• 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.

Fermi has 480 shaders, 675MHz+ clock

Taking that at face value and knowing the core/shader ratio - that would put the 480SP part at between 27.7% and upto 40% faster than a stock 5870.
 
I'm really not bothered by the performance of these cards, I don't intend on buying a first gen GF100 and I don't give a toss if its faster or slower than the ATI 5 series.

And its not that ridiculous conclusion as its based on extrapolating known data.


EDIT: I wouldn't hold too much to that info tho hence why I said taking it at face value, as nVidia is chopping and changing things like mad atm and whats true one day might not be the next.
 
Last edited:
If correct and we're looking at 700 core + 1400 approx on shaders, that's much better than the previous rumours doing the rounds. Like I said in another thread, it seems the 512 shader part will possibly be the 'ultra' model.

On a side note if it can be overclocked (possibly with watercooling) upto the 800 core 1500+ shader area then it could be a beast, if a hot one. I'm thinking the big fire demon out of DMC4. :P
 
Rroff said:
I'm really not bothered by the performance of these cards, I don't intend on buying a first gen GF100 and I don't give a toss if its faster or slower than the ATI 5 series.

And its not that ridiculous conclusion as its based on extrapolating known data.


EDIT: I wouldn't hold too much to that info tho hence why I said taking it at face value, as nVidia is chopping and changing things like mad atm and whats true one day might not be the next.
Click to expand...

You don't have any known data about performance at all, you're extrapolating based on your complete guesses and no, no one else on earth believes it will be that much faster.

You keep missing some vital things when you mention things, clocks haven't been increased massively over earlier leaks, which is the ONLY info you could possibly have. You're taking the early data I assume and ASSUMING you know the clocks of those tests, even though you most certainly do not, as no one does, and upping/decreasing those numbers as you see fit based on the possible clock change.

Its most likely the benchmark results we've seen to date are on 512sp parts at circa 650Mhz clocks, meaning around a 7.5% clock speed increase, with a roughly 7.5% SP reduction...... which would lead a normal person to conclude it would likely be a bit of a wash. More than anything the less sp's and marginally higher clocks probably result in them getting a higher yield.

All along we've been told the 480gtx will be 5-10% faster than a 5870, for months, from many sources and from the most biased Nvidia sources now aswell. Yet you've decided its up to 47% faster, but at least 27.7% faster.

Rubbish.

The less SP clusters, the worse, because Fermi can only put new code through a single cluster at its smallest I believe, similar for AMD but they have more clusters. Meaning to run physx on the gpu the more clusters the better as Physx will always take at least one(or any other gpu code, open cl physics or acceleration of anything).

Consider we're still in the situation where a £200 5850 can likely overclock beyond any overclocked performance the 480gtx can give, which will cost over twice as much with zero supply, and a 470gtx is set to cost £300, so 50% more for 5% more performance and far less performance with both overclocked.

Thats of course assuming the clocks/sp's are really confirmed, as others have confirmed in person, AMD and Nvidia like to leak almost accurate, but slightly incorrect information as a way to track down leaks. But again I'll point out, this could easily be the spec, or even lower than what we've seen benchmarked so far. People are incorrectly reading into a sudden 10% increase in clocks out of no where, its not, its a 10% increase in clocks over what we guessed the clocks would be. You can't take existing benchmarks on UNKNOWN cards and increase the scores 10%.

If AMD came out now and said a 5870 actually has 400Mhz clocks but it lied about clock speeds, it doesn't change the performance just because your knowledge of the specs changed.

THough again I don't think final specs are what we saw benchmarked, not exactly, they were "best guess at what performance they could get", and while they are playing with clockspeed/sp combo's to get the best number of cores they can, I doubt performance will change much more than 2-3% either way.
 
Last edited:
Mr Paul said:
He has a point Rroff
Click to expand...

If he has a point... hes contradicting Charlie Demerjian as one of his articles also backs up the information I've used to come to these figures... take your pick.

All I will say is my information is NOT based on a guess and leave it at that.
 
titaniumx3 said:
This is getting tiresome, I mean isn't it a bit late to be changing such important details? So basically nvidia were misleading people all this time. :mad:
Click to expand...

Who says Nvidia are misleading people. Nvidia have said nothing officially at all. About the only things people have said come from ATI fanboys!
 
Looks like the Marine Iguana will be happy if these figures are close to the truth IIRC hes looking at GF100 cards for folding and they'll be performing at
~2x that of the GTX285
 
drunkenmaster said:
You don't have any known data about performance at all, you're extrapolating based on your complete guesses and no, no one else on earth believes it will be that much faster.

You keep missing some vital things when you mention things, clocks haven't been increased massively over earlier leaks, which is the ONLY info you could possibly have. You're taking the early data I assume and ASSUMING you know the clocks of those tests, even though you most certainly do not, as no one does, and upping/decreasing those numbers as you see fit based on the possible clock change.

Its most likely the benchmark results we've seen to date are on 512sp parts at circa 650Mhz clocks, meaning around a 7.5% clock speed increase, with a roughly 7.5% SP reduction...... which would lead a normal person to conclude it would likely be a bit of a wash. More than anything the less sp's and marginally higher clocks probably result in them getting a higher yield.

All along we've been told the 480gtx will be 5-10% faster than a 5870, for months, from many sources and from the most biased Nvidia sources now aswell. Yet you've decided its up to 47% faster, but at least 27.7% faster.

Rubbish.

The less SP clusters, the worse, because Fermi can only put new code through a single cluster at its smallest I believe, similar for AMD but they have more clusters. Meaning to run physx on the gpu the more clusters the better as Physx will always take at least one(or any other gpu code, open cl physics or acceleration of anything).

Consider we're still in the situation where a £200 5850 can likely overclock beyond any overclocked performance the 480gtx can give, which will cost over twice as much with zero supply, and a 470gtx is set to cost £300, so 50% more for 5% more performance and far less performance with both overclocked.

Thats of course assuming the clocks/sp's are really confirmed, as others have confirmed in person, AMD and Nvidia like to leak almost accurate, but slightly incorrect information as a way to track down leaks. But again I'll point out, this could easily be the spec, or even lower than what we've seen benchmarked so far. People are incorrectly reading into a sudden 10% increase in clocks out of no where, its not, its a 10% increase in clocks over what we guessed the clocks would be. You can't take existing benchmarks on UNKNOWN cards and increase the scores 10%.

If AMD came out now and said a 5870 actually has 400Mhz clocks but it lied about clock speeds, it doesn't change the performance just because your knowledge of the specs changed.

THough again I don't think final specs are what we saw benchmarked, not exactly, they were "best guess at what performance they could get", and while they are playing with clockspeed/sp combo's to get the best number of cores they can, I doubt performance will change much more than 2-3% either way.
Click to expand...

No where reliable has said the 480 is only 10% faster. More likely, based on earlier reliable rumours is the 470 is 10% faster than a 5870 and the 480 competes better with the 5970 although loosing. Thats what we got after A2 silicon benchmarks.
 
Rroff said:
Taking that at face value and knowing the core/shader ratio - that would put the 480SP part at between 27.7% and upto 40% faster than a stock 5870.
Click to expand...

If this was the case surely nvidia (and any subsequent leaks) would be shouting this from the roof tops?

As it stands, all we are seeing is 5-10% banded around everywhere.
 
Oh and because some people don't seem to get it...

I'm not saying the GTX480 will be such and such faster or slower, I'm saying if we take the core speeds suggested here at face value it would indicate, when taken in conjunction with other information, that a 480SP part would fall in a particular ballpark... hence why I said face value...
 
You must log in or register to reply here.
Back
Top Bottom