AMD’s Beijing Technology Summit on Vega has been posted online

AMD's Beijing Technology Summit on Vega has been posted online

AMD’s Beijing Technology Summit on Vega has been posted online

 

A video of AMD’s Beijing Technology Summit on Vega has been posted online, stating that the RX Vega is “just around the corner” and that it will come with various different memory configurations.

AMD’s Vice-president, Scott Herkelman, has stated that Vega GPUs will be available with different memory configurations, with either 4GB or 8GB HBM2 memory stacks. This does not mean that AMD will be releasing chips with either single or dual HBM memory chips (as that would have a huge effect on memory bandwidth), but that they will be using either dual 4GB or 8GB (perhaps 2GB or 4GB) HBM2 memory stacks to create 8GB or 16GB GPUs. 

The wording from Herkelman is not entirely clear here, as it is unclear if 4GB and 8GB is referring to final memory configurations or the capacity of the HBM2 memory stacks used. We do know that 8GB HBM2 memory stacks have been on Samsung and SKHynix company roadmaps for some time, but that doesn’t mean that AMD will be using them. 

While a 4GB GPU could be created with dual 2GB HBM2 memory stacks, we don’t see AMD creating such a high-end GPU with such a small amount of VRAM, especially when considering that Nvidia has recently released a GPU with an 11GB frame buffer.  

 

Vega will use HBM2 that has different capacity stacks

you will see from our board partners different configurations, whether that’s 4 Gig or 8 Gig or those types of memory architectures that will allow you to drive different games and different resolutions based upon what capacity stack they end up using

 

(Video from Videocardz)  

 

Apart what was mentioned above, the presentation does not reveal much new information about the RX Vega, again listing the features of the Vega architecture like Rapid Packed Math, AMD’s new memory cache controller and their other architectural improvements. 

 

You can join the discussion on AMD’s RX Vega GPU on the OC3D Forums. 

 

AMD's Beijing Technology Summit on Vega has been posted online

AMD’s Beijing Technology Summit on Vega has been posted online

 

A video of AMD’s Beijing Technology Summit on Vega has been posted online, stating that the RX Vega is “just around the corner” and that it will come with various different memory configurations.

AMD’s Vice-president, Scott Herkelman, has stated that Vega GPUs will be available with different memory configurations, with either 4GB or 8GB HBM2 memory stacks. This does not mean that AMD will be releasing chips with either single or dual HBM memory chips (as that would have a huge effect on memory bandwidth), but that they will be using either dual 4GB or 8GB (perhaps 2GB or 4GB) HBM2 memory stacks to create 8GB or 16GB GPUs. 

The wording from Herkelman is not entirely clear here, as it is unclear if 4GB and 8GB is referring to final memory configurations or the capacity of the HBM2 memory stacks used. We do know that 8GB HBM2 memory stacks have been on Samsung and SKHynix company roadmaps for some time, but that doesn’t mean that AMD will be using them. 

While a 4GB GPU could be created with dual 2GB HBM2 memory stacks, we don’t see AMD creating such a high-end GPU with such a small amount of VRAM, especially when considering that Nvidia has recently released a GPU with an 11GB frame buffer.  

 

Vega will use HBM2 that has different capacity stacks

you will see from our board partners different configurations, whether that’s 4 Gig or 8 Gig or those types of memory architectures that will allow you to drive different games and different resolutions based upon what capacity stack they end up using

 

(Video from Videocardz)  

 

Apart what was mentioned above, the presentation does not reveal much new information about the RX Vega, again listing the features of the Vega architecture like Rapid Packed Math, AMD’s new memory cache controller and their other architectural improvements. 

 

You can join the discussion on AMD’s RX Vega GPU on the OC3D Forums.