GTX 670 Boost Clock

[rumpleforeskin]

Last week I purchased a MSI GTX 670 Power Edition OC

Having read a few reviews I was expecting the the boost clock to be 1097.5mhz
But when i'm gaming afterburner is reporting my boost clock as 1228mhz

I understand that if the chip is running over 70'c or drawing more than a predetermined power level it will throttle back the speed but this is higher not lower.

Any Ideas?

 

[tviceman]

You have a great card, it's boosting really high at factory settings. What is the problem?

 

[rumpleforeskin]

The problem was im was not sure if there was a problem, all the reviews of this card had the boost clock at 1097. I did not realise it would clock itself higher without any input from me!

Its not a problem if this is how it was designed, i was just a bit surprised to see it running so high without any overclocking.

 

[3DVagabond]

There is no predetermined limit in terms of absolute clocks. Instead there is a combination of parameters that determine what the max boost clock will be. The guaranteed boost clock is just a worst case scenario.

As far as you getting a good chip or not, it's not just the chip that determines max boost clocks. It's the game/application as well. If the game is not very demanding power usage wise, and I'm not sure what determines that, then the card can clock higher in it.

 

[Termie]

The published boost clock is essentially a fallacy. You will almost never see it. In fact, I just had to look up the published boost clock on my GTX670. It's 1084. Yup, never, ever seen it. It's just a theoretical bunch of nonsense that only occurs under conditions that can't occur while gaming.

The problem I have with Nvidia's marketing with this round of cards is that they felt compelled to advertise a boost clock, but nobody, not even Nvidia, could actually say what each individual card would boost to. So they confused millions of video card buyers with a fake boost clock that they knew was well under what any card would ever hit, just to keep them safe from false advertising claims. The truth is that if you buy two GTX670s of the exact same make and model, there's over a 50% chance that they won't boost to the same level, i.e., they are not equivalent cards. So much for consistent benchmarking...

 

[Annisman*]

The published boost clock is essentially a fallacy. You will almost never see it. In fact, I just had to look up the published boost clock on my GTX670. It's 1084. Yup, never, ever seen it. It's just a theoretical bunch of nonsense that only occurs under conditions that can't occur while gaming.

The problem I have with Nvidia's marketing with this round of cards is that they felt compelled to advertise a boost clock, but nobody, not even Nvidia, could actually say what each individual card would boost to. So they confused millions of video card buyers with a fake boost clock that they knew was well under what any card would ever hit, just to keep them safe from false advertising claims. The truth is that if you buy two GTX670s of the exact same make and model, there's over a 50% chance that they won't boost to the same level, i.e., they are not equivalent cards. So much for consistent benchmarking...

This is so true, which made overclocking my (3) gtx 670s a complete PIA. I won't even oc my current two card setup, just not worth the hassle.

I basically just set the fans to 100% and let the clocks go as high as they want. Very depressing for those of us who are interested in fine tuning. The only saving grace is the outstanding performance out of the box.

 

[Ferzerp]

The published boost clock is the absolute minimum that any of the cards will have for a max boost. That's all it is.

 

[Termie]

The published boost clock is the absolute minimum that any of the cards will have for a max boost. That's all it is.

True, but it has little relation to what actually happens in real games.

 

[Ferzerp]

I don't follow.

If demand is there, as long as temperature and power draw don't preclude it, it will sit at max boost. If max boost would require power above the power cap, the speed is downrated until it would not. If temperature is over 70C, it reduces the boost by some number of Mhz (6 or 8 or something) per 5C over 70.

It's not random at all.

 

[Termie]

I don't follow.

If demand is there, as long as temperature and power draw don't preclude it, it will sit at max boost. If max boost would require power above the power cap, the speed is downrated until it would not. If temperature is over 70C, it reduces the boost by some number of Mhz (6 or 8 or something) per 5C over 70.

It's not random at all.

Max boost is not in the specifications. It's not the published boost clock, and differs for every card.

 

[Ferzerp]

Correct, but it will never be less than what is published. You know all cards are at least that...

 

[Termie]

Correct, but it will never be less than what is published. You know all cards are at least that...

...and at the same time, potentially nowhere near the performance of cards benchmarked in published reviews.

 

[notty22]

...and at the same time, potentially nowhere near the performance of cards benchmarked in published reviews.

As long as those review samples have the same voltage as retail cards, which they probably have. In many reviews the gpu-z or voltage is shown or mentioned. And no software adjustments are made to any of the available options in tuner programs. IE power adjust etc. Then identical models will be within a few % of each other.
Not sure why you are suggesting loaded samples. If a card comes with higher clocks or more shaders than it's supposed to, it's picked up on.

edit: The op's card is reflective of many reviews, is it a review sample?

 

[Termie]

As long as those review samples have the same voltage as retail cards, which they probably have. In many reviews the gpu-z or voltage is shown or mentioned. And no software adjustments are made to any of the available options in tuner programs. IE power adjust etc. Then identical models will be within a few % of each other.
Not sure why you are suggesting loaded samples. If a card comes with higher clocks or more shaders than it's supposed to, it's picked up on.

edit: The op's card is reflective of many reviews, is it a review sample?

The sampled cards may be higher or lower than cards at retail, but either way, it's not ideal. The OP's card may be better than review cards, but it's hard to know, as reviews don't always mention actual boost clocks.

I had two reference models previously. One boosted to 1058, the other 1084. That's enough of a difference that I'd call them different products, and I'm sure the actual range in shipping cards is much wider. It gets worse when you start paying for factory overclocked models and are actually looking for a certain amount of overclocking but don't know what you'll get.

 

[rumpleforeskin]

edit: The op's card is reflective of many reviews, is it a review sample?

I don't know the cards history, I purchased it as a B-Grade from overclockers