Not even a year has passed since the launch of NVIDIA’s GeForce RTX 2000, the extremely powerful graphics from a manufacturer that looked to the future and was betting above all on ray tracing. The relevance of this technology in video games is still a matter of debate, but at NVIDIA they have redoubled their efforts and have just presented their new graphics cards.
These are the GeForce RTX 2060 SUPER, the RTX 2070 SUPER and the RTX 2080 SUPER -which will come a little later-. This is a striking twist on the original models: they give more performance for the same prices as their predecessors, and therefore make things more interesting than ever for those who seek maximum performance and that photorealism that raytracing proposes.
RTX 2060 SUPER and RTX 2070 SUPER Specification
The technical improvements implemented in these graphs affect what matters most: the performance of each model, which gains integers thanks to things like the inclusion of more cores of the different types that the Turing architecture handles, but also other striking improvements.
|GPU||GEFORCE GTX 1060 (PASCAL)||GEFORCE RTX 2060 (TURING)||GEFORCE RTX 2060 SUPER (TURING)|
|CUDA CORES||1 280||1 920||2 176|
|NATIVE CLOCK FREQUENCY||1 506 MHz||1 265 MHz||1 470 MHz|
|MAXIMUM CLOCK RATE||1 708 MHz||1 680 MHz||1 650 MHz|
|FLOPS||4.4 TFLOPS||6 + 6 TOPS (6.5 TFLOPS FP32 / 6.5 TFLOPS INT32)||7 + 7 TOPS (7.2 TFLOPS FP32 / 7.2 TFLOPS INT32)|
|FLOPS TENSIONER||Not applicable||51.6 TFLOPS||57.4 TFLOPS|
|RAYS CAST||0.4 Giga Rays||5 Giga Rays||6 Giga Rays|
|TEXEL FILL-RATE||120.5 Gigatexels / sec||201.6 Gigatexels / sec||224.4 Gigatexels / sec|
|MEMORY INTERFACE||192 bits||192 bits||256 bits|
|MEMORY DATA RATE||8 Gbps||14 Gbps||14 Gbps|
|MEMORY BANDWIDTH||192 GB / s||336.1 GB / s||448 GB / s|
|MEMORY SIZE||6 GB||6 GB||8 GB|
|MAX L1 CACHE SIZE||480 KB||1 920 KB||2 176 KB|
|TGP||120 W||160 W||175 W|
In the case of the RTX 2060 SUPER there are more cores to extend the room for maneuver, but there are special benefits in the memory, which happens to be 8 GB of GDDR6 memory instead of the 6 GB of the original models. In addition, the 256-bit memory interface expands the bandwidth.
Consumption does not skyrocket: here NVIDIA prefers to talk about TGP (Total Graphics Power) instead of the traditional TDP, and although it will be necessary to see if the impact of this higher performance is more or less noticeable, it is true that it increases consumption, but Not too worryingly , something that confirms the good balance that NVIDIA has achieved between power and consumption since it launched the GTX 1000 with Pascal architecture.
|GPU||GEFORCE GTX 1070 (PASCAL)||GEFORCE RTX 2070 (TURING)||GEFORCE RTX 2070 SUPER (TURING)|
|CUDA CORES||1 920||2 304||2 560|
|NATIVE CLOCK FREQUENCY||1 506 MHz||1 410 MHz||1 605 MHz|
|MAXIMUM CLOCK RATE||1 683 MHz||1 620 MHz||1 770 MHz|
|FLOPS||6.5 TFLOPS||7.5 + 7.5 TOPS (7.5 TFLOPS FP32 / 7.5 TFLOPS INT32||9 + 9 TOPS (9.1 TFLOPS FP32 / 9.1 TFLOPS INT32|
|FLOPS TENSIONER||Not applicable||60 TFLOPS||72 TFLOPS|
|RAYS CAST||0.6 Giga Rays||6 Giga Rays||7 Giga Rays|
|TEXEL FILL-RATE||180.7 Gigatexels / sec||233.3 Gigatexels / sec||325.7 Gigatexels / sec|
|MEMORY INTERFACE||256 bits||256 bits||256 bits|
|MEMORY DATA RATE||8 Gbps||14 Gbps||14 Gbps|
|MEMORY BANDWIDTH||256 GB / s||448 GB / s||448 GB / s|
|MEMORY SIZE||8 GB||8 GB||8 GB|
|MAX L1 CACHE SIZE||720 KB||2 304 KB||2 560 KB|
|TGP||150 W||175 W||215 W|
With the RTX 2070 SUPER, there are no changes in memory but there are clear improvements in the cores and their working frequency. In fact, there is a much more important change in these models, which are based on the TU104 processor instead of the TU106.
