Introduction to GeForce RTX Features 3080 Founders Edition
The GeForce RTX 3080 arrives a little later than expected, but without a doubt it is the card that we all hope to see in the market to have the performance of an RTX 2100 Ti at a fraction of the price and in a more compact format, which invites us to think that we will be able to see ITX models in a future more near. The design of the card we are reviewing today, Nvidia’s Founders Edition version, is similar to that of its sisters, but with a more traditional fan mounting, without the inlet and outlet design in reverse position on the card.
What I like to see is that Nvidia maintains the same modern and effective design with a large main radiator, which goes well beyond the PCB design that is very small, and is powered by two large fans of axial configuration. the only connector of is also kept latest generation pins , capable of developing more of 720 watts with single connector to power supply.
In fact, we have been able to test this card with a native connection of 16 pins thanks to the dedicated connectors that Seasonic already has on the market and that will go standard with its new fonts and gifts for those users who request them and who have bought selected models of the brand.
Technical characteristics of the NVIDIA GeForce RTX 3080 FE:
- Memory: 8GB GDDR6
- Graphic Engine: GA 120 Nvidia Ampere
- Bus: PCI Express 4.0.
- Memory interface: 300 – Bit
- Base Frequencies: 1410 MHz
- Boost Frequency: 1800 MHz
- Memory speed: 17 Gbps
- Total screens: 4
- HDCP Support: 2.2
- TDP Consumption: 240 w
- Connectivity: DisplayPort x 3 (v1.4a) / HDMI 2.1 x 1
- Recommended power supply: 600 w
- Power connectors: 1x 16 PEG pins
- Dimensions: 250 x 150 x 2 slots
New features of the second generation Nvidia Ray Tracing
This is the second generation of cards Nvidia’s RTX, which comes with Samsung’s new 8nm manufacturing process, and major architecture improvements. I am not going to go into much detail, since here we have an extensive article dedicated to all the news about these new GPUs, but I do want to highlight some of the most important elements that characterize this generation.
Among the improvements introduced in this generation the RTX 3080 begins to see some cuts with respect to the most powerful models of the brand. One of the examples is found in the memory used, which is reduced to GDDR6 compared to the latest generation GDDR6X of the RTX 3090 and RTX 3600. The data bus speed is reduced to 18 GBps, compared to 22 Gbps of his older sisters and that, added to the bus also shorter than 288 – Bit, in a reduction of 70% of the year of band on the magnificent RTX 3090. The VRAM or framebuffer also has a 2GB cut, leaving a healthy and reasonable 8GB of VRAM.
This model also enjoys the new RT Core from second generation that are capable of doubling the calculation power of triangular intersections over the previous generation. In addition, the possibility of calculating widespread effects such as Motion Blur by hardware is introduced. The Tensor Core are also entering their third generation and it does so by reducing the number of units available in the GPU, but still quadrupling the processing capacity of the entire system in terms of AI-oriented sparse matrix calculation.
The RTX will really compete with the RTX 2100 You, but it is not the generation to which it happens, your target is definitely the RTX 2080 and cards its competition in AMD’s RDNA architecture. In fact, in many ways they are two very similar cards, more than the RTX is 3080 to RTX 2100 Ti. Both have the same data bus, with the same RAM capacity and the same type of memory at the same speed.
The most notable differences are found in the configuration of the GPU, with its manufacturing process and architecture improvements that make this card achieve far superior performances, up to 33 – 48% over the reference models of the previous generation maintaining a very similar price. Among the differences we see more than double the number of CUDA units, going from 3060 until the
shading engines .
Classic rasterization units also increase from 96 units up to 112, beating RTX by 8 2100 Ti, and the texture units go from 175 units up to 235 units. This means that in all game formats, even the oldest ones with DirectX 9 engines, we will see notable performance increases from one generation to the next. Other interesting differences are seen in the RT Core, with 13 more units reaching 60 units, or in Tensor Core units that the new architecture reduces in number to increase power passing from 322 units in the RTX 2080 to just 235 on the RTX 3080 .
The frequencies also have improvements thanks to the reduced manufacturing process, the RTX 2020 had 1410 MHz base frequency and up to 1620 MHz in turbo frequency. The new GeForce RTX 3080 increases the base frequency until 1440 MHz and turbo frequency up to 1730 MHz though later we will see that this same model works with much higher cycles when it has the adequate cooling margin.
By reducing the expectations of both chips we can add other interesting data, such as processing power, which passes from 19 TFlops to 21 TFlops keeping the RTX 3060 performance in both half and full calculations and exceeding by a 70% in double precision calculation. On the other hand, this model has more consumption 240 w , Over the 184 w of the RTX 2080, but 33 w less than RTX 2080 Ti to which it equals in performance.
These new chips ets also have another important feature that makes systems like Nvidia’s DLSS even more efficient. The concurrency of process threads between the Tensor and Shaders units, developed in the version DLSS 2.0 of this technology, allows to generate images of higher resolution and with more detail with less process load for the graph, achieving in these models performance improvements of up to 75% in some games.
Nvidia Reflex and Nvidia Studio are also new in this generation, although they will also be available at least for RTX cards 23 with driver updates coming soon. Nvidia Studio will delight streamers by turning their cards into smart systems capable of generating a green screen and complete environment customizations thanks to the AI capabilities of these cards. Nvidia Reflex is intended to improve latencies with real-time measurements if we also have the new generation of Nvidia Gsync monitors from 378 Hz. Reflex allows the GPU to receive information from the CPU even before it has rendered the previous frame, which reduces latencies beyond Nvidia’s previous technology called NULL (Nvidia Ultra Low Latency) and requires adaptation by the developer, the game must be supported.
