AMD's Radeon HD 6870 benchmarked? (updated more screens)

[RPGWiZaRD]

^ Yes I noticed it was 1920x1200 AA4x HD 6870 vs 1920x1080 noAA GTX 480 comparision at first but perhaps it wasn't targeted at me.

[Florinmocanu]

I couldn't find initially the right benches but mindfury pointed the Hexus review.. :P

[Calmatory]

Note that the resolution is 1920x1080 for GTX 480 and 1920x1200 for HD6870. However, the difference should be fairly inisgnificant IMO.

[RPGWiZaRD]

Well it should also be noted that the HD6870 test is used with almost 1GHz CPU overclock advantage (4.2GHz i7-930 vs stock clocked 3.2GHz i7-965, but turbo enabled in Nexus' test) but as what I've seen CPU won't play that much role in the extreme tesselation test anyway so I'd still say around 20~25% advantage in that test for HD6870.

[Florinmocanu]

well, the resolution difference is quite important. You can calculate the number of pixels each GPU has to render.

So, equal graphic settings (except resolution).

GTX 480 has 1920 x 1080 = 2073600 pixels = 30.5 fps

HD 6870 has 1920 x 1200 = 2304000 pixels = 36.6 fps

the resolution difference equals to 11.11% more pixels rendered for the 6870.

The 1 Ghz helps it a bit, but Unigine is not really CPU hungry, especially since it's built to show what tesselation is all about.

Anyway, the performance difference looks solid.

[Olivon]

[RPGWiZaRD]

Mind if I ask the source? Or is that your own setup with the GTX 480 and i7-950? So 24% average fps and 30% minimum fps advantage, talk about tesselation performance improvement. (However if that HD6870 bench would be fake and improvements are way smaller I wanna stab that person if true then he's succeeded in doing early very valuable pre-launch marketing )

[Olivon]

Mind if I ask the source? Or is that your own setup with the GTX 480 and i7-950? So 24% average fps and 30% minimum fps advantage, talk about tesselation performance improvement. (However if that HD6870 bench would be fake and improvements are way smaller I wanna stab that person if true then he's succeeded in doing early very valuable pre-launch marketing )

GTX score brought from a guy on the most famous french forum - Hardware.fr

[RPGWiZaRD]

GTX score brought from a guy on the most famous french forum - Hardware.fr

Thanks.

[Mechromancer]

Wow. If the 6000 series has that kind of tessellation performance, I will buy one as soon as they come out! Maybe, just maybe Stanford will have their ATI OpenCL client out by then so all my dreams can come true at once.

[Chumbucket843]

Wow. If the 6000 series has that kind of tessellation performance, I will buy one as soon as they come out! Maybe, just maybe Stanford will have their ATI OpenCL client out by then so all my dreams can come true at once.

you could always join gpugrid. they will have an ATi beta client out soon.

[saaya]

well, the resolution difference is quite important. You can calculate the number of pixels each GPU has to render.

So, equal graphic settings (except resolution).

GTX 480 has 1920 x 1080 = 2073600 pixels = 30.5 fps

HD 6870 has 1920 x 1200 = 2304000 pixels = 36.6 fps

the resolution difference equals to 11.11% more pixels rendered for the 6870.

pff come on, you know thats not how it works :P
especially in a tesselation geometry benchmarks pixels dont matter that much...

[Calmatory]

well, the resolution difference is quite important. You can calculate the number of pixels each GPU has to render.

So, equal graphic settings (except resolution).

GTX 480 has 1920 x 1080 = 2073600 pixels = 30.5 fps

HD 6870 has 1920 x 1200 = 2304000 pixels = 36.6 fps

the resolution difference equals to 11.11% more pixels rendered for the 6870.

The 1 Ghz helps it a bit, but Unigine is not really CPU hungry, especially since it's built to show what tesselation is all about.

Anyway, the performance difference looks solid.

Pixel/texel fillrate and AA/AF performance(which are per-pixel) has very little to do with tesselation. However, someone with a card with strong tesselation(GTX x) could test this.

[Dumo]

Be ready to cough up some $$$ guys

In a few weeks probly a lot of 480s and 5970s @ F/S section

[RaZz!]

well, the resolution difference is quite important. You can calculate the number of pixels each GPU has to render.

So, equal graphic settings (except resolution).

GTX 480 has 1920 x 1080 = 2073600 pixels = 30.5 fps

HD 6870 has 1920 x 1200 = 2304000 pixels = 36.6 fps

the resolution difference equals to 11.11% more pixels rendered for the 6870.

The 1 Ghz helps it a bit, but Unigine is not really CPU hungry, especially since it's built to show what tesselation is all about.

Anyway, the performance difference looks solid.

as i said in my previous post already, the difference is not only the amount of rendered pixels.
if unigine engine handles different aspect ratios correctly, it uses a higher fov value for the 1920x1080 resolution. so it would have less pixels, but a higher fov and therefore more visible objects etc in one single frame.

no idea what has the bigger impact on performance, more pixels or higher fov. so, yeah...

[Sn0wm@n]

how can displaying pixels not be related to the rendering power of that gpu ????? the bigger the resolution the the more it has to work + the triangles etc... it has to render on that screen .... refresh rate stresses the gpu so i think that florinocanu has a point .. but to what extent maybe someone could run a test on a fermi and a 5800 so we could see the difference ...

[Calmatory]

Bigger resolution doesn't mean more triangles to be processed, unless it affects FOV. It means more pixels to be processed though, so with AA and AF will play bigger role since they are processed per-pixel so there is more processing going on. Higher resolution also stressees TMUs and ROPs more, since they handle texture mapping and pixel handling.

[madcho]

I wonder how it's gonna looks like on 3x 1920x1050

[spursindonesia]

Uh oh, if these stuffs are true, IMHO nVidia gets screwed for the next 3-4 Q until 28 nm process node arrives.

EDIT: And why do i sense there would be tripled or quadrupled FUDs regarding ATi's bad driver, game support, etc, in the next few months or until there's a new rumour regarding nVidia's next gen card surfacing in the web.

[Elmy]

Hopefully they fix Eyefinity flickering with these new cards !

[ajaidev]

image

Excellent find Qlivon

[Manicdan]

Bigger resolution doesn't mean more triangles to be processed, unless it affects FOV.

since they are different aspects it can be worse or better, does 16:9 show more or less than 16:10 in unigine? this in theory could help the 6800 if less polygons are shown, but more pixels are used

[Sn0wm@n]

Bigger resolution doesn't mean more triangles to be processed, unless it affects FOV. It means more pixels to be processed though, so with AA and AF will play bigger role since they are processed per-pixel so there is more processing going on. Higher resolution also stressees TMUs and ROPs more, since they handle texture mapping and pixel handling.

thanks for the exaplaination .. so basicly the ati gpu was stressed more and it still pulled more frames in a benchmark that they had troubles with the 5K series .... so can we safely say that this isnt just a mild refresh bringing good gains????

[Calmatory]

thanks for the exaplaination .. so basicly the ati gpu was stressed more and it still pulled more frames in a benchmark that they had troubles with the 5K series .... so can we safely say that this isnt just a mild refresh bringing good gains????

Most probably yes. However, considering that we don't know the resolution impact for sure, and it is far from sure that these benches are actually valid, I'd hold my breath for now.

[Klarko]

Wow. If the 6000 series has that kind of tessellation performance, I will buy one as soon as they come out! Maybe, just maybe Stanford will have their ATI OpenCL client out by then so all my dreams can come true at once.

Exactly what I'm thinking! ATi needs to get in that game.