What does that mean for users? Well, something very relevant, and it is the fact that if you want you can use two RTX 2070 SUPER in SLI . Until now that could only be done with the RTX 2080 / Ti, but that option will now reach more users.
This also allows for a significant performance leap, but it is true that the impact on consumption is also noticeable, and it becomes 215 W instead of the 175 W of the RTX 2070.
Those improvements in the RTX 2060 SUPER and RTX 2070 SUPER demonstrate NVIDIA’s vocation with these models: to respond to some of the most frequent complaints from users of the first generation of cards: that improvement in the memory of the RTX 2060 SUPER and the SLI support in the RTX 2070 will certainly please many users, who now find a more interesting catalog.
In NVIDIA raytracing goes to more
Those responsible for NVIDIA in our country told us when they told us about this launch how the evolution of ray tracing in their graphics is promising. Whereas in the first titles that supported it -like Battlefield V- that option was applied to reflections, now the effects go to more .
In fact, they gave as an example ‘ Control ‘, by Remedy Entertainment, one of the most anticipated releases in that use of this technology with photorealistic results. In this title the ray tracing serves to give a much more realistic look to the reflections, but it also applies to the shadows and global lighting , much more faithful to what we would expect from a real scene.
The list of games with ray tracing support has been expanding and for example ‘ Metro: Exodus ‘, ‘Wolfenstein: Youngblood’, ‘Modern Warfare’, ‘Doom Eternal’ or the future and expected ‘Cyberpunk 2077’ will integrate it into an ambitious. In addition, they explained at NVIDIA, its integration into the current large APIs – with Unreal Engine and Unity as the absolute protagonists – is important so that more and more titles take advantage of that option.
There were not too many comments, yes, about DLSS. This impressive antialiasing technique continues to advance as well, but its application does not depend only on developers: the game that wants to support it needs to be “trained” on NVIDIA servers so that this technique that makes use of artificial intelligence is applied correctly and allows you to win frames without conditioning the level of detail of the games.
More performance for the same money
This launch follows a similar philosophy to that of the “Ti” versions of families like the GTX 1000. The idea here is to give a twist to the original graphics and do it without impacting the price.
In NVIDIA they highlighted this circumstance by making a summary of the great keys of these three cards. The RTX 2080 SUPER will arrive a little later than the RTX 2060 SUPER or the RTX 2070 SUPER, but this range will be absolute from now on for NVIDIA, which in fact will replace their predecessors in many cases.
In fact, the only ones that will coexist on the market will be the RTX 2060 (which will continue to be produced) with the RTX 2060 SUPER. In the case of the RTX 2070 SUPER and RTX 2080 SUPER, they will replace the original models . We will be able to find those models while supplies last, but they will no longer be manufactured and NVIDIA therefore focuses on the new models although it also keeps the RTX 2080 Ti in the catalog.
An interesting section of this evolution is in energy consumption: here NVIDIA highlighted us how in this section its rival, AMD , is still unable to compete despite using a 7 nm lithography (12 nanometer technology is used in Turing). In addition, those responsible for NVIDIA indicated, the improvement in their families is also important: from the 1080 Ti to the RTX 2080, for example, they managed to reduce consumption and still obtain 40% higher performance.
Prices and release dates of the RTX 2060 SUPER, RTX 2070 SUPER and RTX 2080 SUPER
The range of NVIDIA cards with RTX technology therefore changes, and now we have new models that, as we said, gain in power without the prices varying from those established for the original models.
hus, that family is as follows with their prices and release dates:
- GeForce RTX 2080 Ti (1,259 euros, available for months): Even with the new releases, this model is still the most powerful and ambitious of all that NVIDIA offers today.
- GeForce RTX 2080 SUPER (739 euros, available on July 23, 2019): the manufacturer will give more data later, but keeps the price being faster, for example in terms of the speed of its memory. It will offer 11 + 11 TOPs (FP32 + INT32) and 89 Tensor TFLOPs. According to NVIDIA it is faster than the Titan Xp.
- GeForce RTX 2070 SUPER (529 euros, available July 9, 2019): faster than the GTX 1080 Ti according to NVIDIA, it offers 9 + 9 TOPs (FP32 + INT32) and 73 Tensor TFLOPs. It is on average 16% faster than the RTX 2070 for the same price according to manufacturer data.
- GeForce RTX 2060 SUPER (419 euros, available July 9, 2019): faster than GTX 1080, now with 8 GB of memory instead of 6 GB, offers 7 + 7 TOPs (FP32 + INT32) and 57 Tensor TLOPs . On average 15% faster than the RTX 2060 for the same price. It is close to the performance of last year’s RTX 2070 according to NVIDIA.
- GeForce RTX 2060 (369 euros, available for months): the first step that gives access to all the options of NVIDIA’s RTX technology is still interesting for users somewhat less ambitious or with tighter budgets.