[Hardware] NVIDIA keeps its Ampere GPUs for gaming and HPC, the CMPs are Turing!

[WilkerCSBD]

Yesterday the most anticipated GPU for any gamer with a pocket that shakes at the end of the month, the RTX 3060, was launched. And as a good card in its presentation, it came accompanied by some drivers that, as expected, gave support to this GPU and to two more models: the CMP 30HX and CMP 40HX, both exclusive graphics cards for mining.

But those drivers have revealed more than NVIDIA would want, since these mining cards will not be Ampere, but Turing!

On the one hand, NVIDIA surprised many with the announcement of a new range of graphics cards exclusively for the mining market. Although we already anticipated this at the time, now and after the release of the new driver, NVIDIA has quite interesting plans for these cards and the RTX that are yet to come, segmenting the ranges even more.

NVIDIA resurrects Turing once again and targets it for the mining market
CMP-HX-GPU

Still without official specifications and we already have on the table the first big news that can decongest the mining market: the CMP HX GPUs will not be based on Ampere as the main architecture, but on Turing, as if they were an RTX 2000.

The announcement is far from official, but what it is was the support for the first two models that, seen as seen, will hit the market sooner rather than later. These models are named 30 HX and 40HX, which will have Turing-based processors and being specific, the first will be based on the TU116 and the second on the TU106, so there will be interesting disparities in performance.

At the moment and in the absence of knowing its specifications, we can only count on what was revealed by NVIDIA in its announcement. And is that the CMP 30HX will have a Hash Rate for mining of 26 MH / s, a consumption of 125 watts to house a single 8-pin connector and 6 GB of VRAM.

Instead, the CMP 40HX goes up to 36 MH / s, 185 watts for the same 8-pin connector and raises its VRAM up to 8 GB. Both cards are expected for this Q1.

Why does NVIDIA choose Turing over Ampere? Will the CMPs be manufactured at 8 nm or 12 nm?
NVIDIA-CMP-HX-specs

After the shock of knowing that these GPUs will be based on the Turing architecture, the more important question remains, why? Well, there are several reasons that, although they are not official, they make all the sense in the world. Starting with the obviousness that today's gamer wants and is looking for Ampere GPUs at the right prices, so diverting production line for mining with this architecture is to reduce the number of sales and benefits in the sector that gives NVIDIA the most profitability .

The second factor is very curious: efficiency. From an energy and performance standpoint, Turing is slightly better than Ampere and better handles power spikes, which is vital for any farm farmer to get a return on the investment of GPUs.

NVIDIA-CMP-HX-drivers

Third, we have to look at the price, as NVIDIA could offer these graphics cards at a lower cost than Amperes, which ties in with the fourth and final argument: the manufacturing node.

Although Samsung is responsible for the 8nm of the Ampere GPUs, the 12nm of TSMC is not really that far in performance and efficiency compared to the Koreans, but unlike these, the price per wafer and chip is much lower. It remains to be seen if indeed, NVIDIA has once again considered 12 nm as a reference node for these CMPs, or if, on the contrary, it sees in terms of efficiency and price opting for a higher volume of sales at 8 nm, we will soon know.