New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

With NVIDIA demonstrating real-time ray tracing on GeForce RTX series graphics cards, it's hard to doubt that this technology (in judicious combination with the rasterization algorithm) is the future of PC gaming. However, Turing-based GPUs with dedicated RT cores have until recently been considered the only discrete GPU category that has the processing power to do so.

As tests of the first games that mastered Ray Tracing (Battlefield V, Metro Exodus and Shadow of the Tomb Raider) showed, even GeForce RTX accelerators (especially the younger one - RTX 2060) experience a significant drop in frame rate in hybrid rendering tasks. Despite early successes, real-time ray tracing is not yet a mature technology. Only when not only the most advanced and expensive devices, but also mid-range graphics cards reach the previous standards of performance in new wave games, will it be possible to declare that the paradigm shift launched by Jensen Huang's company has finally happened.

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

Ray tracing on Pascals - pros and cons

But already now, not a word has yet been said about the future successor to the Turing architecture, NVIDIA decided to spur progress. At the GPU Technology Conference event last month, the green team announced that Pascal-based accelerators, as well as lower members of the Turing family (GeForce GTX 16 series), will acquire real-time ray tracing on par with RTX-branded products. Today, the promised driver can already be downloaded from the official NVIDIA website, and the list of devices includes models of the GeForce 10 family, starting with the GeForce GTX 1060 (6 GB version), the TITAN V professional accelerator on the Volta chip, and, of course, the newly arrived mid-range models on the chip TU116 - GeForce GTX 1660 and GTX 1660 Ti. The update also affected laptops with the appropriate GPUs.

From a technical point of view, there is nothing supernatural here. GPUs with unified shader units could perform Ray Tracing long before the advent of the Turing architecture, although at that time they did not have enough speed for this capability to be required in games. In addition, there was no single standard for software methods, apart from closed APIs like proprietary NVIDIA OptiX. Now that there is a DXR extension for Direct3D 12 and similar libraries in the Vulkan programming interface, the game engine can access them regardless of whether the GPU is equipped with specialized logic - as long as the driver allows this. Turing chips have separate RT cores for this purpose, and in the Pascal architecture GPU and TU116 processor, ray tracing is implemented in a general-purpose computing format on an array of shader ALUs.

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

However, everything we know about the Turing architecture from NVIDIA itself suggests that Pascal is not suitable for DXR-enabled applications. In last year's presentation dedicated to the flagship models of the Turing family - the GeForce RTX 2080 and RTX 2080 Ti - the engineers gave the following calculations. If you throw all the resources of the best consumer graphics card of the last generation - the GeForce GTX 1080 Ti - into ray tracing calculations, then the final performance will not exceed 11% of what the RTX 2080 Ti is theoretically capable of. It is equally important that the free CUDA cores of the Turing chip can at the same time be used for parallel processing of other image components - executing shader programs, the queue of non-graphic Direct3D calculations during asynchronous execution, and so on.

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

In real games, the situation is more complicated, because on existing hardware, developers use DXR functions in a dosed manner, and the lion's share of the computational load is still occupied by rasterization and shader instructions. In addition, some of the various effects that are created using ray tracing work well on the CUDA cores of Pascal chips. For example, the mirror surfaces in Battlefield V do not imply secondary reflection of rays, and therefore, are a feasible burden for powerful video cards of the past generation. The same applies to shadows in Shadow of the Tomb Raider, although rendering complex shadows formed by multiple light sources is already more of a challenge. But global illumination in Metro Exodus is difficult even for Turing, and you cannot expect comparable results from Pascal to any extent.

Whatever one may say, we are talking about a multiple difference in theoretical performance between representatives of the Turing architecture and their closest analogues on Pascal silicon. Moreover, not only the presence of RT cores plays in favor of Turing, but also numerous general improvements inherent in new generation accelerators. So, Turing chips can simultaneously perform operations on real (FP32) and integer (INT) data, carry a large amount of local cache memory and separate CUDA cores for reduced accuracy calculations (FP16). All this means that Turing not only handles shader programs better, but can also compute ray tracing relatively efficiently without specialized blocks. After all, rendering with Ray Tracing is so resource-intensive not only and not so much the search for intersections between rays and geometry elements (which RT cores are engaged in), but the calculation of the color at the intersection point (shading). And by the way, the listed advantages of the Turing architecture fully apply to the GeForce GTX 1660 and GTX 1660 Ti, although there are no RT cores in the TU116 chip, so the tests of these video cards with software ray tracing are of particular interest.

But enough of the theory, we've already collected data on the performance of Pascals (as well as younger Turings) in Battlefield V, Metro Exodus and Shadow of the Tomb Raider based on our own measurements. Note that neither the driver nor the games themselves adjust the number of rays in order to reduce the load on the GPU without RT cores, which means that the quality of the effects on GeForce GTX and GeForce RTX should be the same.

Test stand, testing methodology

Test stand
CPU Intel Core i9-9900K (4,9GHz, 4,8GHz in AVX, fixed frequency)
Motherboard ASUS MAXIMUS XI APEX
RAM G.Skill Trident Z RGB F4-3200C14D-16GTZR, 2 x 8 GB (3200 MHz, CL14)
ROM Intel SSD 760p, 1024 GB
Power supply unit Corsair AX1200i 1200W
CPU cooling system Corsair Hydro Series H115i
Chassis CoolerMaster Test Bench V1.0
Monitor NEC EA244UHD
Operating system Windows 10 Pro x64
NVIDIA GPU software
NVIDIA GeForce RTX 20 NVIDIA GeForce Game Ready Driver 419.67
NVIDIA GeForce GTX 10/16 NVIDIA GeForce Game Ready Driver 425.31
Game tests
Game API Settings, test method Full screen anti-aliasing
1920×1080 / 2560×1440 3840 × 2160
Battlefield V DirectX 12 OCAT, Liberte mission. Max. graphics quality TAA High TAA High
Metro Exodus DirectX 12 Built-in benchmark. Ultra Graphics Quality Profile TAA TAA
Shadow of the Tomb Raider DirectX 12 Built-in benchmark. Max. graphics quality SMAA 4x Off

The average and minimum frame rates are derived from the individual frame render time array recorded by the built-in benchmark (Metro Exodus, Shadow of the Tomb Raider) or the OCAT utility if the game does not have it (Battlefield V).

The average frame rate in the charts is the reciprocal of the average frame time. To estimate the minimum frame rate, the number of frames generated in each second of the test is calculated. From this array of numbers, the value corresponding to the 1st percentile of the distribution is selected.

Test participants

The following video cards took part in performance testing:

  • NVIDIA GeForce RTX 2080 Ti Founders Edition (1350/14000 MHz, 11 GB);
  • NVIDIA GeForce GTX 2080 Founders Edition (1515/14000 MHz, 8 GB);
  • NVIDIA GeForce RTX 2070 Founders Edition (1410/14000 MHz, 8 GB);
  • NVIDIA GeForce RTX 2060 Founders Edition (1365/14000 MHz, 6 GB);
  • NVIDIA GeForce GTX 1660 Ti (6 GB);
  • NVIDIA GeForce GTX 1660 (6 GB);
  • NVIDIA GeForce GTX 1080 Ti (1480/11000 MHz, 11 GB);
  • NVIDIA GeForce GTX 1080 (1607/10000 MHz, 8 GB);
  • NVIDIA GeForce GTX 1070 Ti (1608/8008 MHz, 8 GB);
  • NVIDIA GeForce GTX 1070 (1506/8008 MHz, 8 GB);
  • NVIDIA GeForce GTX 1060 (1506/9000 MHz, 6 GB).

Battlefield V

With Battlefield V itself being a relatively undemanding game (particularly at 1080p and 1440p) and having patchy ray tracing, testing the GeForce 10-series with the DXR option was encouraging. However, of all the models without Ray Tracing support at the silicon level, we had to limit ourselves to the GTX 1070/1070 Ti and GTX 1080/1080 Ti models. Electronic Arts games react suspiciously to frequent hardware configuration changes and block the user for a period of one or more days. Therefore, performance measurements for the GeForce GTX 1060 and two GeForce GTX 16 series devices will appear later in this article, once Battlefield V releases restrictions on our test machine.

In percentage terms, any of the test participants experience approximately the same drop in performance at various ray tracing quality settings, regardless of screen resolution. Thus, the performance of video cards under the GeForce RTX 20 brand is reduced by 28-43% with low and medium quality DXR effects, and by 37-53% with high and maximum.

If we are talking about older models of the GeForce 10 family, then at Low and Medium ray tracing levels the game loses from 36 to 42% FPS, and at high quality (High and Ultra settings), DXR eats up 54-67% of the frame rate. Note that in many, if not most, Battlefield V game scenes, there is no discernable difference between Low and Medium settings, and between High and Ultra, neither in image clarity nor in performance. In the hope that Pascal GPUs would be more sensitive to this setting, we ran tests at all four settings. Indeed, certain differences appeared, but only at 2160p resolution and within 6% FPS.

