So how does one OC a 680GTX?

[notty22]

@Ferzerp

I'm just saying that this system isn't working as intended or it's a huge turd for overclockers because it's based on arbitrary factory set caps and don't squeeze all the card's potential.

I don't think that this system is pushing the card more than traditional overclocking would do.

For stock and mild overclock it's great (so it's for me) but for overclockers it's just frustrating and no fun at all.

So from your armchair you are calling the fastest single gpu card ever made a turd. A card/package that every reviewer gave a gold/Editors award to ? Ok.:whiste:

 

[Imouto]

Stop your green fanboyism for a second and read my post again.

 

[Grooveriding]

I did some testing, and it seems that it works like this:

If you leave the voltage alone, it will adjust itself between 0.987 and 1.175
You can adjust the voltage, and it will set that voltage, up to 1.175V

What I didn't have time to check, is if your set voltages are the floor, or the absolute voltage. That is, if I set it to 1.05, and it decides it needs more, will it go up as needed (up to the cap of 1.175)? This seems to suggest that, while you can change voltage, it really gains you nothing, currently.

Well that bodes well if you can adjust voltage downwards. Hopefully a bios unlock would be enough to move it upwards.

 

[notty22]

Stop your green fanboyism for a second and read my post again.

Sorry, but go back and read all your many posts of useless conjecture and conclusions, and who sounds like a fanboy.

If you set the offset at +100 and your card limit the frequency for Heaven for example at 1070 Mhz why in the world would crash setting the offset higher?

Doesn't make sense for me unless the GPU Boost isn't working properly or the whole system is a big flop and it's based in some arbitrary limits set for all cards.

@Ferzerp

I'm just saying that this system isn't working as intended or it's a huge turd for overclockers because it's based on arbitrary factory set caps and don't squeeze all the card's potential.

I don't think that this system is pushing the card more than traditional overclocking would do.

For stock and mild overclock it's great (so it's for me) but for overclockers it's just frustrating and no fun at all.

I'm just trying to understand it lol. Don't get so angry over this.

I don't know if it's caused because of the GPU Boost throttling or what. It makes no sense to have no stability when you set the offset too high but it does fine limiting the very same apps at lower speeds. As I said Nvidia ensured to have stability across all the frequency and voltage combinations.

http://www.techpowerup.com/forums/showpost.php?p=2581204&postcount=25

It looks like something fixable at driver level but I don't know.

 

[ViviTheMage]

Card runs COOL! I wa barely pushing 100 TDP playing LoL, haha.

I did OC it, but everytime I would put in say, 200, it would put it back down to something like 132.

 

[Imouto]

@notty22

Read topic's title. I'm talking every single time about the system and its compatibility with high overclocks.

In fact I said twice already that I'm looking after these cards since I do mild overclocks or none at all.

That's the biggest problem about being a fanboy, you see enemies everywhere.

 

[Ferzerp]

Card runs COOL! I wa barely pushing 100 TDP playing LoL, haha.

I did OC it, but everytime I would put in say, 200, it would put it back down to something like 132.

That is a percentage, so the 100 TDP is a percentage of the default max TDP.

 

[ViviTheMage]

That is a percentage, so the 100 TDP is a percentage of the default max TDP.

you sure, I thought it said something like 195 somewhere in the software?

It was not even running at max clocks, and was barely hitting 50C...seems odd that it would run 100% and not be full clocks.

 

[notty22]

Saving people from themselves is not such a bad idea. I'm referring to o/cing. It's nice, we are spoiled we can take things to the edge in the gpu world, with software sliders. Building complex hardware and software and implementing them, dual bios switches fit in there to, so users can take things close to the edge but not burn cards and then force sad owners to call and ask for warranty not knowing what happened etc. Is a good thing. Less failures and more actual gamers getting more power out of their cards doing almost nothing.
For many years, modders have had to do extra work, to take things to next levels. solder mods for cable/ satellite boxes , not just for illegal use either, or turn wrenches in their cars, change turbo's and not just dial in boost from their seat etc.
I see these complex TDP/boost/ bios implementations, from both companies as new territory.
For those that want more levels of power/speed or whatever it is, there are always ways.

