[ad_1]
Bench components
The latest pilots were used from 19/02/19. GPUs tested:
- NVIDIA RTX 2080 Ti
- MSI RTX 2080 Trio
- EVGA RTX 2070 XC
- Gigabyte RTX 2060
- EVGA GTX 1080 Ti FTW3
- EVGA GTX 1060 6GB SC
- EVGA GTX 980 Ti Hybrid
- EVGA GTX 970 SC
- AMD Radeon VII
- AMD RX Vega 56
- XFX RX 590
- MSI RX 580 8GB Gaming X
- Gigabyte RX 570 4GB
- AMD Fury X
The following is the script of our video. If you prefer to read, continue, but when you encounter references to "on-screen" or video footage, check the video above.
Research and practice mode – Apex Legends
The most important thing to do before starting to test dozens of configurations is to set a certain level of reproducible result in the game. Everyone plays in multiplayer, but the map is huge: it includes desert regions, regions grassy and mountainous, a river with semi-flooded buildings and the first fall of the sky with a more complete view of everything. The number of frames per second is ubiquitous in this game, which makes it functionally impossible to compare with absolute precision of absolute numbers. By simply browsing the clips in our video, you can see how varied the environments are. The best we can do is to establish a reliable benchmark that accurately scales relative performance, determining the "best" choices in each price category, rather than trying to look for absolute performance. everywhere in the game
Let's start with a 4K performance study to check consistency of matches. We will look at frametime charts after browsing the framerate charts.
With 1080 Ti and 4K at High settings, we observed an average SPF of 47 when we fell in the match, with the framerate in particular pointing down when looking at the landscape at a cast angle. . Viewing angles perpendicular to the ground have greatly stimulated framerate – not shown here – but wider views of the landscape and horizon have reduced the SPF to about 47FPS AVG. For 5 minutes in the hills and burned forest, we controlled the framerate at about 60FPS AVG, with minimums at 45FPS 1% and 39FPS 0.1%. It was just a coincidence if the average was 60 frames per second (Vsync and Adaptive Sync were off). In some game areas, as we will show in the game-time tracks, we have seen peaks of up to 70-80 AVP / AVG, with minimums never falling below 52FPS AVG. It seems that the framerate range is relatively constant when the graphics processor is subjected to such a constraint, but our 1080p tracking test will show that the framerate is less consistent when the constraint is removed from the pixel processing pipeline.
Finally, by testing with the training mode instead of the multiplayer mode, we find that our frame rate is high: we are at 73FPS AVG instead of 47-60FPS. At the end of the day, we consider that the loss of FPS is a minimal concern, as it only happens once per game and slight adjustments to the parameters to display the distance and the scaling level of detail can contribute. The 60 range is what we are looking for to get the maximum accuracy in multiplayer. 73FPS is close – it's reasonable, but – it's clearly not realistic. It is a performance gap of 20%, and even ignoring it, the question of whether other cards evolve accurately in practice mode remains. The answer is no, as we will show you later. So we finally abandoned the training mode to opt for multiplayer tests. In multiplayer mode, as long as the tests are performed exactly the same place each time, they are accurate, from one run to the other. The tests only vary and become variable when tested in random locations. As long as control is exercised, the multiplayer mode is a much better test environment for accurate data. It's ultimately landscape and prefabricated objects that impact performance, not necessarily actors, and test paths can simply prevent players from solving this problem.
At 1080p, the initial drops are placed between 126FPS AVG and 137FPS AVG, this difference is mainly due to the fact that we lowered the head or a wider angle of view of the horizon. Our two 5-minute test passes covered the entire map, tracing between 162FPS AVG and 182FPS AVG, with functionally equal sockets between them. The equality between the lows is good for testing and shows that few actions in the game cause severe falls. Nevertheless, the difference between these two figures allows the 2nd test to achieve an improvement of 12% compared to the first one. This is a big step forward that clearly shows that simply testing randomly on a multiplayer map is not enough for an identical comparison. The practical mode also poses a problem, even with our heavier load load during the tests at 206FPS AVG landing. This gives it a 13% lead on the fastest gaming test – not unreasonable, really – or 27% on the slowest test. This is where the real challenge lies.
