• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The first "proper" Kepler news Fri 17th Feb?

Ok, regarding transistor densities, I have a few questions:

A comparatively lower transistor density in the 7970 explains why AMD has had better yields and could get to market months ahead of Kepler, right?

Does this also explain why the 7970 can overclock so well, and in comparison why we shouldn't expect Kepler to proportionately overclock as much, right?

Not saying it's a bad chip, it could end up thrashing the 7970 for all we know.

No. If you simply take greater spacing between transistors it still depends on things like the amount of resistive coupling between them, and many more factors. The clock speed of a circuit is determined by too many factors to simplify it down like that. But the greatest contributor to higher clocks is feature size. The smaller the features, the higher the max attainable clocks.
 
Last edited:
With launch drivers according to HC,the HD7970 3GB is around 41% faster than an HD6970 2GB:

http://www.hardwarecanucks.com/foru...s/49646-amd-radeon-hd-7970-3gb-review-25.html

According to one of the links in post 1600,the GTX680 is supposedly 25% faster than an HD7970 3GB. If this is true,this would mean the GTX680 is around 36% faster than a GTX580 if HC numbers are used.

If the GTX680 is less than 35% to 40% faster than a GTX580 it would actually be less of a generational jump than between the HD6970 and HD7970 and hopefully this is not the case.

Hopefully,it isn't because at least it means the HD7900 series and hence the HD7800 series will drop in price.
 
Last edited:
If the GTX680 is less than 35% to 40% faster than a GTX580 it would actually be less of a generational jump than between the HD6970 and HD7970 and hopefully this is not the case.

Frankly, I'd be very surprised if GK104 IS that far ahead of the GTX580. I'm expecting more like a 20-25% overall performance bump.

You have to remember that this is, essentially, Nvidia's mid-range card - at least from an architectural perspective. It seems that, given the performance of the 7970, Nvidia are choosing to market it as their high end replacement for the GTX580, with a price to match. This helps them to cover the massive delays on their 'true' high-end part (GK110).

But still, from an architectural point of view, the GTX580 is 77% larger than the GTX680 (according to this chart), whereas the 6970 is only 6.6% larger than the 7970. It's hard to compare the two directly in terms of generational increase.


I know I said earlier I'm not interested in the pricing, but if the GTX680 really is a $549 retail card (which will translate to £420+ over here), then I'd say it's over-priced - no matter how it compares to the 7970. I can't look at the specs for the GTX680 (294mm^2 die, 3.5Bn transistors, 2Gb memory, 256-bit interface) and see it as a high-end Nvidia 28nm card. Everything about it screams that it's an upper mid-range part - except for the price. Which is a shame.
 
Yes indeed, and this is what makes it all look like fail.

/assumption mode...

I would assume that they never intended to run Kepler at those clocks. I also assume that the 700mhz clock was true, but then they got to see Tahiti.

Scrambling with their technology they spent the last two months trying to make it perform as closely to or slightly better than Tahiti.

But, more worryingly if they haven't beaten Tahiti then Kepler = fail. Total fail.

Why? because they have had time to sit down and fiddle with things and put Kepler into a clear lead. If they haven't ? then the fact is they can't.

And can't = fail.

The whole "wafer shortage thing" could simply be Nvidia being punched to the ground by AMD and then saying "wait, no wait, NO WAIT" before getting up and throwing back a feeble punch.

Kepler needs a damn good lead if it is to be successful. It needs a lead that leaves AMD incapable of jiggling Tahiti and coming back with a faster card.

And all signs currently point to no.

Or, the more likely, the 680 is what the 660 would have been clocked up as it bests the 7970, because the 7970 is disappointing.
 
Frankly, I'd be very surprised if GK104 IS that far ahead of the GTX580. I'm expecting more like a 20-25% overall performance bump.

You have to remember that this is, essentially, Nvidia's mid-range card - at least from an architectural perspective. It seems that, given the performance of the 7970, Nvidia are choosing to market it as their high end replacement for the GTX580, with a price to match. This helps them to cover the massive delays on their 'true' high-end part (GK110).

But still, from an architectural point of view, the GTX580 is 77% larger than the GTX680 (according to this chart), whereas the 6970 is only 6.6% larger than the 7970. It's hard to compare the two directly in terms of generational increase.


I know I said earlier I'm not interested in the pricing, but if the GTX680 really is a $549 retail card (which will translate to £420+ over here), then I'd say it's over-priced - no matter how it compares to the 7970. I can't look at the specs for the GTX680 (294mm^2 die, 3.5Bn transistors, 2Gb memory, 256-bit interface) and see it as a high-end Nvidia 28nm card. Everything about it screams that it's an upper mid-range part - except for the price. Which is a shame.