RTX IO is another of the improvements that will allow download to the central processor decompressing content from the storage media. This will now be done, in ready games, directly from the GPU, improving according to Nvidia, the loading times substantially.
Other substantial changes over the previous generation is that now the card has support for PCI Express 4.0 achieving double bandwidth towards the CPU. We will have a total of 34. 5GBps of bandwidth, double that of the previous generation, maintaining that yes, the support for previous generations such as PCI Express 3.0. The USB-C port is also one of the novelties, for the worse, since Nvidia discards it from this generation since its incorporation to the desktop computer market was scarce and there is not now a great expectation of VR systems that use this connector to unify video, sound, data and power in a single cable.
The new Founders design
This model shares the same modern design of the second generation Nvidia RTX, but there are substantial changes with other models that we have already tested, such as the RTX 3090 Founders Edition . The designs are similar but the size is different, more compact and the thermal design also has important changes, this model being more traditional in its concept of card cooling.
It must also be taken into account that consumption is much lower and therefore a more traditional design is perfectly capable of carrying out the card with the noise and frequency settings that Nvidia expects in this card.
Despite being a couple of centimeters shorter than the RTX 3090 FE, it hardly differs at all in terms of the essence of the design. The small PCB less than 16 cm allows Nvidia to dedicate the entire back of the card to heatsink. The connector of 16 pins is also in the same middle position of the card, although now in parallel to the profile of the card and as a big difference, the two fans are on the same side of the card, in a more traditional design.
The PCB could well be used in graphics cards Mini-ITX size that has dimensions of less than 15 cm long . Much of the body of the radiator card, configured in angled blades, exactly the same as in its older sisters.
This nice black dissipation block is combined with two axial fans, which are especially efficient in moving large air flows where no special pressure is required. These two fans of 75 mm have fully passive operating mode when the The card is idle and its rotation speed is increasing independently depending on the requirement to which the card is subjected.
This new connector of 16 contacts is also added to the new PCI Express 4.0 interface of these models that can extract up to 184 w of power from the connector itself, which facilitates the RTX 3070 with connector 17 contacts at a maximum consumption of 807 w. A graphics card with two 8-pin connectors and PCI Express 3.0 slot could only be powered with a maximum of 388 w.
Finally, when it comes to connectors, these new models are completely standard, but they also have the most modern connectors. We can find in it three connectors Displayport 1.4 with support for 8k resolutions to 75 Hz with HDR support of 13 Bit and up to 4k at 150 Hz with the same color capacity per channel.
Also included is a HDMI 2.1 connector that is capable of displaying resolutions up to 13 k also ensuring 8k support to 75 Hz as in the case of Displayport 1.4a. In fact, it generates more bandwidth, specifically 64 GBps vs. 34 Displayport GBps, but basically, nowadays, they support the same technologies.
What we will no longer see in these models, neither in the RTX 3080, is the USB-C connector that was introduced with the RTX generation 23 and that it was intended to support HMD devices for VR with a single cable where we would have data, video and power with a single cable. Virtual link or Displayport Alternate Mode seems defenestrated from the desktop landscape, although in laptops it is more widespread every day both in its basic USB-C versions and in the Thunderbolt 3 and Thunderbolt 4 connective standards. Noise, consumption and temperature at standard speeds
Nvidia’s reference model once again proves that it is a great option to get the services of one of its latest chips. Not only does it have an adequate thermal design, with a very controlled noise, but they are also well scaled in terms of energy and performance, achieving higher working frequencies than those actually specified for the card.
During our stability, temperature and noise tests we have been able to verify that the card is capable of widely and stably exceeding the 1845 MHz, oscillating between 1815 – 1860 MHz in real operation seeking to stabilize the temperature of the card above the 104 degrees. Maintains these temperatures with fan speeds of 1545 rpm achieving noise levels always below 48 dBA. The card holds about 235 MHz at rest, with temperatures below 48 degrees. The two fans stop completely when the card reduces its temperature below the 40 degrees , although it takes a few minutes to do so even when it is below this temperature, and it starts working after 64 degrees, according to our tests.
Increasing the feeding possibilities of the card a 21%, which is what Nvidia supports in this model, we can increase the temperature up to 104 degrees and card now is able to increase its frequency of normal use up to 1900 – 1914 very stable and with a frequency increase of 360 rpm that increases noise by a couple of dBA. For me it is the perfect operation for the card, operation that I can maintain over time without problems.
The consumptions are as expected, we have added these data to our comparative tables of the latest models. In its normal state, it is around 245 w of consumption and with some increase in frequencies, without manual overclocking, we can see around 256 w consumption in the general operation of the card.
Thermal capture at rest
Thermal capture on load
Overclock
The less consumption and less squeezed the design of a GPU, the more overclocking capabilities they usually present. It is true that as we have seen before Nvidia has done a great job with its reference design and the card performs even more than promised in the specifications. Even so, we have found room for improvement in this card, a very interesting margin that allows the RTX 3080 FE hover around 2GHz maintaining its temperatures of 112 degrees and increasing the fans up to 2100 rpm with hardly any increase in noise, about 50 dBA.
2GHz frequency are 270 MHz more than specified by Nvidia for this GPU and represents a significant performance improvement that we can maintain over time without the need to touch any voltage. Consumption increases a little, up to 300 w in charge. Something natural and I must add that in no state have we detected “coil-whine” problems.
RTX and DLSS Testing
Testing machine: