The GeForce RTX 2080 that was launched last September was undoubtedly repaired. But it was much faster than the GeForce GTX 1080 Ti. More problematic, the Founders Edition model sold for $ 800, which means it paid an additional $ 100 for its slight performance advantage over the flagship of the previous Nvidia generation.
The $ 699 GeForce RTX 2080 offers scales in favor of players with high-update QHD monitors and 4K screens at average frame rates that are 6% higher than GeForce RTX 2080 (and 17% better than GTX 1080 Ti) point priced at $ 100 less expensive than the previous 2080 Founders Edition card. It also offers performance that is way ahead of the new Radeon 5700XT with AMD Navi technology, but that's not surprising given that the Nvidia card costs 75% more.
It will not make anyone with a high-end graphics card want to update. But if you were looking for an upgrade to get a better value than the one offered by the first round of Turing-based cards, know that GeForce RTX 2080 offers you 21% better performance per dollar than GeForce RTX 2080 Founders Edition (or scan an improvement of 6 % over 2080 less expensive).
Editor's Note: The title "Leaving Navi in the dust". We have revised the title to better reflect the content of the article, which mainly compares the RTX 2080 with other cards in its price range.
Meet the GeForce RTX 2080 Super
Even more than the GeForce RTX 2060 Super and the 2070 Super, the 2080 Super looks very, very similar to its predecessor. Nvidia now has a wide range of impeccable TU104 processors that can be used to build GeForce RTX 2080 without deactivating any of its resources.
As a brief review of the TU104 and its vital specifications, TSMC manufactures the GPU on its 12 nm FinFET node. A total of 13.6 billion transistors are piled up at 545 mm², which is naturally narrower than Nvidia's massive TU102 processor but still slightly larger than the flagship of the latest generation of 471 mm² (GP102).
TU104 is built with the same building blocks as TU102; Simply present less of them. Transmission multiprocessors still have 64 CUDA cores, eight Tensor cores, one RT core, four texture units, 16 load / large units, 256KB or log space, and 96KB or L1 cache / shared memory. The TPCs are still composed of two SMs and a PolyMorph geometry engine. There are four TPCs per GPC and six GPCs throughout the processor. Therefore, the fully enabled TU104 controls 48 SM, 3,072 CUDA cores, 384 Tensor cores, 48 RT cores, 192 texture units and 24 PolyMorph engines. TU104 therefore loosens an eight-lane NVLink connection compared to TU102, limiting it to an x8 link and 50 GB / s or bidirectional performance.
A narrower corresponding back-end feeds the computing resources through eight 32-bit GDDR6 memory controllers (256-bit aggregates) connected to 64 ROPs and 4 MB or L2 cache. But instead of populating that 256-bit path with Micron 14GB / s GDDR6 modules, Nvidia changes to 8 GB or Samsung K4Z80325BC-HC16, part of 16 Gb / s with clock speed at 15.5 Gb / s for GeForce RTX 2080 Super. Why adjust the data rate? Jumping to 16 Gb / s would have required a PCB modification, and the gain would not have been worth the added cost and complexity. However, Nvidia says that the memory should still be overclocked at 16 Gb / s manually. What's more, GPU overclocks are more effective in improving performance, since the chip does not lack bandwidth.
|GeForce RTX 2060 Super||GeForce RTX 2070 Super||GeForce RTX 2080 Super|
|Architecture (GPU)||Turing (TU106)||Turing (TU104)||Turing (TU104)|
|FP32 Compute Peak||7.2 TLFOPS||9.1 TFLOPS||11.2 TFLOPS|
|RT Cores||3. 4||40||48|
|Base Clock Rate||1470 MHz||1605 MHz||1650 MHz|
|GPU increase rate||1650 MHz||1770 MHz||1815 MHz|
|Memory capacity||8GB GDDR6||8GB GDDR6||8GB GDDR6|
|Memory bus||256 bits||256 bits||256 bits|
|Memory bandwidth||448 GB / s||448 GB / s||496 GB / s|
|Transistor count||10.8 billion||13.6 billion||13.6 billion|
|The size||445 mm²||545 mm²||545 mm²|
|SLI support||do not||yes||yes|
The cumulative effect of a more capable TU104 is amplified at higher clock speeds. Because GeForce RTX 2080 Founders Edition had a 1,515 MHz base and a 1,800 MHz GPU Boost GPU, the GeForce RTX 2080 Super starts at 1,650 MHz and generally operates about 1,815 MHz. The maximum FP32 calculation performance increases from 10.6 TFLOPS to 11.2 TFLOPS. And the memory bandwidth increases to 496.1 GB / s, compared to 448 GB / s.
These more aggressive specifications influence the consumption of the GeForce RTX 2080 super power, which makes its board reach up to 250W. However, Nvidia is offered with eight and six pin auxiliary power connectors.
In this sense, GeForce RTX 2080 Super and GeForce RTX 2080 Founders Edition are almost identical from the outside, apart from some Super marks on the back plate and the RTX 2080 Super logo on a reflective adhesive label that is applied on the front. A pair of 8.5 cm axial fans on each side uses 13 blades to keep heat away from TU104 as quickly as possible.
The same forged aluminum cover keeps them in place on a good weld pile in a steam chamber.
Inside, you are seeing the same phase 8 (GPU) + 2 (memory) power supply. Six of the GPU phases are powered by the power connectors, while the other two consume current from the PCIe slot.
On the front, Nvidia shows the same screen outputs: three DisplayPort 1.4 connectors, an HDMI 2.0b port and a USB Type-C interface with VirtualLink support.
How we tested the GeForce RTX 2080 Super
With the GeForce RTX 2060 Super and 2070 Super reviews behind us, along with our Radeon RX 5700 and 5700 XT coverage, we were able to fill in some blanks in our reference data using a completely new platform. The machine we are testing now is powered by the six-core Intel Core i7-8086K CPU on a Z370 Aorus Ultra Gaming motherboard with 64GB or a Corsair CMK128GX4M8A2400OC14 kit. We are still using a pair of 500 GB Crucial MX200 SSD for our game suite, along with the Noctua NH-D15S fan / heatsink combo.
Our latest data library already included GeForce RTX 2080, GeForce RTX 2070, GeForce RTX 2060, GeForce GTX 1080, GeForce GTX 1080, GeForce GTX 1070 and GeForce GTX 1070. To that list, we add GeForce RTX 2080 Ti. All these cards are represented by the Nvidia Founders Edition models themselves, except the 1070 Ti, which is an MSI GeForce GTX 1070 Ti Gaming 8G. AMD Radeon VII itself is also part of the comparison, together with Sapphire's Nitro + Radeon RX Vega 64 and Nitro + Radeon Radeon RX Vega 56. These associated cards ensure that we do not see the frequency / regulation problems presented with our reference models.
Our reference selection includes Battlefield v, Destiny 2, Far Cry 5, Final Fantasy XV, Forza Horizon 4, Grand Theft Auto V, Metro Exodus, Shadow of the Tomb of the Raider, Strange brigade, Tom Clancy's Division 2, Ghost Recon Wildlands by Tom Clancy, The sorcerer 3 duck Wolfenstein II: The New Colossus.
The test methodology we are using comes from PresentMon: Performance in DirectX, OpenGL and Volcano. In summary, these games are evaluated using a combination of OCAT and our own internal GUI for PresentMon, with registration via GPU-Z.
We are using the 431.16 driver for Nvidia's GeForce RTX 2060 and 2070 Super and version 430.86 for all Nvidia cards of the previous generation. Earlier this month, Nvidia published the 431.36 driver build, which affected performance Tom Clancy's Division 2, Strange brigadeY Exodus meter. As a result, we had to try GeForce RTX 2080, 2070 and 2060 again using that update. On the AMD site, we are using Adrenalin 2019 Edition 19.6.3 for all three cards.
MORE: The best graphics cards
MORE: Desktop GPU Performance Hierarchy Table
MORE: All graphic content