thanks for the informative input technical expert of GPUs... LOL

LOL, a year too late and they still try spinning it like a superior gaming product. Sorry, but they have lost this round a year ago. HBM2 was a wrong bet.

While all of these GPUs are focused on the same application set, they cut across multiple architectures. The MI25 is built on the new “Vega” architecture, while the MI8 and MI6 are based on the older “Fuji” and “Polaris” platforms, respectively. The top-of-the-line MI25 is built for large-scale training and inferencing applications, while the MI8 and MI6 devices are geared mostly for inferencing. AMD says they are also suitable for HPC workloads, but the lower precision limits the application set principally to some seismic and genomics codes. According to an unnamed source manning the AMD booth at ISC, they are planning to deliver 64-bit-capable Radeon GPUs in the next go-around, presumably to serve a broader array of HPC applications. For comparison’s sake, NVIDIA’s P100 delivers 21.2 teraflops of FP16 and 10.6 teraflops of FP32. So from a raw flops perspective, the new MI25 compares rather favorably. However, once NVIDIA starts shipping the Volta-class V100 GPU later this year, its 120 teraflops delivered by the new Tensor Cores will blow that comparison out of the water. A major difference is that AMD is apparently building specialized accelerators for deep learning inference and training, as well as HPC applications, while NVIDIA has abandoned this approach with the Volta generation. The V100 is an all-in-one device that can be used across these three application buckets. It remains to be seen which approach will be preferred by users. The bigger difference is on the software side for GPU computing. AMD says it plans to keep everything in its deep learning/HPC stack as open source. That starts with the Radeon Open Compute platform, aka ROCm. It includes things such as GPU drivers, a C/C++ compilers for heterogeneous computing, and the HIP CUDA conversion tool. OpenCl and Python are also supported. New to ROCm is MIOpen, a GPU-accelerated library that encompasses a broad array of deep learning functions. AMD plans to add support for Caffe, TensorFlow and Torch in the near future. Although everything here is open source, the breadth of support and functionality is a fraction of what is currently available to CUDA users. As a consequence, the chipmaker has its work cut out for it to capture deep learning customers.

Agreed, this should have been out before Christmas. I'm not sure who is willing to buy a product that sits between a year old 1070/1080 if we go by the FE results. the price difference between the two is not even that high anymore. unless they go and undercut the gtx 1070 price... HBM2 sounded lovely back when we still had plan GDDR5 now we got GDRR5x hitting 12ghz on some of the models, so the bandswidth gap is tiny

HBM was the worst thing AMD thought to bring to consumers.

Agreed, this should have been out before Christmas. I'm not sure who is willing to buy a product that sits between a year old 1070/1080 if we go by the FE results. the price difference between the two is not even that high anymore. unless they go and undercut the gtx 1070 price... HBM2 sounded lovely back when we still had plan GDDR5 now we got GDRR5x hitting 12ghz on some of the models, so the bandswidth gap is tiny

Nvidia has their own kind of Infinity Fabric coming soon, and Intel will surely follow suit if they want to stay relevant. http://research.nvidia.com/publication/2017-06_MCM-GPU%3A-Multi-Chip-Module-GPUs The PDF is an interesting read for anyone who wants it. VEGA will need lots of optimization and at launch will not be 100% ready. Will likely take 6 months for the drivers to mature to the point that it can be close to 1080 Ti in some small cases, and by then Volta will arrive.