Intel is building entire Xeon lines with embedded GPUs for serving web imagery. And these Intel Xeon's with embedded GPUs don't need to transfer data to the co

Intel is building entire Xeon lines with embedded GPUs for serving web imagery. And these Intel Xeon's with embedded GPUs don't need to transfer data to the coprocessor, since they operate off of system memory, so transfer latencies are a non-issue.