leoneazzurro
Regular
There are profiles in CCC, so you can create a profile with AI on and another with AI off.
LucidLogix Hydra, madness?
- Thread starter Citrous
- Start date
leoneazzurro
Regular
Not really. There are probably some optimizations that are not accessible by the CCC and are always on, but there is an "AI Off" option that turns off most of the standard optimizations. Moreover you can choose the "standard" or "Advanced" AI levels. But this is OT, I think.
DegustatoR
Legend
So how's that helping me to get AA working in Oblivion without texture filtering optimisations? Or get rid of filtering optimisations without loosing half of performance of my 4870X2?
You're experiencing massive Texture issues?
DegustatoR
Legend
I do. Anisotropic filtering quality on my 5870 even with AI Off sucks compared to my GTX280 in HQ mode. Which is kinda strange considering that it's supposed to be the best on the market, no?
leoneazzurro
Regular
I have a 4850 Mobility with Cat 9.9, I'm playing Oblivion in these days and I don't experience massive issues (nor minor issues, for what's worth). And a 4850 Mobility is more than enough to play Oblivion smoothly at 1680*1050 (my notebook screen resolution) with AA 8x.
So, could you please stop going OT and write in the appropriate thread if you are facing issues?
You own both a X2 and a 5800 series GPU? Oo
Don't forget his X1950XTX
DegustatoR
Legend
I still has my 3870X2 too, yes. Why, you don't believe me?
Performance preview:
http://pcper.com/article.php?aid=815&type=expert&pid=1
http://pcper.com/article.php?aid=815&type=expert&pid=1
Seamless scaling of ocl kernels on a quad 5970?
Ok, prolly not what I am thinking. It seems that they are distributing different draw calls to different gpu's and them compositing them. I wonder what they have in mind for scaling ocl, dxcs. But yeah, this could be wonderful if integrated right on cpu's. More likely, it'll go on intel only cpu's which will trash your performance (but not kill right away like nv) if you use green/red chips.
Seamless scaling of ocl kernels on a quad 5970?
Ok, prolly not what I am thinking. It seems that they are distributing different draw calls to different gpu's and them compositing them. I wonder what they have in mind for scaling ocl, dxcs. But yeah, this could be wonderful if integrated right on cpu's. More likely, it'll go on intel only cpu's which will trash your performance (but not kill right away like nv) if you use green/red chips.
I don’t think that they will split single process calls.
Who say no AFR? I don’t have seen any proof that they can manage modern games by distribute a single frame to multiple GPUs and merge the results.
I know, but let’s take a look at the Direct3D Compute shader interface. The process call is Dispatch which takes the dimension of the group cube. A shader can get the position in the group cube as parameter. If you want to split the cube to multiple chips you will run into the problem that at least one chip get’s wrong positions. You need to add additional instructions to the shader. This isn’t that easy as DirectX shaders are signed since DX10. But this is the easier problem. Shader Model 5 allows random read write access to the memory. Currently I cannot see how shaders that make use of this can be safe split to multiple GPUs without explicit hardware support. And even in this case any third party solution can only call functions that are supported by public accessible APIs.
I dont see much advantage at all. The only real advantage over SLI/Crossfire is the ability to keep your old card for added performance.
But even that is dubious, as typically, for example in my case, I ebay my old card upon buying a new one. So it would probably be a hypothetical question like, keep my old card with Hydra and get 30-50% more performance, or sell it and buy a 30-50% better new card in the first place with the proceeds? The second option would obviously have huge benefits in power consumption, and doesn't require a motherboard with multiple PCI-E slots either (as my current motherboard is).
There's also the issue of DX revisions, I assume pairing a DX 11 card with a DX 10 card would downgrade the whole setup to DX 10? Again major downgrade, and would severely limit the appeal of pairing older cards with new.
Would no AFR mean no SLI/CF input lag though? That would be nice, but I doubt many people would actually care.