AMD DX11 RV8xx Card Exposed

[>HyperlogiK<]

Geforce 4, 6 and 7 and 8 series all offered near doubled performance, as did the original Radeon, the 9 series and the 4 series.

[Oberon]

I think power consumption may still be up in the air. Cypress is only supposed to require one PCI-E power connector, right?

[Chad Boga]

I don't agree with that beeing the normal progression rate in the videocard arena. Given that manufacturers come with more or less one generation a year, it would be much higher than usual tech progression (doubling each two years). Indeed, I think that only the great surprise/success/hit chips have managed to double over their previous generation (R700, G80...). I'd say that a much more realistic and paired to other technologies and Moore's Law formula would be:

next gen = 1.5*previous gen

Given that a new generation is coming every year (your formula could be good if there were a 2 years gap between generations)

Video card generations have been coming out in recent years, 18 months or so apart, not 12 months.

[Manicdan]

i was really happy with how well the 4870 did for its chip size, i think they may be going larger simply to have single chip solution that may still rival all competition, but hopefully keep with their small cores at a great price that worked so well for the 4000s

[BigDisk]

I don't agree with that beeing the normal progression rate in the videocard arena. Given that manufacturers come with more or less one generation a year, it would be much higher than usual tech progression (doubling each two years). Indeed, I think that only the great surprise/success/hit chips have managed to double over their previous generation (R700, G80...). I'd say that a much more realistic and paired to other technologies and Moore's Law formula would be:

next gen = 1.5*previous gen

Given that a new generation is coming every year (your formula could be good if there were a 2 years gap between generations).

Said that, it seems that this generation both ATI and NVIDIA are going to double the performance of their chips... as it seems that performance/watt it's not doubled (or close to that), I'm a little scared about the power consumption of the bigger chips of both ATI and NVIDIA this round.

Except that 1.5* previous gen would surpass Moore's law of doubling the performance every 24 months (it would result in 2.25x the inital performance every 24 months).

If we assume the new generations of card are coming out every year, and that they respect Moore's law of transistor density (and that transistor count equals performance), then the performance would be

nextgen = sqrt(2) * previous gen
or roughly nextgen = 1.41 * previous gen

Just sayin' ...... (sorry about being anal)

[Manicdan]

Except that 1.5* previous gen would surpass Moore's law of doubling the performance every 24 months (it would result in 2.25x the inital performance every 24 months).

If we assume the new generations of card are coming out every year, and that they respect Moore's law of transistor density (and that transistor count equals performance), then the performance would be

nextgen = sqrt(2) * previous gen
or roughly nextgen = 1.41 * previous gen

Just sayin' ...... (sorry about being anal)

its honestly tough to compare across generations when there is a major change in how things are handled, ie DX versions. if we look at just a framerate change, it may be totally different from the changes in furmark rendering.

there is however a different way to look at it, which it overly simplified. the real changes only occur with new processes, going from 55nm to 40nm etc. if there are any delays, then everything can be held back by months, and when expecting a new one every 24 months, 2 month delay is pretty massive in that formula.

looking at AMD, the 45mn quads came out at the beginning of this year, and intel is showing off a 32nm already. the performance jump per year for AMD has been quite sucky since their attempt to switch to 45nm, and they felt the pain of using 65nm for like what, 3 years?!?

so to end it, im not a big fan of trying to say one generation will be x more powerful than the previous. as there are only a few real things that go into it, and each of them can sway the numbers very easily.

[Baron_Davis]

To the guy who wanted the price on a 5890 x2 or w/e it will be called.

I'm gonna stay predictable and say it will be the same as a 4870 x2 so around 600.

I say their main stream card (read: 5890) will be around 300-350. The 5850 (read: 4850 equivalent) will probably be around 200-250. The rest...I don't care about.

[labs23]

To the guy who wanted the price on a 5890 x2 or w/e it will be called.

I'm gonna stay predictable and say it will be the same as a 4870 x2 so around 600.

I say their main stream card (read: 5890) will be around 300-350. The 5850 (read: 4850 equivalent) will probably be around 200-250. The rest...I don't care about.

Those prices are very reasonable, I'd hope it will be then. But higher than those figures won't be a surprise.

[flippin_waffles]

So the USS Hornet is where AMD is going to launch Evergreen from from. Very interesting and very original..

[TurboDiv]

They might be planning on dropping down some cards with attached bombs on Nvidia headquarters... "World Card War I" ??... **runs and hides in the basement with a tin foil hat*

[cegras]

Except that 1.5* previous gen would surpass Moore's law of doubling the performance every 24 months (it would result in 2.25x the inital performance every 24 months).

If we assume the new generations of card are coming out every year, and that they respect Moore's law of transistor density (and that transistor count equals performance), then the performance would be

nextgen = sqrt(2) * previous gen
or roughly nextgen = 1.41 * previous gen

Just sayin' ...... (sorry about being anal)

Performance != transistor density. You're drawing the wrong conclusion from that. For example, increased transistor density = more cache = linear, corresponding increase in performance? No.

[eric66]

http://vimeo.com/6182068

r8xx in action dirt 2

[Chad Boga]

So the USS Hornet is where AMD is going to launch Evergreen from from. Very interesting and very original..

Perhaps they will unfurl a large "Mission Accomplished" banner.

[jaredpace]

Perhaps they will unfurl a large "Mission Accomplished" banner.

HAHAHA

[>HyperlogiK<]

and that transistor count equals performance

That's quite a big assumption.

[Sh1tyMcGee]

http://vimeo.com/6182068

r8xx in action dirt 2

Seems like a pre rendered video to me. Not actual gameplay.

[Manicdan]

Seems like a pre rendered video to me. Not actual gameplay.

that would be one of the worst prerenders i ever saw

but it was too many short flashes in a low quality video to really see dx11 effects

[Mad1723]

that would be one of the worst prerenders i ever saw

but it was too many short flashes in a low quality video to really see dx11 effects

if you compare with the actual gameplay footage we have up to now, I'd say the video looks legit

[trinibwoy]

http://vimeo.com/6182068

r8xx in action dirt 2

Nice, glad to see AMD's marketing department finally grew some balls.

[Sly Fox]

if you compare with the actual gameplay footage we have up to now, I'd say the video looks legit

Agreed, GRID and DiRT are already beautiful games on the proper hardware (which luckily only really needs a GTX 260 or 4870), no reason this sequel can't look even better.

[Talonman]

That is a good video...

[MirageSys]

Nice, glad to see AMD's marketing department finally grew some balls.

And teeth to go along with them. Good thing they aren't foolish enough to furnish a can of whoopass though.

[zerazax]

Forget all the DX11 haters... I can't wait to see DiRT2 in DX11 !!

[LordEC911]

Stolen from neliz- (who really wants to reg at chiphell?)





4770 "reference" cooler being put to good use.
Estimated to be between a 4870/4890 performance for a "current" 4850 price.

[zalbard]

VRam chips on both sides... Does it mean it's 2gb? Or is the PCB just too short?