[ad_1]
In autumn 2021, the Lyon division of Arkane Studios pleased us with the game Deathloop — an original project with a number of familiar elements from Dishonored. This is an immersive adventure in which we are gradually looking for the optimal solution to the main task. Two heroes confront each other on a small island within a closed time loop.
Visually and stylistically, the game resembles Dishonored 2, although the place of action has been moved to the middle of the 20th century. Deathloop uses its Arkane Studios engine called Void Engine, which is a modified version of id Tech 5. The game only works in the DirectX 12 API environment. Support for ray tracing for more realistic shadows and Ambient Occlusion is claimed. AMD FSR and NVIDIA DSR scaling available.


Consider the performance of the game using the example of several video cards. Let’s compare their results with different resolutions and evaluate the impact of tracing on overall performance. Let’s see the results with different scaling options.
Ray tracing in Deathloop
First, let’s talk briefly about ray tracing. Below is a comparison of the same scene with simple shadows and trace-based shadows.


With the «rays», the shadows from the towers have blurred contours, and here and there the darkening in the crevices of the rocky surface intensifies.


In another scene, it is even more difficult to notice the difference in the contours of the shadows, the difference is barely perceptible. How such minimal changes will affect performance will become clear from the results of testing.
Test participants
Five video cards were involved in the testing:
The top segment is represented by reference models AMD Radeon RX 6800 and NVIDIA GeForce RTX 3070 Founders Edition. Other video cards were adapted to standard frequencies.
Full technical specifications are given in the table.
Video adapter | GeForce RTX 3070 | GeForce RTX 3060 | Radeon RX 6800 | Radeon RX 6600 XT | Radeon RX 6600 |
---|---|---|---|---|---|
Core | GA104 | GA106 | Navi 21 | Navi 23 | Navi 23 |
Number of transistors, million pcs. | 17400 | 13250 | 26800 | 11060 | 11060 |
Process technology, nm | 8 | 8 | 7 | 7 | 7 |
Core area, sq. mm | 392 | 276 | 520 | 237 | 237 |
Number of stream processors | 5888 | 3584 | 3840 | 2048 | 1792 |
The number of tensor kernels | 184 | 112 | — | — | — |
The number of RT cores | 46 | 28 | 60 | 32 | 28 |
Number of texture blocks | 184 | 112 | 240 | 128 | 112 |
Number of ROP blocks | 96 | 48 | 96 | 64 | 64 |
Base core frequency, MHz | 1500 | 1320 | 1700 | 2064 | 1626 |
Game Clock frequency, MHz | — | — | 1815 | 2428 | 2044 |
Boost frequency, MHz | 1725 | 1882 | 2105 | 2607 | 2491 |
Memory bus, bit | 256 | 192 | 256 | 128 | 128 |
Memory type | GDDR6 | GDDR6 | GDDR6 | GDDR6 | GDDR6 |
Memory frequency, MHz | 14000 | 15000 | 16000 | 16000 | 14000 |
Memory capacity, GB | 8 | 12 | 16 | 8 | 8 |
Supported version of DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
Interface | PCI-E 4.0 | 16x PCI-E 4.0 | PCI-E 4.0 | 8x PCI-E 4.0 | 8x PCI-E 4.0 |
Energy consumption, Vt | 220 | 170 | 250 | 160 | 132 |
Further, the GeForce graphs show the full frequency range — from base to peak Boost value, for Radeon — the range from Game Clock to peak Boost value.
Test stand
The configuration of the test bench is as follows:
- processor: Intel Core i9-9900K;
- cooling system: be quiet! Silent Loop 280mm;
- motherboard: Asus Rog Maximus XI Formula;
- memory: Kingston Fury Renegade KF436C16RB1K2/32 (DDR4-3600);
- system disk: Kingston SSDNow UV400 480GB;
- additional drive #1: Kingston A2000 NVMe PCIe 1000GB;
- additional drive #2: Kingston KC2000 NVMe PCIe 1000GB;
- case: Antec NX800;
- power supply unit: Antec HCG850 Gold;
- monitor: ASUS PB278Q (2560×1440, 27″);
- OS Windows 10 Pro x64;
- AMD Radeon Adrenalin Edition driver 21.12.1;
- driver for NVIDIA GeForce 496.76.
