Contrary to early leaks and rumors, the GeForce RTX 4090 turned out to be very energy efficient and even more economical than the RTX 3090. But Nvidia’s new product can be made even less voracious.

The authors of the QuasarZone resource decided not only to achieve less power consumption from the RTX 4090, but also to achieve this in two different ways. The first is reducing the power limit, but the second is more complex – it is undervolting or reducing the voltage on the GPU.

As you can see from the diagrams, both methods are effective, and in both cases you can achieve an impressive reduction in power consumption with a slight drop in performance. In the case of power limit management, power consumption can be reduced to 268 W, with a loss of only about 6% of performance. Reducing the voltage allows you to achieve power consumption of only 232 watts, sacrificing less than 8% of performance.

Thus, the RTX 4090 can be turned into a mid-range card in terms of power consumption, while keeping its monstrous performance almost unchanged.

Intel was expected to endow Raptor Lake processors with DLVR (Digital Linear Voltage Regulator) technology. Information about it appeared last year. However, it has now become known that current Intel CPUs do not support it.

Information about DLVR was neither in Intel’s presentation, nor somewhere on the site describing Raptor Lake. However, the CPU DLVR Bypass Mode Enable item has appeared in the new BIOS of Asus motherboards. True, its activation does not lead to anything.

One of the overclockers from the Asus team said that the technology does not work because Intel eventually disabled it in the new generation of processors. For what reasons, it is not clear, but from the tests we know that, perhaps, in vain, since Raptor Lake turned out to be even more voracious than Alder Lake. However, next-generation CPUs may well support the new technology already fully, however, there is no clarity on this matter.

DLVR is a digital linear voltage regulator that complements the main one. This should allow a significant reduction in energy consumption.

Samsung has introduced a new memory for video cards. The new solution is called GDDR6W.

As the name implies, this is not something completely new, but it is not at all the same as GDDR6/GDDR6X.

GDDR6W deserves attention both for its characteristics and for the technologies behind it. In particular, the new packaging technology FOWLP (Fan-Out Wafer-Level Packaging) is used here. It is thanks to her that the new memory is allocated with much more bandwidth and twice the capacity.

In particular, compared to the usual GDDR6 with a frequency of 24 GHz, the bandwidth of GDDR6W is twice as high, as is the capacity. Thanks to this, the volume of one microcircuit was increased to 4 GB, the number of I / O operations also doubled, and the microcircuits themselves became much thinner, since the crystal is mounted directly on a silicon wafer, and not on a printed circuit board. As a result, the height of the chip with doubled capacity is only 0.7 mm versus 1.1 mm for conventional GDDR6.

Speaking of bandwidth, Samsung compares GDDR6W to HBM2e memory, which is 1.4 TB/s for the chip at 22 GHz.

The company completed standardization for the new memory in the second quarter. But it is not yet clear when GDDR6W will hit the market.