Nvidia updated its GTX 1050 3GB graphics card this week. This dedication to the Pascal architecture also proves that we are still quite far away from the next generation of products.
In fact, the latest information that I learned from AIC is that they can get the GPU from NV in August. That is, the fastest version of the public version is also the end of August and the appearance of September. The non-public version will be later.
A key variable that has been delayed for so long is the GDDR6 memory. Although SK hynix, Micron, and Samsung all claim to get it, they can only achieve stable supply in the summer.
According to Appuals, The GDDR6 supply agreement between SK hynix and NVIDIA resulted in an increase of 6% in SK Hynix's share price and reached the highest level in 17 years. The first batch was a single-stack 8Gb version with speed ranging from 10Gbps to 14Gbps. As for the final choice, we must look at the NV debug results.
Compared to GDDR5, GDDR6, in addition to double the capacity / pin bandwidth, the voltage is also reduced from 1.5V to 1.35V (or 1.25V), which is more power. However, the performance increase achieved after the conversion is 40%.
Somewhat unintelligible is that This new card using GDDR6 memory is based on the Volta architecture. Instead of Ampere (Ampere) and Turing (Turing).
Frankly speaking, NV should be psychologically prepared to squeeze toothpaste in 2018. One aspect that can be verified is AMD's 'dummy fire'. 2018 does not have a high-end card at all. Navi will only be on a large scale next year, and AMD may not fail to do 'competitive intelligence'. analysis.
