NVIDIA GeForce RTX 2080 and 2070 cards will soon move to the most successful Turing chips

by



[ad_1]

EVGA GeForce RTX 2080
You may not realize it, but not all GeForce RTX 2080 and GeForce RTX 2070 graphics cards are the same. We are not only talking about custom cooling solutions and printed circuit boards (PCBs), but GPUs actually present in each model: NVIDIA splits its GPU Turing arrays from these models into two categories, or at least that is how things were over. Not anymore, apparently.
We reported on this separation of GPUs last November. Also known as "binning", NVIDIA has up to now offered its hardware partners a variant of a low-end (and probably cheaper) GPU for non-overclocked cards, as well as a low-end GPU. a top-of-the-range GPU for overclocked models at the factory. Here's how things have rocked up to now …
  • TU104-400A-A1: Overclocked GeForce RTX 2080 cards
  • TU104-400-A1: Current GeForce RTX 2080 cards
  • TU106-400A-A1: Overclocked GeForce RTX 2070 cards
  • TU106-400-A1: GeForce RTX 2070 Card Stock
EVGA GeForce RTX 2070

For both GPU families, revision "A" represents the arrays that have been grouped together to achieve higher clock speeds. Steve Burke at Nexus Gamers He had discovered the split GPU families and had noted during his tests on the GeForce RTX 2070 XC ultra-graphics games (TU106-400A-A1) and the GeForce RTX 2070 (TU106-400-A1) black game. EVGA, that there was a significant gap between the clocks.

"The XC Ultra measures an average of about 1935 MHz for the entire test, while the Black 2070 experiences a clock fall from about 1815 MHz to 1785 MHz." The average frequency varies between 1785 MHz and 1800 MHz. for Black 2070. This is a significant difference for -The-Box, and indicates this performance advantage resulting from the best binned GPUs, "wrote Burke at the time.

And now? Igor Wallossek at TomsHardware Germany indicates that NVIDIA has announced to "multiple sales channels" that it was turning to the use of only high-end GPUs. The new variants include the TU104-410 (GeForce RTX 2080) and TU106-410 (GeForce RTX 2070).

The good thing about this decision is that they will all be qualified for overclocking. The discretion will be left to each individual hardware partner, but from the consumer's point of view, all future cards will be created equally, in terms of the actual GPU. This could potentially improve the overclocking performance of the end user, even on stored cards.

There are some caveats, though. Firstly, even if GPUs will no longer be separated into higher performance A-versions and non-A less powerful variants, other factors may affect overclocking – PCB designs, power scheme, cooling, etc. And secondly, there will be no way of knowing if you are buying a newer card with the latest graphics processor under the hood or an existing batch that uses one of the older, split graphics processor.

We would not be surprised if some retailers specify which cards use a newer GPU, although if that happens, you can probably expect higher prices. However, this will work over time as existing stocks are depleted and replaced by new batches.

[ad_2]

Source link