Yea, but that doesn't mean that 5870X2 really is two full-spec 5870s... or that it is going to be released right now.
AMD: R8xx Speculation
- Thread starter Shtal
- Start date
Yea, but that doesn't mean that 5870X2 really is two full-spec 5870s... or that it is going to be released right now.
Well, our DX11 hardware already has DirectCompute on by default and so downlevel support for DirectCompute 10 / 10.1 is already there, hence Anand's test of NV's DirectDompute demo. A DirectCompute 10.1 driver, for some DX10.1 hadrware, has been available to developers for many months as well.
I know, but when are you planning to release this DX10.1 driver to the general public?
nVidia has already has its driver out for a while now (the last two official 190-releases).
Its scheduled, I don't know when for though. D3D software teams's focus over the last few months has obviously been on getting a compliant DX11 stack ready.
But it could be 5850 X2 rather than 5870 X2. After all, both would be Hemlock.
I hadn't seen that, it's pretty nice! I know the aim isn't the same, but it still looks much better than Larrabee's jelly/water reflections.
Strange thing is though... the screenshot shows it running at 152 fps on a GeForce GTX 280(?).
DirectCompute performance seems to be all over the place. The tessellation sample from the DX SDK seems to run slower on a GTX285 than on an 8800GTS in some cases.
Correct, but from what I've heard, the 5870X2 will have two full spec'ed Cypresses but at 5850 speed, hence 725Mhz. I agree with you though that 125Mhz is quite a large hit.
It remains to be seen though, what ATI has done for the communication/scaling of the two chips. Since I heard about the Hydra 100, i had hoped that a dual GPU card would come, using some Lucid technology instead of the PLX/NF200 chips.
I'am pretty excited with Anandtech's article regarding the launch of the Hydra 200 chip next month and i wonder if there could be any chance that the 5870X2 could use one such chip for the communication of the Cypresses. If I am not gravely mistaken, i think both Intel and AMD are shareholders of Lucid!
Silent_Buddha
Legend
That would make the card quite expensive, as currently the price for dual 8 lane PCIE Hydra is 40+ USD (I think 48 was the number being tossed around) and the dual 16 lane PCIE Hydra was 70+ USD (I think 72 or 76 was the number being tossed around).
I'm not sure you'd be able to make a card using either of those for under 600 USD for 5850x2 or 800 USD for 5870x2.
Regards,
SB
Done first Vantage test:
http://service.futuremark.com/compare?3dmv=1438780
GT1 has exactly same FPS as my previous HD4870CF @850/1000, but GT2 is slower by 5FPS!
Anyone wishes particular tests? I'm happy to help tonight! :smile:
http://service.futuremark.com/compare?3dmv=1438780
GT1 has exactly same FPS as my previous HD4870CF @850/1000, but GT2 is slower by 5FPS!
Anyone wishes particular tests? I'm happy to help tonight! :smile:
Congrats! I wish you to have a blast time with it!
Crysis and/or Crysis Warhead benchmarks with their respective tools please. The standard gpu test for Crysis and Ambush benchmark for Warhead.
DX10 Very high/Enthusiast and DX10 High/Gamer at 1920X1080/1200. And for the love of God, NOAA please!
Try the oldie ArchMark.
Code:
[b][url=http://www.zeckensack.de/archmark/]ArchMark 0.50[/url][/b]
Driver ATI Radeon HD 5800 Series v2.1.9016
Resolution 1024x768 @ 47.86Hz
Comment Compliled by Lightman
Method Flush
[b]Fillrate[/b]
--[b]32 bits[/b]---------------------------------------
Mode R8G8B8A8 Z24 S8
Col 26.104 GPix/s
Z 77.160 GPix/s
ColZ 26.034 GPix/s
ZPassColZ 17.797 GPix/s
ZCullLEqual 146.801 GPix/s
ZCullGEqual 152.478 GPix/s
ZCullEqual 151.533 GPix/s
S 157.017 GPix/s
SCull 157.039 GPix/s
----[b]stencil test passed[/b]-------------------------
S 70.088 GPix/s
ZFailS 49.776 GPix/s
------[b]z test passed (LEQUAL)[/b]--------------------
S 69.900 GPix/s
ZS 47.025 GPix/s
Col 26.062 GPix/s
ColZ 20.608 GPix/s
ColS 19.585 GPix/s
ColZS 16.726 GPix/s
--[b]16 bits[/b]---------------------------------------
Error no pixel format
[b]Bandwidth[/b]
Mode R8G8B8A8 Z24 S8
--[b]available to buffer clears[/b]--------------------
All 247.946 GB/s
Color 109.708 GB/s
ZAndStencil 135.348 GB/s
Z 101.060 GB/s
Stencil 29.733 GB/s
Draw 171.889 GB/s
BurnedByRAMDAC 150.566 MB/s
Physical 172.039 GB/s
[b]Geometry[/b]
Error no pixel format
[b]Texturing[/b]
Error no pixel format
[b]Readback[/b]
Mode R8G8B8A8 Z24 S8
--[b]Whole buffer[/b]----------------------------------
R8G8B8A8 538.574 MPix/s
B8G8R8A8 747.988 MPix/s
R8G8B8 429.639 MPix/s
B8G8R8 19.770 MPix/s
Zuint 57.713 MPix/s
Zfloat 619.303 MPix/s
S8 437.411 MPix/s
--[b]32x32 region[/b]----------------------------------
R8G8B8A8 8.385 MPix/s
B8G8R8A8 8.474 MPix/s
R8G8B8 6.269 MPix/s
B8G8R8 2.476 MPix/s
Zuint 4.338 MPix/s
Zfloat 7.122 MPix/s
S8 6.852 MPix/s
[b]Texture cache[/b]
Mode R8G8B8A8 Z24 S8
RGBA 128 kiB
[b]Tiling[/b]
Mode R8G8B8A8 Z24 S8
--[b]preferred block alignment[/b]---------------------
----[b]updating all buffers[/b]------------------------
Width 16
Height 2
----[b]in color buffer[/b]-----------------------------
Width 4
Height 2
----[b]in depth buffer[/b]-----------------------------
Width 8
Height 8
----[b]in stencil buffer[/b]---------------------------
Width 64
Height 16Vista x64 and Phenom II 940 3.5GHz. Card at default 850/1200
Crysis 1920x1200 VH DX10 64bit noAA
Compared to my CF4870 512MB this is a blast!
Crysis 1920x1200 H DX10 64bit noAA
now I will try to bench Warehead but how to do it from Steam?
Last edited by a moderator:
Am I the only person who noticed, that the RGSS modes cause unsharp edges (silhouettes)? In theory the edges should be the same as with MS (or only slighly different - I'm not sure, if the SS mode supports per-sample gamma correction), but the difference is really huge - results are far from the RGSS of VSA-100...
Similar threads