In absolute terms, any of the top accelerators based on Pascal chips can maintain a frame rate above 60 FPS in 1080p mode with reduced quality of reflections, and the GeForce GTX 1080 Ti claims a similar result even with tracing at the High level. But once you jump to 1440p, only the GeForce GTX 1080 and GTX 1080 Ti provide comfortable framerates of 60 FPS and above in Low or Medium ray tracing quality, and in 4K mode, none of the previous generation cards has suitable processing power ( as, however, and any Turing with the exception of the flagship GeForce RTX 2080 Ti).

If we look for parallels between specific accelerators under the brand name GeForce GTX 10 and GeForce RTX 20, then the best model of the previous generation (GeForce GTX 1080 Ti), which is analogous to GeForce RTX 2080 in standard rendering tasks without DXR, dropped to the level of GeForce RTX 2070 with reduced quality ray tracing, and at high it can only fight with the GeForce RTX 2060.

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

Battlefield V Max Quality
1920×1080 TAA
RT Off R.T. Low RT Medium R.T High RT Ultra
NVIDIA GeForce RTX 2080 Ti FE (11 GB) Present in several = 100% -28% -28% -37% -39%
NVIDIA GeForce RTX 2080 FE (8 GB) Present in several = 100% -34% -35% -43% -44%
NVIDIA GeForce RTX 2070 FE (8 GB) Present in several = 100% -35% -36% -46% -45%
NVIDIA GeForce RTX 2060 FE (6 GB) Present in several = 100% -42% -43% -50% -51%
NVIDIA GeForce GTX 1660 Ti (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1660 (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1080 Ti (11 GB) Present in several = 100% -40% -39% -54% -58%
NVIDIA GeForce GTX 1080 (8 GB) Present in several = 100% -41% -41% -57% -61%
NVIDIA GeForce GTX 1070 Ti (8 GB) Present in several = 100% -40% -41% -57% -59%
NVIDIA GeForce GTX 1070 (8 GB) Present in several = 100% -38% -39% -57% -61%
NVIDIA GeForce GTX 1060 (6 GB) Present in several = 100% ND ND ND ND

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

Battlefield V Max Quality
2560×1440 TAA
RT Off R.T. Low RT Medium R.T High RT Ultra
NVIDIA GeForce RTX 2080 Ti FE (11 GB) Present in several = 100% -33% -34% -44% -45%
NVIDIA GeForce RTX 2080 FE (8 GB) Present in several = 100% -37% -38% -47% -49%
NVIDIA GeForce RTX 2070 FE (8 GB) Present in several = 100% -36% -36% -48% -48%
NVIDIA GeForce RTX 2060 FE (6 GB) Present in several = 100% -41% -42% -51% -52%
NVIDIA GeForce GTX 1660 Ti (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1660 (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1080 Ti (11 GB) Present in several = 100% -40% -40% -59% -62%
NVIDIA GeForce GTX 1080 (8 GB) Present in several = 100% -36% -39% -59% -63%
NVIDIA GeForce GTX 1070 Ti (8 GB) Present in several = 100% -39% -39% -58% -62%
NVIDIA GeForce GTX 1070 (8 GB) Present in several = 100% -38% -38% -59% -63%
NVIDIA GeForce GTX 1060 (6 GB) Present in several = 100% ND ND ND ND

New article: GeForce RTX no longer needed? Ray tracing tests on GeForce GTX 10 and 16 accelerators

Battlefield V Max Quality
3840×2160 TAA
RT Off R.T. Low RT Medium R.T High RT Ultra
NVIDIA GeForce RTX 2080 Ti FE (11 GB) Present in several = 100% -30% -30% -44% -47%
NVIDIA GeForce RTX 2080 FE (8 GB) Present in several = 100% -31% -32% -46% -49%
NVIDIA GeForce RTX 2070 FE (8 GB) Present in several = 100% -40% -38% -53% -52%
NVIDIA GeForce RTX 2060 FE (6 GB) Present in several = 100% -28% -30% -44% -53%
NVIDIA GeForce GTX 1660 Ti (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1660 (6 GB) Present in several = 100% ND ND ND ND
NVIDIA GeForce GTX 1080 Ti (11 GB) Present in several = 100% -36% -37% -60% -63%
NVIDIA GeForce GTX 1080 (8 GB) Present in several = 100% -40% -43% -64% -67%
NVIDIA GeForce GTX 1070 Ti (8 GB) Present in several = 100% -38% -42% -62% -65%
NVIDIA GeForce GTX 1070 (8 GB) Present in several = 100% -36% -42% -63% -66%
NVIDIA GeForce GTX 1060 (6 GB) Present in several = 100% ND ND ND ND

Source: 3dnews.ru

Add a comment