NVIDIA Pronounces GeForce RTX 2060: Beginning At $349, Out there January 15th
On the launch of CES 2019 with a surprisingly anticipated keynote session, NVIDIA as we speak introduced the subsequent member of the GeForce RTX household of graphics playing cards: the GeForce RTX 2060. The newest and now the most affordable member of the RTX 20 sequence continues the cascade of Turing architectural product variations in the direction of cheaper and better quantity market segments. The brand new card is anticipated to hit the market on January 15 subsequent week, priced at $ 349.
We wouldn’t have top-down map specs but, however primarily based on the knowledge NVIDIA has launched up to now, the GeForce RTX 2060 seems to be primarily based on a slimmed-down model of the TU106 GPU. The GeForce RTX 2070 is already in use , That is noteworthy as a result of NVIDIA has used a distinct GPU for every RTX card – TU102 / 2080TI, TU104 / 2080, TU106 / 2070 – making it the primary such card within the household. It's additionally a small shift from the established order for GeForce xx60 elements typically, which historically at all times have their very own GPU, however NVIDIA will get decrease for price causes.
Comparability of NVIDIA GeForce specs
RTX 2060 Founders Version
GTX 1060 6 GB
FE: 1710 MHz
14 Gbps GDDR6
Eight Gbps GDDR5
Eight Gbps GDDR5
14 Gbps GDDR6
Reminiscence Bus Width
Single Precision Perf.
FE: 7.9 TFLOPS
N / A
N / A
Manufacturing Course of
TSMC 12nm "FFN"
TSMC 12nm "FFN"
Date of introduction
RRP: $ 249
FE: 299 US
RRP: $ 379
FE: 449 US
RRP: 499 USD
FE: $ 599
Let's dive into the numbers anyway. The GeForce RTX 2060 has 1920 CUDA cores. This implies we're a 30 SM configuration in comparison with the RTX 2070's 36 SMs. Since Turing's core structure is designed to scale with the variety of SMs, which means all of the capabilities of the core computation are scaled down in the same method. The 17% lower in SMs means a 17% lower in RT core numbers, a 17% lower in tensor cores, 17% much less texture models, 17% fewer L0 / L1 caches, and so forth.
No marvel the clockspeeds are very near the opposite RT106 card from NVIDIA (RTX 2070). The bottom clock charge is barely decrease than 1365 MHz, however the increase clock is barely larger than 1680 MHz. General, the RTX 2060 is able to delivering about 87% of the RTX 2070's processing energy / RT / texture efficiency, creating an unusually small hole between an xx70 card and an xx60 card. In different phrases, the RTX 2060 is in place to outperform its efficiency / shading efficiency.
Nonetheless, TU106 has gotten extra customization within the backend, and for workloads that aren’t pure calculations, the decline shall be harder. The cardboard comes with solely 6GB of GDDR6 VRAM, versus 8GB for the larger brother. In consequence, NVIDIA doesn’t occupy two of the TU106's Eight reminiscence controllers, leading to a 192-bit reminiscence bus. Which means that the RTX 2060 gives solely 75% of the RTX 2070's reminiscence bandwidth when utilizing 14 Gbps GDDR6. To place it in figures, the RTX 2060 provides 336 GB / s bandwidth for the RTX 2070's 448 GB / s.
As a result of reminiscence controllers, ROPs, and L2 cache are all very tight within the NVIDIA structure, this additionally signifies that ROP throughput and the quantity of L2 cache are additionally diminished by 25%. Thus, for graphics workloads, the sensible efficiency degradation shall be larger for the throughput charge of 13%, however for the ROP / reminiscence throughput typically lower than 25%.
I even have particular issues about solely 6GB of VRAM – particularly in a time when gaming consoles are shipped with Eight-12GB of VRAM – however we'll look into that later within the later evaluation.
NVIDIA charges the RTX 2060 for a TDP of 160W. That is barely decrease than the RTX 2070, however solely marginally, as these playing cards are rated for 175W. With reduced-power GPUs having restricted power-down capabilities, it's not unusual for such a card to eat practically as a lot vitality as its full-fledged counterpart.
If you happen to have a look at the NVIDIA specs, there aren’t any purposeful variations between the RTX 2060 and the RTX 2070. For instance, the latter didn’t have SLI help, so there's nothing to vary right here. Not solely is the RTX 2060 slower and cheaper than its greater brother, it additionally provides all of the options we anticipate from the Turing structure household.
By way of card design, subsequent week's launch shall be each a reference and customized card sharing. NVIDIA will launch a Founder's Version card with the same old styling – and within the footage that NVIDIA has launched, it appears to be like similar to the RTX 2070 – whereas board companions have been utilizing the RTX 2070 for a number of months now TU106 GPU work has taken the time to realize the expertise they should design their very own boards. Just like the customized RTX 2070 boards which have come out since then, you possibly can anticipate these playing cards to deal with the vary from small, mITX-sized single fan playing cards to giant tri-fan monsters.
apart: Whereas NVIDIA calls this xx60 class, the value and basic energy necessities of the RTX 2060 appear misplaced. The xx60 sequence was historically the mainstream card from NVIDIA. Till the introduction of the GTX 1060 6GB they have been often round 200 US . The GTX 1060 6GB reached the highest finish of the $ 249 scale for customized playing cards (and a whopping $ 299 for the Founders Version). Nonetheless, the $ 349 RTX 2060 is now effectively past the same old worth vary. With three different GeForce playing cards above it might not be a high-end system, however the definition is a map for fans.
This additionally signifies that efficiency comparisons with the GTX 1060 seem equally misplaced. With 1920 CUDA cores, the RTX 2060 is considerably sooner than the GTX 1060, but in addition prices 100 US extra and consumes 30 W extra energy. All in all, this seems like the brand new GTX 1060 on the brand new GTX 1070 (pristine MSRP 379). We'll need to see what the actual efficiency appears to be like like after we evaluation the brand new card, however up to now it appears to be like like this As if NVIDIA maintained its general worth / efficiency curve for the RTX 20 sequence, which signifies that the cardboard in current video games is simply a minor enchancment over the GeForce GTX 10 Collection card, it replaces this worth stage.
Despite the fact that the advance in efficiency is low, it will considerably change the aggressive panorama. If the NVIDIA GTX 1070 Ti's efficiency claims are legitimate, AMD's Vega 56/64 playing cards shall be undermined, and the corporate might want to reply if it desires to be a part of that market.
All in all, the worth proposition argument for the RTX 2060 seems to be similar to the remainder of the RTX 20 sequence: NVIDIA expects shoppers to be ready to pay a premium for the subsequent technology Turing options. Structure to pay – particularly raytracing acceleration and the assorted purposes of tensor cores. That's why NVIDIA must additional promote these options, get builders on board, and usually promote picture high quality enhancements. Even NVIDIA appears to appreciate that this won’t be simple. Subsequently, they’re additionally launching a brand new GeForce recreation pack program that can embrace the brand new RTX 2060, the place consumers should buy Battlefield V without spending a dime or anthem
The GeForce RTX 2060 shall be launched subsequent week on January 15th. Costs begin at $ 349. We’ll have a look at this new map very quickly, please keep tuned.