At the beginning of July, Microsoft together with AMD and NVIDIA presented the so-called Hardware Accelerated GPU Scheduling, known in Spanish as “Hardware Accelerated GPU Scheduling” within Windows itself. We already did the corresponding article explaining with hair and signs what it consisted of and the advantages it was going to bring for gaming (previous link). After some problems with first-time drivers, we believe that it is already fully implemented and the time has come to test it, is it worth it in gaming? Is performance gained? Let’s see it.
First of all and so that no one wonders why, the tests were developed with an NVIDIA GPU and not AMD. The answer to this is simple: today AMD does not give official support for this feature and therefore it cannot be enabled or disabled in Windows as it does in the case of Huang’s, so once again we only have an option available to test for reds delays.
Similarly, NVIDIA has not had it easy either, the first supported driver left serious doubts and the following ones have emphasized offering better performance. As is often the case with these things, being late can mean arriving better, since the data is more conclusive when everything works as it should.
This is our case and after recapping several scenarios and games, we will update when we can with more hardware available in successive articles, since it affects a multitude of configurations and it will not be easy or fast to analyze them more or less in volume and in a representative way .
Hardware Accelerated GPU Scheduling, a step forward or harmless for gaming?
We are not going to explain the technology itself, (that’s what the article linked above is for), so we are going to go directly to the data, not without first naming our test team:
- Intel Core i7-8700K (Stock)
- ASUS Maximus X Formula
- Corsair Vengeance Pro RGB 2 x 8 GB 3200 MHz C16 (XMP ON)
- ASUS STRIX RTX 2060 OC
- Corsair MP510 960 GB
- ASUS STRIX Helios
- ASUS Ryujin 240
- Corsair AX1200i
- EK Vardar EVO RGB x 7
The games tested have been Shadow of the Tomb Raider, Battlefield V and The Witcher 3, all under 1080p resolution and with different presets and APIs, so that you can see how HAGS affects performance in multiple scenarios under the same hardware, already that, in theory although our CPU is very fast for gaming, a high FPS rate could generate a certain bottleneck, at least theoretically according to Microsoft.
Of course, the operating system used has been Windows 10 version 2004 19041.450 and the drivers used in our RTX 2060 are the most current at the time of writing: 452.06 WHQL . That said, let’s go with the data and its analysis, will Microsoft be right or will low-end hardware be necessary to see its benefits?
Shadow of The Tomb Raider
As always, both graphs are representative of each other, they belong to the same sample collected and are therefore complementary. What we can see is that the FPS are quite similar to each other, but there seems to be more stability between them if HAGS is active. The average shows this, 92.2 vs 92.3 FPS, the minimums also equaled by the minimum and the maximums saved by the specific peak are not representative as such.
Regarding the FrameTime, very even, very compact and linear until the scene changes, where it is triggered in both cases, there does not seem to be a clear winner when looking at the graph and it would be necessary to enlarge by sections to really see differences. The averages also demonstrate this with minimal differences, at least under DX12.
Their percentiles seem to give a slight advantage when HAGS is active, but they are still fairly even and we will have to wait to see the data regarding the variations to get something clear.
Little difference in 0% and 1% Low, two and three FPS above for HAGS ON, which, although it is already something appreciable as gain proper, is really the biggest difference that seems to be seen in this title under the Ultra preset. On the other hand, the deviations and variability of FPS and Stuttering do not show what we expected, that is, some difference that we decided to opt for on or off. And it is that the differences are to the minimum in both cases, so the most representative data in this scenario continues to be the differences that we can find in 0% and 1% Low, something that is not decisive to say if it is worth activating it or disable it.
What we make clear is that under DX12 in 1080p and Ultra it seems that leaving it active does not deteriorate the performance and that instead we can gain something.
Battlefield v
Here we have selected a completely different scenario. Battlefield V is a very demanding game on any PC, but its optimization leaves a lot to be desired when it comes to spitting FPS under DX12, so we have opted for DX11 which offers less problems and more consistency under 1080p and a Low preset, which will shoot the FPS and will strain the CPU. So much so that we have reached the maximum rate of the game, 200 FPS, after which it varies while we are playing.
