Based on Radeon Vega graphics, the AMD Vega 10 GPU, which has demonstrated at many levels including the CES in which it has been pictured in a video published by LinusTechTips. The upcoming technology was seen running a DOOM at 4K resolution alongside Ryzen CPU.
AMD Vega 10 GPU Pictured, Based on Radeon Vega Graphics.
The Graphic Card which is pictured is still in an engineering state as the product itself is months away from the debut for consumers. Nevertheless, there are many of the interesting things that Raja Koduri and Linus talks about in the video that can be seen below
To start with up, we realize that AMD Vega 10 design cards will be accessible in shopper, workstation and server variations. The customer variations will be marked Radeon while the expert cards include the new FirePro WX and Instinct marking. The designing example which was running at CES ’17 used a more drawn out rendition of the cooler cover included on Radeon RX 480 and it searches truly slick for the illustrations card itself. To prevent the press from sneaking their cameras inside the PC case, all its events covered up in taking the pictures of the Vega Samples and Ryzen. Nevertheless, Linus managed to get special permission from AMD and took off the side panel to leak out an interesting Vega 10 based Radeon Graphic Card.
Board Speculations, AMD Vega 10 Based on Radeon Vega Sample Graphics.
There are other things to note down that the PCB on the back remains uncovered. Having a dual slot body, the card utilizes a blower style cooler. The logo on the side has a red LED which lights up when the card is running. The assembly of the cooling has various parts, above the GPU, the heatsink block is attached and held in a place through a metal bracket which is easily viewed on the back. The winding sheet is held by various screws that are located on the PCB. Likewise, the PCB doesn’t has any follows for GDDR5 memory as all the memory is hung on the interposer itself since it’s HBM2 based.
The power associations are still a mystery as AMD chose to put tape on the info ports. The reason they did this is on the grounds that the Vega test board is not characteristic of the last GPU and representation board. So it’s there to stay away from any perplexity amongst test and retail sheets. AMD likewise said that they might want their AIB accomplices to get ready little ITX sheets in light of Vega so it’s feasible we may not see a R9 Nano-esque item from AMD themselves in the coming Radeon Vega lineup. Another fascinating thing to note is that there’s a moment PCB toward the finish of the card which includes a USB port.
This port won’t be incorporated into the buyer adaptation and Raja Koduri gave a really decent clarification of the motivation behind this port. The port gives engineers at AMD the capacity to screen every one of the signs, in particular power and electrical signs leaving this GPU, remotely and with no overhead. As of now, all representation cards with the VEGA 10 GPUs running at AMD HQ highlight this port so they can decide the quality (execution per watt) of the last item.
Specifications of AMD Vega 10
To consumers, AMD’s high end Vega 10 GPU will be available soon in the first half of 2017.
A die size of over 500mm2 chip spans from early calculations and features a HBM2 stacks. Assimilating up to 16GB of HBM2. the variants of consumers which was then demonstrated using the Star wars and DOOM featured 8GB of HBM2 VRAM. for the consumer variant the specific device ID is 687F:C1.
The representation chip will use the most recent 14nm GFX9 center engineering which depends on the NCU (Next Compute Engine) plan. The representation card will include 64 Compute Units or 4096 stream processors. AMD anticipates expanding the throughput of the chip through expanded clock speeds. This will permit AMD to pump numbers superior to the Fiji GPU which depends on 28nm GCN 3.0 engineering and accompanies a similar number of centers, e.g. 4096 SPs. The server part with the full chip is required to highlight a TDP of 225W with clock speeds around 1526 MHz. A shopper situated representation card can highlight much higher clock speeds since server parts for the most part do not have the cooling abilities of purchaser level cards which transport with better coolers and PCBs intended to permit overclocking of the GPUs. You can read our Vega GPU design see here.
Official Architecture Slides of AMD Radeon Vega
Vega 11 GPUs and AMD Vega 10
Specifications of AMD Vega 10 Memory:
Such as the Radeon R9 Fury X was limited to just 4GB of VRAM, the first generation HBM graphic card which has a bandwidth of 512 GB/s. it consist of a 4- layers per stack which will continue with the latest Vega GPUs since the AMD will have to maximize value on these cards for the audience of gaming. In case of 4 layers we might be looking for a higher density per layer. The speed also increases the pin with HBM2. the new memory standard can clock up to 2 GB compared to 1 GB on HBM1. the increased clock speed allows the same memory bandwidth as four HBM1 stacks on just the two HMB2 stacks.
The increased density also helps in AMD to decrease the cost in designing larger interposers. HBM2 it self takes more n more space in comparison with HBM 1 with a die size around 92 mm 2 while the HBM1 has just 35mm2 in size.
NVIDIA Pascal GP100 and AMD vega 10 GPU Specification Comparison.