It looks like NVIDIA will be replacing all meat grill as we have been announcing for months. Although this is also a lie, for now it all points to the fact that Huang, led by Samsung, will hunt down AMD which is likely to share the Greens’ enthusiasm for the type of memory used.
Exclusive: Features of NVIDIA Ampere RTX 3090, RTX 3080 and RTX 3070 release dates, price, specs.
Ampere can come with two new types of VRAM
Rumors point to two types of VRAM depending on the chip used, as was the case with Pascal and GDDR5 vs GDDR5X at the time. Although there is no official data from any NAND Flash developer for this new memory type, i GDDR6X It would not be a great plan if the industry was ready for a new speed depending on the new lithographic systems currently available, led by South Koreans as well.
Once again, last month we gave the latest rumors that NVIDIA might have 18 Gbps chips in GDDR6 and that these will come with Ampere, not surprisingly, we’ve already seen the first RTX 2080 Ti GPU in the form of 16 Gbps and NVIDIA introduced the RTX 2080 SUPER is GDDR6 at 15.5 Gbps.
The logical step is not only to increase the frequency and improve the build, but also to combine this with a new type of memory that provides greater bandwidth without compromising the voltage and hence the usage.
GDDR6X, the big unknown with NVIDIA’s hope to do it without HBM2E
Despite rumors, it is alleged that AMD could put some HBM-type memory into it Navi 2x low is RDNA 2, which will also reflect the higher cost of the company but the benefit of having greater bandwidth (nothing useful in gaming systems as we’ve been seeing since the introduction of HBM on the market, Game Cache in between).
If all the information is true and NVIDIA managed to overcome the 18 Gbps barrier to 20 Gbps by Samsung, we’re talking 2,500 MHz traditionally and under the 384-bit bus will appear 960 GB / s, something unimaginable just two years ago when the Train entered the market.
At the moment it is assumed (if this is true) that NVIDIA would have to diversify its range, when it would begin to understand the concept’s implementation. GA103, a completely new chip in terms of width is affected.
Maybe it will be split with GA102 and GA103 for GDDR6X and more for GDDR6 at 18 Gbps? Yes, it may not be conducive to the company’s move, but as is often the case with this type of rumor, let’s be cautious and touch them with tweezers, the reality can be as it is or very different, only time will tell.