 

[Ferzerp]

you sure, I thought it said something like 195 somewhere in the software?

It was not even running at max clocks, and was barely hitting 50C...seems odd that it would run 100% and not be full clocks.

The precision software does not show absolute TDP anywhere. It only shows % of default max.

 

[railven]

Saving people from themselves is not such a bad idea. I'm referring to o/cing. It's nice, we are spoiled we can take things to the edge in the gpu world, with software sliders. Building complex hardware and software and implementing them, dual bios switches fit in there to, so users can take things close to the edge but not burn cards and then force sad owners to call and ask for warranty not knowing what happened etc. Is a good thing. Less failures and more actual gamers getting more power out of their cards doing almost nothing.
For many years, modders have had to do extra work, to take things to next levels. solder mods for cable/ satellite boxes , not just for illegal use either, or turn wrenches in their cars, change turbo's and not just dial in boost from their seat etc.
I see these complex TDP/boost/ bios implementations, from both companies as new territory.
For those that want more levels of power/speed or whatever it is, there are always ways.

I'd have to agree with Notty. I think GPU Boost is a big winner for nVidia. My GF doesn't let me OC her PC, but after I told her about GPU Boost she seemed all for it.

Of course she still wouldn't let me raise the offset, however, I will when she isn't looking haha.

 

[ViviTheMage]

The precision software does not show absolute TDP anywhere. It only shows % of default max.

Gotcha, but I assume if it's at 100% in the software, it is maxed out if it's 100?

 

[Ferzerp]

I would assume so. I leave mine set on 132%, and have seen values up to 128% in 3dmark11.

100 corresponds to 195W I believe.

 

[Grooveriding]

There is currently a way to add a lot more voltage to a 680 if you own an Asus board with the vsense feature. I've also read the 1.21 voltage will be unlocked with a BIOS at some point soon.

The vsense mod came from here http://kingpincooling.com/forum/showthread.php?t=1680 can be done with the instructions here :

http://www.datafilehost.com/download-e84c3638.html



You attach a wire from the green circled location on your 680 to your vsense port on your mobo to add voltage

And attach a wire here to read the 680 core voltage

You can then add voltage to the 680 in your mobo's VGA Paradise settings.

A little too OTT for me, but if you really want to add some volts and have an asus mobo with this feature...

 

[bunnyfubbles]

wow I definitely love this new way of overclocking

I know for those who prefer the manual way might not like it if only because it can be more fun to sort of tweak things on your own and figure out what your own card's sweet spots are, but I'm all for making things easier

just like I loved transitioning to a K series CPU because it was just so much less headache trying to study up on all the exact settings I'd need to tweak, I'm loving the 680 for the same reason. A ~20-25% overclock was as simple as tweaking a few sliders and that was that.

Granted, I think nVidia kind of neutered the 680's potential by limiting it to dual 6 pin power in the reference design, however I also think it will only be a matter of time before non reference designs start showing up to buck that decision, that is unless nVidia really doesn't want the 680 to be taken that much further and denies partners from doing so. Because right now the 680 really does seem to be limited more by how much juice it can be fed more-so than temps being a problem.

Or who knows, maybe there's some simple limitation somewhere along the line where more power just won't do any good unless you have some extreme cooling (sub zero, etc, much like how Ivy seems to be unfortunately shaping up) and might not even be stable (shader clock might just be too high and become unstable before the GPU clock for instance)

 

[Grooveriding]

wow I definitely love this new way of overclocking

I know for those who prefer the manual way might not like it if only because it can be more fun to sort of tweak things on your own and figure out what your own card's sweet spots are, but I'm all for making things easier

just like I loved transitioning to a K series CPU because it was just so much less headache trying to study up on all the exact settings I'd need to tweak, I'm loving the 680 for the same reason. A ~20-25% overclock was as simple as tweaking a few sliders and that was that.

