AMD: R9xx Speculation

Maybe you want to check the differences in AA? And btw, the ones I linked before had 4xAA at 1920x1200.

edit - checking the german guys, not sure he ran his with 4xAA though. It doesn't show if you have run it with AA on the final score?

 

No, I didn't, but thank you to the link.

Well... I was being ironic...
Speaking of the best case for AMD, "overshaded" looks an excuse to not say "too much geometry", "overshaded" pixels may hurt nVidia much more than it hurts AMD.

 

oh are you still whining ? i gave you same settings as you want from a respected reviewers.. you are hopeless damn..

i dont have faith but you can see the difference look where AA is written ?

 

I think comparing numbers in tessmark should bring better conclusions about tesslation power in Barts, compared to Cypress and Fermi: http://www.gpu-tech.org/content.php...sellation-punch-does-the-Geforce-GTS-450-have

I hope Carsten does an update on that.

 

I hope the author will update his tessmark that you can change the input mesh and the tessellationfactor.
Now he is using four different settings with different input values.

 

I honestly don't know what that means .
It could mean that there will be improved AF and MLAA in CCC for the 5000 series.
OR
No improved AF and MLAA in CCC for the 5000 series .

Oh well thanks for replying though. We will find out soon enough.

Interesting response, thanks!

 

I sure as hell hope they aren't doing tessellation with the GS. It's pretty easy to do tessellation with a geometry shader on any DX10 hardware, but the parallelization creates a huge buffer requirement because you're starting with a bunch of tessellation primitives (I assume 64) and then as you go through the GS code you generate one triangle from each primitive simultaneously.

Good fixed-function tessellation would take one primitive as input and generate all of its tessellation coordinates triangles before moving to the next primitive. You only need to buffer these before going into the domain shader, which should require far less space per wavefront than a normal vertex shader.

 

And "too many polygon edges for MSAA" also tells us that the MSAA compression algorithm is no longer compressing, I guess (it's adaptive compression).

Why do you say that overshaded pixels would hurt NVidia more?

 

What the hell are you on about "whining"? We've got a bunch of different benchmarks, with 1ghz+ clocks and different levels of AF and AA. Legitreviews for some unknown reason changing AA levels between reviews, it's not exactly hard to make a mistake, ie seeing the 4 anisotropy as 4xAA? Especially with Hexus remarking "Here's the standard benchmark run at 1,920x1,080 with 4x AA and 16x AF."

Instead of spending 2 pages being a total dickhead, you couldn't just have pointed out the differences in AA? Jesus.

 

I wasn't suggesting that GS was used for tessellation. Merely that GS is "always on" if TS is active. But in light of what you say next, plus a smidgen of thinking time, I think that's irrelevant to TS, per se.

I think the buffering would only need to be to the extent of being able to fully populate a hardware thread.

I haven't studied the document, but this "small buffer" for post-TS data might correspond with the use of a small buffer that's being proposed for post-GS data. i.e. the same buffer suits either.

To be honest I can't see why LDS can't be used as this "small buffer", for the naked vertices produced by TS. Similar to the way that vertex attributes are buffered in LDS for consumption by interpolation instructions during pixel shading (and, as you say, taking less space).

I think the "small buffer" approach was the crux of NVidia's design for GS buffering, using shared memory (I reckon) to capture a batch of data from a hardware thread, before optionally submitting for setup or writing the triangles to memory (stream out). In Fermi shared memory is also known as L1 cache and because it's integrated with the L2 cache there's a fairly robust architecture in place to deal with the clumps of data that TS can produce, and to deal with problem of delivering vertices/triangles to the screen-space tiles for rasterisation, as the tiles are spread across all the SIMDs.

 

Gee, I dunno. What about lower price?

 

Holy cow! Looks like Nv will be cutting prices for the 470 and the 460 to fight Barts. [H] is reporting that those 2 will face big cuts soon. So will Nv bring the 470 price down to fight the 6870, and then let the 460 1Gb fight the 6850? Gotta say wow to a company willing to fight a 255mm chip with a 530mm chip.

 

It's not that they're willing, it's just they have no other choice.

 

That's quite interesting. I wonder if this will trigger a full blown price war…

 

6870 isnt a replacement for 5870, Cayman will be. Therefore AMD effectively moved mid-high end cards to lower-mid level. Therefore it is an improvement

 

Lost Planet 2 benchmark would be a nice test too. It runs much better on my 460 than on my 5850.

Game consoles are way ahead of AMD in that mate.

 

So in your opinion asking people not to use for example under 2x2 pixel triangles, when no card can show any difference between 2x2 pixel triangle and smaller ones, somehow limiting the devs

 

Looks like it! http://www.fudzilla.com/graphics/item/20573-gtx-470-available-for-€207-in-dach

GTX 470 for €190 in Holland… nuts!

 

http://www.techpowerup.com/133109/N...60-and-GTX-470-to-Counter-HD-6800-Threat.html

edit:
Gotta love Fuad's logic

Err, cutting the prices of your current lineup indicates exactly the opposite - that you don't have anything new to counter the HD6800's, so you lower your current prices

 

Only the Gainward GTX 470 so far. Other brands are still at €240+.