What do you think of nVidia locking down voltage?

[SirPauly]

It is kinda confusing based on the innovation and dynamic nature of GPU Boost.

 

[Grooveriding]

Even with AMD having the edge performance wise in most brackets there is not enough of a difference to put much pressure on nvidia to do anything price wise or in terms of releasing a proper robust high end card.

Hopefully the 8970 is strong and/or releases after the GK110 780 card, if even the 780 is going to be a GK110 card and hopefully not another clocked to its limits tiny die. The 8970 being strong could put pressure on nvidia to release a proper and robust high end card with a proper voltage control chip as in the past.

This whole GPU boost junk and lack of voltage control on the 680 is totally underwhelming. The 680 looked decent at release with its price/performance, but these Kepler cards are really underwhelming with the current state of the better performance, price and flexibility of the 7970/7950.

 

[Keromyaou]

Gtx680 seems to be a heavily overclocked mid-range card after all, which was suspected by many in the beginning. Gpus (or cpus) should allow overvoltage to give customers confidence that they are operating well inside the safe limit. The default core voltage of Gtx680 should have been less than the actual default voltage and the card should have operated well below its safe limit from the beginning. However this conservative approach didn't shine Gtx680 in terms of performance. So it seems that Nvidia overvoltaged/overclocked the card to the limit. I hope that Gtx780 will be a true high-end card. If Gtx780 doesn't allow overvoltage like Gtx680, I probably interpret it as an indication that the card is merely an overclocked mid-range one again.

 

[Silverforce11]

I think your heading in a good direction. Finally some people are really looking at this a little deeper......

I dont think theres a major fundamental difference at the transistors being used thats causing this, for two reasons:

1. Entire different architectures, yet Kepler and GCN seems to top out at ~1.3ghz +/- a few percent. This is true for every single 28nm card currently, from bottom to top, except for the 660 which is crippled due to board TDP specs. This suggests the node is similar for both.

2. There's a pretty good reason NV is capping at 1.175vcore; we know based on extensive OC/Ovolt for GCN that its very efficient when its <= 1.175vcore.. once you crank it above that, power draw skyrockets.

Combined with the "just enough" board specs for reference kepler cards, its extremely risky to have a scenario where power draw is skyrocketting. Cards burning/exploding is not what NV wants to see repeated. AMD can release their lame boost bios with 1.25v is because their designs are overkill on power circuitry components and can handle it.

However, custom boards should not have any issue, which is why MSI has said NV originally gave them the go ahead for overvolt, but due to competitor complaints, NV decides to even the playing field for a few reasons (which is being speculated in this thread).

 

[sontin]

Again this is like saying base voltage for a 3570k CPU is 1.1v. Processors are not rated at idle speed from the manufacturer. Notice how nobody says the clock speed of a 3570k is 1600Mhz or a GTX 670 is 350Mhz. You have zero control over your GPU above 1.175v. This means that if your limit is not temps, not tdp, and not the chip itself...you are stuck because you cannot put 1.2v. You don't have the ability to test your GPU to see the limits.

And what's different to previous generation? There was a limit, too. You were not able to give a Fermi card more than ~1,2V. Now you are complaining about the exact same limitation. A little bit late, i guess. :whiste:

Kepler's Boost is boosting clocks and voltage. At base clock nVidia is using ~1V. That is the stock voltage. You can force on every Boost step the 1,175V if you want. In the highest boost mode you get the highest voltage. And 1,175V is not stock because that the card runs with 1,175V depends on the power and temperature limit.

 

[blackened23]

You're taking the voltage of the LOWEST idle state and calling that the stock voltage? Just to correct your error: 1v is the voltage used in idle state. That is 315-350mhz. I believe you're confusing the under voltage used during idle states. That is not stock voltage, the chip will undervolt itself when not in full use. All CPUs and GPUs do this for the most part - if they're not being utilized, they will under volt.

Let's review:

Idle state voltage:

315-350 mhz, .95V - 1V this is being UNDERVOLTED during idle state

Full usage:

1216mhz (avg) 1.175V. This is stock voltage.

Extrapolating from this, we learn that the 7970 has a stock voltage of .95V (idle state) as well, and the ivy bridge has a .9V (idle state) stock voltage. Thanks for that clarification. All kidding aside, referring to an idle state voltage (an undervolt) as "stock" is just incorrect.

 

[blackened23]

My GTX 680s in the lowest idle state, 324 mhz. .992V.

 

[wand3r3r]

Now that you've cleared that up can we return to the topic at hand.

 

[blastingcap]

And what's different to previous generation? There was a limit, too. You were not able to give a Fermi card more than ~1,2V. Now you are complaining about the exact same limitation.

You do realize that Fermi was built on an entirely different 40nm process right?

 

[Silverforce11]

The problem with clearing up sontin's "mistake" is that he isn't even man enough to admit when he's blatantly wrong and apologize or withdraw such statements.. you will see his same misleading crap again in another thread soon enough.

Its not great to call out an AT member, but its the sad truth from many of his recent posts.

 

[sontin]

You're taking the voltage of the LOWEST idle state and calling that the stock voltage? Just to correct your error: 1v is the voltage used in idle state. That is 315-350mhz. I believe you're confusing the under voltage used during idle states. That is not stock voltage, the chip will undervolt itself when not in full use. All CPUs and GPUs do this for the most part - if they're not being utilized, they will under volt.

Let's review:

Idle state voltage:

315-350 mhz, .95V - 1V this is being UNDERVOLTED during idle state

Full usage:

1216mhz (avg) 1.175V. This is stock voltage.

Extrapolating from this, we learn that the 7970 has a stock voltage of .95V (idle state) as well, and the ivy bridge has a .9V (idle state) stock voltage. Thanks for that clarification. All kidding aside, referring to an idle state voltage (an undervolt) as "stock" is just incorrect.

So, you have Kepler cards for more than 6 months and don't know how the Boost is working? That makes it clear for me why people have no clue.

Stock is not 1,175V. That is the voltage for the highest Boost step. Stock is base clock. That is the clock you get when the last is not over 60% or so:


What you talking about is the Boost clock. And that using different voltage from the default up to 1,175V, for example that step:


You see i'm over my base clock with the stock voltage but not using the highest Boost step.

 

[SirPauly]

It's great to see features like GPU Boost over-all but certainly can see the view of moving backwards, specifically for an OC enthusiast point-of-view that may desire to have the control more-so than dynamics that nVidia may control. However, I see GPU Boost as moving forward, too, can there be compromises made between nVidia and their partners for some middle ground?

The key is if one feels so strongly, certainly share a view, vote with one's wallet and if there is enough momentum or lost sales --- nVidia may rethink this moving forward.

Personally like AIB differentiation, very much so and was odd to see what seems like in-fighting with nVidia and their partners on this.

 

[sontin]

nVidia is not moving backwards. You had no full control over the voltage with Fermi. There was a software limit depended on the stock voltage of every card. Only a new bios or the modification of the card could gave more options.

People don't understand that nVidia is now overvolting their Kepler cards with the Boost function. Instead of only using one voltage step they are using every one which are available. Nothing is different from the previous. Things changed with Kepler but there was no step backwards.

 

[SirPauly]

Personally considered the volts offered by GPU boost, 1.175, the highest default volt, and was hoping AIB's could offer a bit more flexibility with some manual adjustments over this. It was the reasoning why I purchased a MSI power edition GTX 670. Sadly, this may not be the case and certainly an unideal situation, where some AIB's were promoting adjustable volts and now can't.

I'm not a huge finger pointing or blame poster but this kinda sucks over-all.