Granted, I think nVidia kind of neutered the 680's potential by limiting it to dual 6 pin power in the reference design, however I also think it will only be a matter of time before non reference designs start showing up to buck that decision, that is unless nVidia really doesn't want the 680 to be taken that much further and denies partners from doing so. Because right now the 680 really does seem to be limited more by how much juice it can be fed more-so than temps being a problem.

Or who knows, maybe there's some simple limitation somewhere along the line where more power just won't do any good unless you have some extreme cooling (sub zero, etc, much like how Ivy seems to be unfortunately shaping up) and might not even be stable (shader clock might just be too high and become unstable before the GPU clock for instance)

That's an interesting take on it. If you check around on the overclock.net, xtremesystems, [h] and nvidia's own forums, there is a lot of interest in being able to disable gpu boost altogether.

It's a nice feature for someone who doesn't tweak or do anything and just plugs in their card and goes. If you like to overclock it's more a hindrance than a benefit. It's constantly pushing clocks and voltages around, and in tandem with the inability to increases voltages, feels to be working against good overclocks on these cards.

I've had a few crashes/TDRs in BF3 which were precipitated by the cards jumping around with voltages. It seems to be dropping voltages as the load lightens and then if the load quickly increases, it can crash on you. This isn't happening at stock, but does when you start trying to overclock. If you give the overclocked card a steady high load looping a benchmark or 3DMark11, totally stable. Gives me the impression when you're using offset overclocks that the gpubooost clock and voltage fluctuations at times do not react fast enough to increased load and you get a crash/TDR.

It's a decent feature and has its place, but they need to provide the option to disable it. With straight voltage and clock control, overclocking would be more stable and likely yield better results. It would just be more satisfying to have control over the hardware rather than an automated system.

 

[ViviTheMage]

would be nice if they had a switch, or software switch you could disable it, if you wanted to.

 

[Elganja]

I got sick and tired of waiting for the Hydro's and got 3 SC ones from newegg. Ordered my ek blocks from performance-pc's as well...

Vivi, what did you end up doing for oc'ing? Would like to pick your brain for tips and hints! Groove you too... i know you are a oc'ing fool

 

[Grooveriding]

I got sick and tired of waiting for the Hydro's and got 3 SC ones from newegg. Ordered my ek blocks from performance-pc's as well...

Vivi, what did you end up doing for oc'ing? Would like to pick your brain for tips and hints! Groove you too... i know you are a oc'ing fool

Well I have both of mine under water now, I am also using the EK blocks. Initially I felt they somehow made a massive difference because it allowed one of my cards that initially only could use a +135 offset to do +200. Turned out that was not the case and it was a lucky run of stability

The trouble with overclocking the 680 is that you can't raise the vcore. Over on the EVGA forums they have confirmed none of their custom cards will allow it. I think what is going to happen is there will be custom cards with hard set core voltage increases with binned chips for better clocks.

For the ref. 680s you basically push the power threshold to 132% and start with a +100 offset and test stability. If it is stable, raise it +15 to core and test again. Battlefield 3 seems to be the best test of stability for me because it fails where other programs do not initially (3dmark11) If you are clocking too high you should get a TDR, that is what I have happen at least.

As far as the memory, it overclocks like stink. I can do +500mhz on both of my cards (1000mhz effective), but keep it at +400 because I prefer to take it easy on the memory. The memory overclock is more important than the core clock as it delivers better FPS gains, at least this is what I find @ 1600P.

The lack of voltage control is pretty much a bummer, but it is what it is.