Test run of the first location with a complex detailed landscape. Several repetitions. Measurements were made by MSI Afterburner.
Standard maximum quality profile is selected. Next, only the shadow settings were changed for additional activation of tracing.
Test results
Let’s consider the performance of younger test participants in Full HD resolution.
In normal mode without tracing, the Radeon RX 6600 XT is in the clear lead, and the GeForce RTX 3060 is only slightly ahead of the Radeon RX 6600. Activating shadows with tracing has a minimal effect on the results of the NVIDIA video card, while AMD’s performance drops significantly. This allows the GeForce RTX 3060 to come out on top. At the same time, all models of this class showed good performance in 1080p at maximum graphics settings. In normal mode, the game uses 7 GB of memory and more, with rays it can load more than 8 GB.
Let’s go to the 1440p format. We compare all video cards from the list of participants.
The Radeon RX 6600 is not enough for this format even without tracing, and the Radeon RX 6600 XT provides around 60 frames. The GeForce RTX 3060 has a stronger position — the video card is minimally inferior to the AMD competitor in normal mode, and shows a serious gap with tracing. At the same time, for a comfortable frame rate with rays on the GeForce RTX 3060, you will still have to turn on DLSS scaling. The older Radeon RX 6800 and GeForce RTX 3070 handle the game perfectly regardless of whether tracing is enabled or not. A senior AMD representative provides the best results.
The 4K format is difficult for the Radeon RX 6800 and GeForce RTX 3070. Even without tracing, they barely reach 60 frames, and with «rays» their results are 30-33% lower. On the AMD video card, the memory load in the most difficult mode reached 11 GB.
You can achieve comfortable indicators in 4K when scaling is enabled. AMD FSR (FidelityFX Super Resolution) and NVIDIA DLSS (Deep Learning Super Sampling) algorithms are available in the game. We compare the performance of Radeon RX 6800 and GeForce RTX 3070 video cards in different modes at 4K resolution.
When combined with maximum quality tracing and AMD FSR, the final performance of the Radeon RX 6800 is higher than in regular 4K without advanced shadows. Switching to normal quality FSR mode boosts results to 70fps and above.
For the GeForce RTX 3070, only the highest quality FSR and DLSS options are used. And here it is necessary to note the clear advantage of NVIDIA technology in the final performance. As a result, you will be able to comfortably play in 4K with tracing and DLSS Quality.
Conclusions
The use of ray tracing in Deathloop is not impressive. The difference between normal shadows and shadows with tracing should be looked for with a magnifying glass. At low resolutions, turning on tracing gives a small performance drop, especially on NVIDIA graphics cards. At high resolution, the frame rate can drop by a third, and enabling this option already looks like a dubious decision. Scaling will help increase performance in heavy modes. For NVIDIA video cards, it is better to use DLSS technology, for AMD and older video cards, you can use FSR. At the same time, it should be understood that with a decrease in resolution, the deterioration of the picture during scaling will be more obvious.
Video cards of the modern generation easily cope with the game in Full HD format — even Radeon RX 6600 is enough here. For settings with tracing, GeForce RTX 3060 and Radeon RX 6600 XT are enough. The latter allow you to play Deathloop in 1440p format as well, if you do not include enhanced shadows. The potential of Radeon RX 6800 and GeForce RTX 3070 allows you to comfortably play in 4K with scaling and «rays». Obviously, more powerful video cards give more freedom in choosing settings for maximum resolution, and flagship solutions can handle Deathloop without activating FSR or DLSS.
[ad_2]