For these reasons, we will perform multiplayer tests in a heavily loaded controlled area, using the river region of the map. This will give us the data that best represents the actual gaming experience, but that remains controllable for variables and reproducible during testing.
Apex Legends Frametimes
For a 1080p framing between the two test passes in play, we now get the chart. Do not forget that frames are representative of the interval between frames. So we're looking at about 50,000 data frames and trying to keep an eye on the above-average peaks. Spikes higher than 8-12 ms become perceptible by the player. Consistency is the most important factor, with the most important time being the most important time. These two tests were real gaming sessions without death for 5 minutes, with no particular spirit for specific test runs. They have been tested in different areas of the game. As you can see, playing as a real player, the two players do not deviate much from each other in real time. The red line is more thorny on average, which leads to the worst overall average SPF. It also rises above the blue line towards the beginning, where we start to hit 10-12ms while the blue is still at 6-8ms in an instant.
Both are perfectly acceptable, of course, but the goal is to show how these previous averages are created and that the game has some natural variance depending on the type of game encountered. It's impossible to simulate this game realistically – you can not get exactly the same test path every time in multiplayer, so the single-player mode remains our best current option for consistency. A reliable test is more important than a random test that seems closer to the actual game, because we can not fairly judge one card over another. Keep in mind that the absolute numbers displayed in the next tables should be considered lower for the actual multiplayer mode. instead, we suggest focusing on the relative scaling.
Apex Legends 1080p & High Reference Settings
Finally, here are the results: 1080p with multiplayer tests near the fluvial part of the map, where we see a medium to heavy load without introducing point charge inducers such as grenades – for more. information about it after the cards. Also, note that we have completely disabled the framerate limit.
The 2080 Ti ends at around 161FPS AVG, with the Radeon VII card nearby at 153FPS AVG. We did not trust these numbers, so restart the analysis of the two devices, then again at 2080 Ti for parity. In the end, the numbers are repeated. The Radeon VII card hardly falls as the resolution increases, curiously and contrary to what one would expect with its memory bandwidth, but this game seems to work well in 1080p compared to 2080 Ti. We may be experiencing a processor framerate limit of 1080p, but our peak FPS numbers are worth 240FPS and more, so these tips will not be shaved too fast. While some of the intensive attraction appeal areas may use more CPU resources, on average we are quite far from the FPS bottleneck. Radeon VII seems to be doing well here.
The Trio 2080 finishes at 145FPS AVG, which gives an advance of 11% to the 2080 Ti and the Radeon VII at 5%. The 1080 Ti is further behind the 2080 than in the other titles, ending at 128FPS AVG at 145FPS of 2080, producing a gap of 13%, favoring the 2080 and matching the performance of the 2070. For this title, it seems that a RX 580 with a light overclock – that is to say a RX 590 – or a RTX 2060 would be a good modern option for medium to high price categories. The 980 Ti still holds up, ranking at 80fps AVG with 1080p / High settings, and the 970 seems to be doing fine. Keep in mind that the settings you use can be ignored to improve performance. Like the RX 570, the Fury X encounters consistency problems at the time of cadence, while the frame rates are low and they fall below the average distance from the average.
Apex Legends 1440p GPU Benchmark: RTX 2080 vs. Radeon VII, Vega 56, GTX 970, etc.
1440p has the 2080 Ti at around 124FPS AVG, so we are seeing a rise of about 62% over the AVG FPS of our next 4K figures. The poor performance is still constant and good on the 2080 Ti, with sometimes a few milliseconds on average. The 2080 Ti is 12.7% ahead of the RTX 2080, about 3.4% on the Radeon VII. This corresponds roughly to the performance we observed for the Radeon VII during our first exam. For the rest, the 2060 seems to be a division line against the bottom of the table, offering a good upgrade path for those who want to play Apex Legends at 1440p with high settings. 82FPS AVG with a minimum around 60FPS is a largely acceptable performance. It is interesting to note that the performances of Fury X and RX 570 are enormous, the limited VRAM of Fury X seeming to cause a considerable impact on the regularity of the shooting, falling to 22FPS 0.1% and 36FPS 1%. The RX 570 has a similar behavior. The GTX 970 does not seem to have the same impact, but we think it's because it's so busy elsewhere in the pipeline that memory is less of a problem. The card has trouble working even at 30FPS AVG for 1440p / High.
