Originally Posted by renethx
The point seems to be not the CPU but the HyperTransport bus
. As you know, the memory controller is integrated in the CPU
and the gaphics core communitcates with the system memory via the HT bus (thus implementing IGP is harder in the AMD platform than the Intel platform where memory controller is located in the same chipset). Higher bandwidth of HT 3.0 helps a lot in 3D performance (communication between shader processors and memory; 110% increase, that's not a mere speculation but the fact!), then could help deinterlacing.
If it's the HT bus that is important, the B2 stepping and BIOS update should be fine for these specific purposes.
There are about two different things going on with my comments that aren't necessarily related and I think have you confused about what I've been saying.
1. My comments about the Phenom aren't related to whether it helps the 3D graphics performance, clearly it does. And the increased HyperTransport bandwidth may indeed help with deinterlacing. My hesitation isn't about the performance increase. It's about buying a CPU we know to be flawed. Yes with the TLB patch off in the BIOS it's fine, still I would be cautious and wait an extra couple of months for the B3 stepping. I would tend to agree that just getting the 9100e is fine, however consider the time frame of when the 9100e comes out, which is supposed to be around April, and then sometime in summer AMD will get the B3 steppings out, so you basically gave into impulse and bought the old stepping within a few months of a new stepping. Not the end of the world, but just a recommendation of caution if you can wait to build your HTPC.
2. Again, I think you're confused about what I'm saying. The 3D Mark scores do not necessarily correlate to improved video anything. You'll recall that the VC-1 and H.264 decoding are done in dedicated silicon, as is the noise reduction, the designs are borrowed from ATI's Xilleon line of DTV chips. What is done in the shaders is deinterlacing and MPEG2 decode. So yes those areas may improve, but we don't have any tests of this yet, all we see is it failing to score well in the HD HQV tests (which are compressed in VC-1, so it doesn't test the MPEG2 shader code, only the deinterlacing portion). We need the same test bench with an Athlon 64 X2 and then a Phenom swapped in.
Now the post-processing is only what the video drivers expose to the video decoder filter. If the video drivers don't even expose the ability to choose ATI's advanced deinterlacing then it will fail, regardless of the CPU. I posted earlier on why this could be happening, I'm not saying this is the reality of it, but the tactic has been done before by ATI and NVIDIA, when they felt their low-end parts didn't have the speed for a type of deinterlacing or post-processing they simply disabled it in the drivers. Now hopefully ATI has done a load-balancing type of code in their drivers for the deinterlacing and it properly recognizes that you have a beefy CPU which can help bear the brunt of advanced deinterlacing.