Kepler Overclocking 101: The GTX 670/680 Overclocking Master-Guide (Overclock.net)

[toyota]

I still will never understand where they csme up with 70c as the first throttle point. Why not 80c? I think for a gpu, 70c is really low.

With my setip I know I am at the overclock limits. A few mhz more and I will crash somewhere in some game or benchmark. A lot of trial and error to find this.

I agree that the first throttle should be at 80 C not 70 C. it makes me wonder if Nvidia knows these gpus are actually not that durable.

 

[cmdrdredd]

I agree that the first throttle should be at 80 C not 70 C. it makes me wonder if Nvidia knows these gpus are actually not that durable.

I thought about that but honestly I don't know. I think it's more like they don't want people overclocking to the point that the $400 GPU is beating the $500 GPU.

In my situation, a few extra Mhz isn't going to be a game changer at all.

 

[VisceralM]

I agree that the first throttle should be at 80 C not 70 C. it makes me wonder if Nvidia knows these gpus are actually not that durable.

I think the problem is there are two schools of thought on this.

1.) Nvidia is being upfront and honest that high temps and volts will degrade these cards.
or
2.) Nvidia has set up artificial restraints regarding the 70c limit and voltage controls to artificially set up and lock-in cards so that they can maintain specific performance profiles for their 6xx series cards.

There are more and more custom bios's out there for the 6xx cards, most involve upping volts and removing the 70c restraint. I'm sure it would be possible to have one that does nothing but change the 70c to 80c.

As for which of the above it is, 1 or 2, I'm not sure. You read about a few people here and there having problems when aggressively overclocking, but statistically it's all anecdotal. I lean toward #2 myself, but it's just an opinion.

 

[Rifter]

Thanks for the link OP i find it a very interesting read.

Whats with all the Gordon Freemen quotes? i know the guy is a idiot sometimes but why quote posts from other threads and bash him in this thread for them?(since i see none of his post in this thread, not even one, im assuming all the above quotes came from elsewhere)

 

[toyota]

Thanks for the link OP i find it a very interesting read.

Whats with all the Gordon Freemen quotes? i know the guy is a idiot sometimes but why quote posts from other threads and bash him in this thread for them?(since i see none of his post in this thread, not even one, im assuming all the above quotes came from elsewhere)

he did nothing but derail the whole thread and I guess all of his posts have been removed. heck I think he has been removed too...

 

[lopri]

I did have a suspicion on usefulness of overclocking prior to testing GTX 670 and I feel the same after using it for a couple weeks. Overclocking it is pretty much a useless exercise on these cards and I do consider it as a positive.

Even if you overclock it to go outer bounds of "boost" frequency, the reality is that the card will clock itself according to the load and the environment (heat/power) without your consents. While there is some perceived negatives on this - which I haven't made my mind up yet - but I still think the positives greatly outweigh them.

Besides which, as several members have testified, there is a potential for degradation when these things are overclocked - another reason not to bother with overclocking.

 

[Silverforce11]

I think the problem is there are two schools of thought on this.

1.) Nvidia is being upfront and honest that high temps and volts will degrade these cards.
or
2.) Nvidia has set up artificial restraints regarding the 70c limit and voltage controls to artificially set up and lock-in cards so that they can maintain specific performance profiles for their 6xx series cards.

There are more and more custom bios's out there for the 6xx cards, most involve upping volts and removing the 70c restraint. I'm sure it would be possible to have one that does nothing but change the 70c to 80c.

As for which of the above it is, 1 or 2, I'm not sure. You read about a few people here and there having problems when aggressively overclocking, but statistically it's all anecdotal. I lean toward #2 myself, but it's just an opinion.

There's no reason why their GPU dies are more vulnerable to heat, its manufactured from the same company, same process and node. What i do think is a problem, is because a lot of gtx680 cards initially were all reference designs, and the VRMs on that looks to be JUST enough for a 225W TDP card. Hotter die = more power usage. So NV knows its not a good idea to stress the power circuitry. See, i dont even understand why they don't include an extra mosfet/choke, it hardly costs anything in comparison to the selling price at $500.

It could well be their original power/perf target was lower and they had to change the clocks late in the launch cycle, well after the cards were made.. thus its now running near the redline so to speak. Its well represented in how little OC headroom there is, default boosts hit ~1.2ghz, with most OC not even reaching 1.3ghz boost.. so little headroom/power design suggests the scenario above.

 

[tviceman]

I agree that the first throttle should be at 80 C not 70 C. it makes me wonder if Nvidia knows these gpus are actually not that durable.

The throttle is pretty minimal. I doubt the 13mhz throttle or whatever it is will OMG extend the life of the chip. I do agree that 75-80 degrees would make more sense, but I think that either Nvidia engineers were a little too agressive (accidentally) on the throttling mechanism, OR this is another way they are limiting overclocking.

 

[tviceman]

There's no reason why their GPU dies are more vulnerable to heat, its manufactured from the same company, same process and node. What i do think is a problem, is because a lot of gtx680 cards initially were all reference designs, and the VRMs on that looks to be JUST enough for a 225W TDP card. Hotter die = more power usage. So NV knows its not a good idea to stress the power circuitry. See, i dont even understand why they don't include an extra mosfet/choke, it hardly costs anything in comparison to the selling price at $500.

It could well be their original power/perf target was lower and they had to change the clocks late in the launch cycle, well after the cards were made.. thus its now running near the redline so to speak. Its well represented in how little OC headroom there is, default boosts hit ~1.2ghz, with most OC not even reaching 1.3ghz boost.. so little headroom/power design suggests the scenario above.

http://www.hardocp.com/article/2012/07/30/msi_geforce_gtx_680_lightning_overclocking_redux/4

Kepler up-clocks on the core just as high as Tahiti when a) neither card gets voltage adjustments OR b) when both cards get voltage adjustments. I know the latter is a huge catch-22 since there are only 3 cards that can take voltage adjustments, but gtx680's are hitting 1.3ghz just as often as hd7970's are.