draw 100% tdp, but it takes some sort of secret technique that nobody knows about to actually reach said power target.


Edited by sack_patrol, 30 March 2013 - 08:46 PM.

Timestretch2.png

#1: DFI 790FXB-M3H5 | Phenom II 965 BE @ 3.6Ghz | Sunbeam CCF 4 | Vapor-X 5870 1GB 940/1290 | Ballistix Tracer 2x2GB 1333Mhz 7-7-7-24 | Razer Barracuda AC-1

#2: MSI Z77-GD65 | Intel Core i5 3570k @ 4.4Ghz | Noctua NH-D14 | WindForce GTX 680 2GB 1306/1688 | Ballistix Tracer 2x4GB 1600Mhz 8-8-8-24 | Xonar Essence STX

Chassis: Antec 902 | Corsair Force GT 90GB | Seagate 1TB | WD Caviar Blue 320GB | OCZ ModXStream Pro 600W

Peripherals: Logitech G400 | Rosewill Cherry MX Red | audio-technica ATH-AD700 | Asus vh238h 23' LED