It is difficult to explain how HAGS works here, since, although we will only have 3.2 FPS of final difference between both modes, the minimums are much better when we activate it and the maximums do not change too much, but it is true that when in OFF maintains 200 FPS without problem, in ON those FPS are much lower. On the other hand, in the global computation there is more stability of frames, curious.
Analyzing the FrameTime, it seems that in ON it is more compact, which means more smoothness. Saving two specific moments, both settings remain very similar in terms of milliseconds, but it must be pointed out that, although the average and the minimums are practically nailed, the maximums have shot up in the case of having it deactivated. It is true that it has been for less than a second, but that has represented a very clear jerk on the screen, a drop in FPS that having it active does not occur. In addition, there are more high peaks with OFF than with ON, so it seems that the stability with the parameter active is greater and softer, something that we will check later.
Their percentiles are pretty even if we just look at the average, but actually the graph shows that in OFF there are more global samples at lower milliseconds. The problem is that there are more samples with high milliseconds in OFF than in ON, so the final result ends up equalizing.
The following two graphs are much more representative, since in the case of 0% and 1% although the average only shows the aforementioned 3 FPS, the minimums in both cases are much better for ON than for OFF with 6 and 15 FPS respectively . To this must be added the chilling difference in deviations in FPS and Stuttering.
With HAGS activated the values are on average 60% better than if we kept it deactivated in Battlefield V. That is, we will have much more stability in FPS and much more smoothness on the screen with HAGS ON, in addition to the aforementioned rise in FPS in Low.
The witcher 3
The Witcher 3 is a very well optimized game that is curiously affected by very small details. You have to have an optimal and very specific configuration for it to go tremendously fluid, which is almost more important than achieving a high FPS rate, because saving Hairworks, the game spits FPS without too many problems as the High preset shows.
Again FrameRate and FrameTime are very close in both HAGS settings. It should be mentioned that in the 31-32 second the peak that can be seen in both graphs corresponds to a horse that crossed our path and that made the data collection vary, not easy to nail it in this way, by the way. But saving this, the timing is perfect as can be appreciated. It should be noted that, although the FPS are very equal in almost the entire section, the FrameTime does seem to show more stability with ON compared to OFF, at least that is how it is seen when seeing the green lines behind the orange ones, which means worse smoothness. Again we will have to wait for the following data to draw conclusions, because in the rest they are totally equal.
The percentile is practically perfect in both settings, so there is little we can comment on here, let’s see what it is capable of doing in the% and variability.
Here we also have equality, although again HAGS ON beats OFF. The 0% and 1% Low are even with -1 and +2 FPS respectively, but the deviation of FPS and Stuttering give HAGS ON the winner also by the minimum. The stuttering is the same in both settings, but there is more homogeneity in the FPS, which will make our perception of the game on the screen more pleasant, more constant, something that is not decisive in this game, but that we always like.
Conclusion on Hardware Accelerated GPU Scheduling
At the moment and according to what we have seen and tested, either with G-SYNC or G-SYNC Compatible, is that Hardware Accelerated GPU Scheduling does not bring a counterproductive effect to the PC or the gaming sensation. Rather the opposite, it improves almost everything and in some cases by a lot. It seems that in poorly optimized games this is enlarged and it is possible that in more resource-efficient teams these are amplified even more, so it is a win-win by Microsoft and a success that NVIDIA already offers it against AMD.
In the case of greens, it is clear that it is improving and that any gamer should leave it active and try at least in their favorite games to see how it works. It cannot be categorically assured that what is seen here is extrapolated to any game as such, but there are three triple A titles with different scenarios, graphic load, API and FPS, so the CPU reacts differently and in all cases it has been positive, which creates, at best, a precedent. So since it only costs a PC restart, why not give it a try? It’s free and we may earn more than we think.
