A few game recording tools support Quicksync for realtime converting. Performance impact is much less than via CPU and files much smaller than uncompressed formats. However Intel does not support Direct3D pipeline capturing natively. This would be useful especially under Windows 8 with headless iGPU support.I mean gaming with dedicated AMD/Nvidia and then recording via Quicksync. Nvidia's Shadowplay on Kepler via NVENC support d3d capture native and the overhead while game capture is lower than Quicksync.
Using hardware accelerated Inband Frame Readback (NvIFR) technology. NvIFR is low-latency GPU/DMA accelerated render target readback interface, which provides video capture applications an ability to capture frames from Direct3D pipeline with much lower performance penalty than any traditional Direct3D based render target readback implementations.
I totally agree, a big part of the the bottleneck when it comes to game recording is the capture part. Unfortunately, in the context of Intel Graphics, a direct frame buffer capture capability such as the one you describe from Nvidia is currently not available. Technically this can be achieved but it is currently not a feature exposed to developers. The feature has been discussed and may be come part of future Intel Graphics driver releases.
However, one thing to keep in mind here is that, even if Intel Graphics would support direct frame buffer capture it does not help in the case where a game is played on a discrete graphics card (e.g. Nvidia or AMD) since the buffer will first have to be physically copied from that device context/memory to the Intel Graphics context/memory before it can be encoded using Intel QuickSync technology. (thus in this case no benefit of a direct frame buffer capture feature)
Technology has advanced in past few years. Currently there are game recording software, like D3DGear (http://www.d3dgear.com), that supports both discrete GPU (AMd/Nvidia) and integrated GPU (Intel).
This means D3DGear game recording software can capture game scene from discrete graphic card, while using Intel iGPU to encoding into H.264 movie. This become the ideal recording solution because it fully and effectively utilize all system resource and avoid to make CPU or discrete GPU as bottleneck.