Linkedin profiles of the experts in hardware can often reveal very interesting information. And 'the case of this profile, made public by 3DCenter site and in the meantime removed, an AMD R & D Manager in Shanghai, indicating that it operated within the project indicated by the code name of "Greenland".
With this name indicates AMD internally of the next-generation GPU architecture positioned in the top segment of the market range so we speak of the chip known as Vega 10, as indicated in the latest of AMD GPU roadmap shown by ' American company to coincide with the GDC in San Francisco in recent weeks.
In Linkedin profile indicates that the Greenland GPU integrates within itself up to 4,096 shader processors: this implies a given identical to the integrated GPU from AMD in Fiji used in Radeon cards of the Fury family. Changes little, then? An identical number says nothing of architecture at the base, and then the performance capabilities: AMD plan is already confirmed as the Vega solutions will use 14-nanometer production technology FinFET type allowing complexity levels well above those of 28-nanometer technology adopted for Fiji solutions.
AMD has also remarked as with Polaris solutions will be able to introduce significant improvements in terms of overall efficiency, thus achieving higher performance by balancing the overall consumption. With Vega 10 AMD will use the second generation of HBM or High Bandwidth Memory memory, able to offer a level of twice the theoretical maximum bandwidth than the first generation HBM memory coupled to the family Fiji GPU as well as increase the total amount of memory for each GPU.
With this name indicates AMD internally of the next-generation GPU architecture positioned in the top segment of the market range so we speak of the chip known as Vega 10, as indicated in the latest of AMD GPU roadmap shown by ' American company to coincide with the GDC in San Francisco in recent weeks.
In Linkedin profile indicates that the Greenland GPU integrates within itself up to 4,096 shader processors: this implies a given identical to the integrated GPU from AMD in Fiji used in Radeon cards of the Fury family. Changes little, then? An identical number says nothing of architecture at the base, and then the performance capabilities: AMD plan is already confirmed as the Vega solutions will use 14-nanometer production technology FinFET type allowing complexity levels well above those of 28-nanometer technology adopted for Fiji solutions.
AMD has also remarked as with Polaris solutions will be able to introduce significant improvements in terms of overall efficiency, thus achieving higher performance by balancing the overall consumption. With Vega 10 AMD will use the second generation of HBM or High Bandwidth Memory memory, able to offer a level of twice the theoretical maximum bandwidth than the first generation HBM memory coupled to the family Fiji GPU as well as increase the total amount of memory for each GPU.
Comments
Post a Comment