Originally Posted by Lee Patekar
They'll do like intel. You'll have overpriced gaming cards with little gains like the desktop quad cores and then you'll have the high end overpriced cards for 1000$ like the extreme edition CPUs, and then you'll have your workstation / server GPUs for deep learning that will be very expensive, like Xeons.
If the CPU division is profitable and the graphics division seems too far gone, we may see them stabilize the company by concentrating on CPU sales and scaling back the graphics department to smaller offerings.. like the Polaris launch.
Unlike Intel Nvidia still has room to increase performance. There are several node shrinks incoming. From 16nm to 12nm. Then first to 10nm and/or 7nm. The numbers may differ depending on the marketing department of the foundry. The only reason Nvidia had to repeat the Intel scheme would be, if they have a unpresented lead like back in teh 8800 days. After the big hit they had years of refreshes and even some rebrands. Yet they can always aim higher. First get 4k resolution undner control with the mainstream, then go over 5k with the final goal of 8k super-duper-high resolution. This is the path they probably will take. In here no regard to upcoming memory technologies.
In my mind it is still questionable if AMD can keep up with all those needed development. If they only release entire new cards every 2 years, they may fall behind. You can not think of everything and be lacklaster in exactly those areas.
Originally Posted by Lee Patekar
Its a real possibility and, as a consumer wanting competition for better prices, I hope it doesn't happen.
As the years with Intel having no competition showed, the customers have to bite the sour apple or don't buy at all. People don't need new hardware every years like it was used to be. If your scenario happens, the customers only option is to stay on the same hardware longer. Something AMD users are practically used to nowadays.
Originally Posted by New green
"With Vega 10 clocked at 1.5GHz, we could expect a monstrous 12.5 TFLOPs of FP32 compute performance, and the high-speed 8GB of HBM2 with what I think will be the start of the show in High Bandwidth Cache (HBC), AMD could have one of the fastest graphics cards on the market with its Radeon RX Vega.
RX Vega from $399-$1000
AMD Radeon RX Vega (GTX 1070) - 4GB HBM2, cut down Vega 11 GPU
AMD Radeon RX Vega (GTX 1080) - 4GB HBM2, cut down Vega 10 GPU
AMD Radeon RX Vega (GTX 1080 11Gbps) - 8GB HBM2, cut down Vega 10 GPU (slightly higher clocks)
AMD Radeon RX Vega (GTX 1080 Ti) - 8GB HBM2, full Vega 10 GPU
AMD Radeon RX Vega (TITAN Xp) - 16GB HBM2, dual full Vega 10 GPUs"
What things could look like.
Who in their right mind would buy a 4GB card in 2017?
HBM has proven to be doomed with low memory capacity. Even the HBCC crap will not save this fact. If memory is full, it is full. Period!
I also doubt Vega will have that many version. Maybe 2 +1, one big like Fury X - one cut down like Fury and a special edition like the old NANO was. If AMD really has so many editions, they will most likely have low availability too. They can't produce chips infinitely. The line-up is confusing none the less. Seeing the suggestion for the prices is laughabel too. having the same chip from 400 to 1000 bucks is too much. An additional 4Gb are not worth several hundret bucks more. The market would not fall for that scheme.