I agree with what you are saying,but the fact if it is being sold as a GTX680 with a high end price it has to be judged compared to the GTX580 and the HD7970. Outside of power consumption if it is £400+ and around 20% to 25% faster than the GTX580,it is only a more efficient HD7970 then.
 
Last edited:
Or, the more likely, the 680 is what the 660 would have been clocked up as it bests the 7970, because the 7970 is disappointing.

I guess it all depends how you look at it.

Performance wise over the last gen it does not disappoint. At all. It slaughters a 6970.

It competes with dual GPU cards. That is also far from disappointing.

The only thing about the 7970 I have seen people disappointed about is the price of it, and that's just the way it is.

Reviews do not slate the card for being disappointing, in fact they are very kind to it.

As for what the 680 was supposed to be? well, a statement from Nvidia themselves suggests it was supposed to be their top card and supposedly it will be.

AMD's 28nm tech will only be truly disappointing from a tech standpoint if it can not beat Kepler once and for all. We will find that out in the next month.

Do that many people overclock GPUs anyway? I know most people here would/might but some people just want the best and wont even know you can overclock a GPU... People still buy Bulldozer CPUs :p

Not every one overclocks a GPU no. However, a product these days is made or broken on its ability to do so. Reviews will all have a part in them where they overclock the card.

Which is why the 6970 did not bode too well in initial reviews, because for overclocking it sucks ass.
 
I can't agree more. Nvidia does have a very strong following and regardless if the 7970 is the faster, people will still buy Nvidia and vice versa.

Indeed. Even if the 7970 is faster and cheaper. I've seen a lot of people on BF3 believing that the 590 is the fastest single GPU Nvidia card available.

And I have no doubt that some people out there will buy a card based solely on which has the most memory.
 
Not every one overclocks a GPU no. However, a product these days is made or broken on its ability to do so. Reviews will all have a part in them where they overclock the card.

Which is why the 6970 did not bode too well in initial reviews, because for overclocking it sucks ass.

my point is, do that many people care? They go onto a PC website, and pick what looks best. Sure this is a high end card, but I know a number of people with 'high-end' PCs that they know nothing about.

Indeed. Even if the 7970 is faster and cheaper. I've seen a lot of people on BF3 believing that the 590 is the fastest single GPU Nvidia card available.

And I have no doubt that some people out there will buy a card based solely on which has the most memory.

Haven't you heard? Memory is the most important thing on a GPU now :p
 
my point is, do that many people care? They go onto a PC website, and pick what looks best. Sure this is a high end card, but I know a number of people with 'high-end' PCs that they know nothing about.
:p

You'd be surprised.

Firstly if a product does not overclock it won't get a score as high as it would if it did. There are very few people who do not overclock.

I mean example. Intel sell a couple of their CPUs unlocked at a higher price.

Seen many people buying non K versions? because I haven't.

Not only that but it's the people with high end PCs that know little that spend the most money. See also - Alienware.

They go on the internet, see the fastest best reviewed parts and then proceed to buy them. They may not know bugger all about how these parts work, but they sure as hell know which ones demand big willy status.
 
There are very few people who do not overclock.

...

Seen many people buying non K versions? because I haven't.

Well, clearly those who post about their hardware on internet forums - especially those who post on an overclocking forum - are highly likely to overclock their kit where possible.

This doesn't mean that the majority of people overclock. My experience in the wider computing field suggests that overclocking of CPUs is restricted to a relatively small subset of enthusiasts, and those who overclock GPUs are an even smaller minority.

If you have any hard figures that suggest otherwise then I'd be interested to see them, but my experience suggests that overclockers represent a very small minority of sales.
 
You'd be surprised.

Firstly if a product does not overclock it won't get a score as high as it would if it did. There are very few people who do not overclock.

I mean example. Intel sell a couple of their CPUs unlocked at a higher price.

Seen many people buying non K versions? because I haven't.

Not only that but it's the people with high end PCs that know little that spend the most money. See also - Alienware.

They go on the internet, see the fastest best reviewed parts and then proceed to buy them. They may not know bugger all about how these parts work, but they sure as hell know which ones demand big willy status.

I disagree, but okay. I meant if its faster [at stock] quite a few people wont care about its OC ability.

There are very few people who do not overclock.
How do you know? I'm sure the majority of the Alienware-type never OC.


Well, clearly those who post about their hardware on internet forums - especially those who post on an overclocking forum - are highly likely to overclock their kit where possible.

This doesn't mean that the majority of people overclock. My experience in the wider computing field suggests that overclocking of CPUs is restricted to a relatively small subset of enthusiasts, and those who overclock GPUs are an even smaller minority.

If you have any hard figures that suggest otherwise then I'd be interested to see them, but my experience suggests that overclockers represent a very small minority of sales.

Exactly!
 
Back
Top Bottom