Apex Legends 4K Benchmark – RTX 2080 Ti vs. RTX 2080, 1080 Ti
At 4K / High, the 2080 Ti ends at 76FPS AVG, standing about 16% ahead of the RTX 2080 Trio and 34% ahead of the Radeon VII. The 2080 ends 16% in front of the Radeon VII card, functionally related to the 1080 Ti. Interestingly, in this game, we actually see the 2080 place itself reasonably above the 1080 Ti. In our launch tests, the 2080 was closer to parity than the 1080 Ti in almost every game, and sometimes late. Here, the 2080 leads the 1080 Ti. Speaking of parity, the RTX 2070 is not far from the 1080 Ti in these tests, the 1080 Ti having a 3.4% advance. This also corresponds to what we saw in the test tests, even though we consider that these tests are not valid for a broader performance analysis due to scale limitations.
The RTX 2070 leads the 2060 by around 20%, and at this point, we probably recommend that the 2060 switch to much lower settings or 1440p. The 2070 does not pose a problem, but in a game as fast as Apex Legends, it would probably benefit from a slight reduction in settings. Vega 56 is struggling to keep up the pace here, although her real-time performance is still good in this title, so she deserves praise. The R9 Fury X is behind the Vega 56, with lows falling in the inconsistency, probably because of the limitations of 4 GB. The X Fury also ends up being tied with the aging 980 Ti and King, which is only 8% ahead of the XFX RX 590.
Obviously, these cards are not really for the 4K game, but some of them could play Apex Legends at a speed close to 60 frames per second with enough reduction in graphics quality – it would probably be better to play at 1440p.
Grenades
These numbers are accurate for average multiplayer use and can therefore be used reliably, even for relatively absolute tests. This will become less true if the patches start or if the game changes significantly, but for now, they can be used both as absolute and relative performance scale numbers. This becomes a little less true in the most heavily loaded scenarios, but for the vast majority of multiplayer players, the numbers are good performance. An example of where our numbers would no longer be reliable as an absolute metric would be the one involved in very heavy fights where several grenades are thrown. For example, the fire grenades we tested in practice, where you have an infinite amount, would lose about 23% of their flow by amplifying several in a row. This is a big performance problem that has not been taken into account in our graphics, so keep that in mind for decision making. If you want to be sure that objects reach a specific target of SPF in all scenarios, including a high number of fire grenades, consider an additional 20% drop in SPF. To be fair, this scenario will not be common, but it is an important scenario to know.
Example of scaling performance tests (invalid test path)
Finally, here is what a test in training mode would look like. First of all, let's be very clear that this is invalid data – we will also mark the graph in the same way, because otherwise people would stick it everywhere without listening and presume it was accurate. This data is only valid for training mode, but is not suitable for multiplayer mode when you are looking at card-to-board relative performance. Fury X is a good example: in multiplayer, we encountered frametime consistency issues due to memory limitations, but these limitations did not appear in the much more limited practice mode. This is a huge difference in the game experience, even ignoring the higher average SPFs. The next difference is the TI 2080, where we see almost 300 frames per second in practice, but about half of that in multiplayer. Even if the scaling was perfect card-to-card – which is not the case – it is a point where it is so far from reality that the numbers are no longer useful.
This scaling is quite different. The Radeon VII, for example, is nowhere near where it should land, the Ti 2080 behaves badly in an unusual way compared to the 2080, the Fury X does not suffer as it should, and and so on. Some numbers, such as 580 versus 1060, are actually relatively representative of reality, but they are still imperfect. In the future, we will test only with our multiplayer test path, because the drive mode is just not representative of the game. We hope that Apex Legends introduces a read mode to facilitate tests, provided that it is accurate.
Conclusion
Apex Legends is relatively lightweight to operate at high settings, despite some very painful scenes. Heavy fire grenades will charm FPS, but they are also rare in real multiplayer games. The drop-in is also intense if you look towards the horizon (longer view distances), but this only happens once per game. Overall, this leaves us with gameplay in the field, and is globally playable on just about everything in the class ~ $ 150 and up to 1080p / High. Of course, there is more room for FPS increases if the settings are lowered, but the High settings are not so intensive.
Editorial, responsible for the test: Steve Burke
Additional tests: Patrick Lathan
Video: Andrew Coleman
[ad_2]
Source link