As far as the EK blocks you are getting... my 680s idle @ 25-27C and full load in BF3 is 38-41C :awe: They run super cool under water. I run them in a parallel config. This is with my fans full blast using a 360 & 240 rad. If I run my fans silent temps go up about 5C. They run ridiculously cool with water cooling.

 

[ViviTheMage]

I got sick and tired of waiting for the Hydro's and got 3 SC ones from newegg. Ordered my ek blocks from performance-pc's as well...

Vivi, what did you end up doing for oc'ing? Would like to pick your brain for tips and hints! Groove you too... i know you are a oc'ing fool

I actually didn't do any, just left it default...runs just as good, if no tbetter, as my old setup, 2x6950's.

 

[Elganja]

Well I have both of mine under water now, I am also using the EK blocks. Initially I felt they somehow made a massive difference because it allowed one of my cards that initially only could use a +135 offset to do +200. Turned out that was not the case and it was a lucky run of stability

The trouble with overclocking the 680 is that you can't raise the vcore. Over on the EVGA forums they have confirmed none of their custom cards will allow it. I think what is going to happen is there will be custom cards with hard set core voltage increases with binned chips for better clocks.

For the ref. 680s you basically push the power threshold to 132% and start with a +100 offset and test stability. If it is stable, raise it +15 to core and test again. Battlefield 3 seems to be the best test of stability for me because it fails where other programs do not initially (3dmark11) If you are clocking too high you should get a TDR, that is what I have happen at least.

As far as the memory, it overclocks like stink. I can do +500mhz on both of my cards (1000mhz effective), but keep it at +400 because I prefer to take it easy on the memory. The memory overclock is more important than the core clock as it delivers better FPS gains, at least this is what I find @ 1600P.

The lack of voltage control is pretty much a bummer, but it is what it is.

As far as the EK blocks you are getting... my 680s idle @ 25-27C and full load in BF3 is 38-41C :awe: They run super cool under water. I run them in a parallel config. This is with my fans full blast using a 360 & 240 rad. If I run my fans silent temps go up about 5C. They run ridiculously cool with water cooling.

thanks for the info. I'm not looking for something insane or anything, just a little extra juice. I'll probably set it to +100 and +400 mem for all 3 of them and call it a day... or so I tell myself now... lol

as far as the waterblocks go, I'm going to use the ek bridge, but in series not parallel. There was some reason to my madness, but long ago I decided to go serial instead of parallel

 

[Ferzerp]

thanks for the info. I'm not looking for something insane or anything, just a little extra juice. I'll probably set it to +100 and +400 mem for all 3 of them and call it a day... or so I tell myself now... lol

as far as the waterblocks go, I'm going to use the ek bridge, but in series not parallel. There was some reason to my madness, but long ago I decided to go serial instead of parallel

Fewer tubes, worse cooling.

 

[Grooveriding]

thanks for the info. I'm not looking for something insane or anything, just a little extra juice. I'll probably set it to +100 and +400 mem for all 3 of them and call it a day... or so I tell myself now... lol

as far as the waterblocks go, I'm going to use the ek bridge, but in series not parallel. There was some reason to my madness, but long ago I decided to go serial instead of parallel

I read up a bit on parallel vs serial. Seems there are pros and cons to both, but in general, parallel seemed better to me.

I have a lot of 90 degree fittings and a few different blocks, so parallel offers less restriction than serial. Also I am a bit OCD, so I like that parallel gives consistent temps across your GPUs, whereas serial will have the first GPU getting hit with water cooler than the following ones. At least this is what I have read.

I'm using an EK bridge as well, actually have the triple parallel with a blank in the middle in case I decide to get another 680 at some point :ninja:

 

[thilanliyan]

Also I am a bit OCD, so I like that parallel gives consistent temps across your GPUs, whereas serial will have the first GPU getting hit with water cooler than the following ones. At least this is what I have read.

Yes, your 1st GPU in the loop would get cooler water than the 2nd, but the difference should only be a couple of degrees.