Our Sponsors

Thursday, January 24, 2019

Nvidia vs. AMD GPUs when used with an Adaptive-Sync display, how...



Nvidia vs. AMD GPUs when used with an Adaptive-Sync display, how they compare | Part 2 of 2

Contents

  • Introduction
  • Test Setup and Methodology
  • Monitors Tested
  • HP Pavilion 27q
  • Nixeus NX-EDG27
  • Nixeus NX-EDG27240
  • G-Sync Compatibility Certification
  • Conclusion

Introduction

This is a conclusion to the prior post linked here - https://www.reddit.com/r/hardware/comments/afqnei/nvidia_vs_amd_gpus_when_used_with_an_adaptivesync/

The purpose of this was to define, explain, test, and validate the differences between using an AMD GPU and an Nvidia GPU with an adaptive-sync capable display that is not G-Sync Compatible certified. And while numerous tests from respected reviewers and the community at large have already shed a lot of light on this subject, I feel as if I should finish what I started by posting my findings here. For those of you who have already read through more than enough tests, I hope there’s something new and interesting in this post to make it worth your while.


Test Setup and Methodology

Testing on all monitors was done with an EVGA GTX 1060 SSC 6GB and a Sapphire RX 580 Pulse 8GB. The purpose of using these two models was to ensure that performance was virtually identical. If I had used two GPUs with different performance levels, I may have difficulty telling if the additional fluidity was due to adaptive-sync, or just superior frame pacing from one of the GPUs.

The game World of Warcraft (DX 12 API) was used for all of my testing. WoW and a few others games made my list because their visual environment allows for easy detection of tearing and judder with V-Sync off. Benchmarking shows that the 1060 and 580 had virtually identical performance in specific areas, within 1-2% of each other. This used to not be the case, but the 8.0 and 8.1 patches brought with it significant improvements for AMD hardware.

Nvidia driver version 417.71 was used as this is the first driver with adaptive-sync support. AMD driver 18.12.2 was used as it was the most recent WHQL release at the time of testing. Had I been doing performance benchmarks, a newer driver would have been used.


Monitors Tested

While at least one of the displays listed below supports Freesync over HDMI, I did not test this. Nvidia’s implementation only works over DisplayPort.

. HP Pavilion 27q Nixeus NX-EDG27 Nixeus NX-EDG27240
Size 27" 27" 27"
Resolution 2560x1440 2560x1440 1920x1080
Panel Type IPS (PLS) IPS (AHVA) TN
Refresh Rate 75hz 144hz 240hz
VRR Range 48-75hz 30-144hz 35-240hz
Adaptive Overdrive No Yes No
LFC No Yes Yes

Most of the above is self-explanatory, but I’ll try to explain anything that seems out of sorts or uncommon.

PLS is Samsung’s variation of IPS technology. AHVA is AU Optronics’ (AUO’s) version of IPS.

Adaptive Overdrive is the display’s ability to adjust the overdrive on the fly based on the currently displayed refresh rate/frame rate. All G-Sync module-based displays have this feature in hardware (called variable overdrive). So far the Nixeus NX-EDG27 is the only Freesync certified display confirmed as having the feature. It is not yet clear to me if the method used by the EDG27 is the same as what Nvidia employs.

LFC is AMD’s branding for the feature that allows a monitor to multiply the refresh rate over the frame rate to allow for additional VRR support. Basically, the panel/scaler can’t operate below the minimum VRR range, so instead, a multiple is used. For example, if the range is 30-144hz and you run at 28fps, the display will run at 56hz. For the HP Pavilion 27q, LFC is not supported because the range is too narrow, IE, 47fps would be 94hz, which is out of range. G-Sync displays do this automatically, despite the lack of additional branding. Nvidia’s driver supports it in G-Sync Compatible mode for both certified and non-certified displays, provided the range is wide enough.


HP Pavilion 27q

Product Page | https://store.hp.com/us/en/pdp/hp-pavilion-27q-display

Please note that this display was purchased with my own funds and not provided by the manufacturer.

  • By default, the monitor has adaptive-sync disabled. Running at 75hz may cause frame-skipping as the scaler isn’t designed for that. My sample worked fine at 75hz, but this is not normal behavior and should not be expected in the default setup.
  • Enabling “Gaming-Freesync” under View Modes makes adaptive-sync work. You lose the ability to adjust the RGB values in the OSD. Color accuracy takes a hit as well. Delta E Average went from 1.62 to 3.4. Delta E Max went from 5.3 to 11.84. Contrast took a small hit, going from about 950:1 to 920:1. However, 75hz works with no frame-skipping. This is an intended tradeoff.
  • Once enabled in the OSD, AMD’s drivers automatically enable Freesync and reflect a range of 48-75hz, or 48-60hz if you drop down to 60hz.
  • Nvidia’s drivers require that you enable G-Sync for the selected display (one additional step over AMD). I could not find a spot within the driver that shows the VRR range.
  • Both GPUs performed identically within the stated range.

Conclusion: Performance is identical for both GPUs. Nvidia requires an extra step for enabling adaptive-sync compared to AMD. The display’s picture quality takes a small but noticeable hit when Gaming-Freesync is engaged. This allows the scaler to run at 75hz without frame skipping.


Nixeus NX-EDG27

Product Page | https://www.nixeus.com/nx-edg27

Please note that this display was provided free of charge by the manufacturer for the purpose of this comparison.

  • The monitor does not have a toggle for adaptive-sync. It is always on and available to the GPU.
  • AMD’s drivers have Freesync on by default and reflect a range of 30-144hz. Nvidia’s drivers do not show a range, and must have G-Sync enabled manually for the selected display.
  • Both GPUs performed identically within the stated range. The monitor’s OSD allows you to look at the currently displayed refresh rate, which should closely match the in-game frame rate if within 30-144 (it may not be an exact match due to difference in polling rates).
  • When below 30fps, the monitor’s refresh rate was a multiple of the currently displayed frame rate, confirming that LFC worked fine on both GPUs.

Conclusion: Performance is identical for both GPUs. As with before, Nvidia requires one minor, additional step for adaptive-sync to work.


Nixeus NX-EDG27240

Product Page | https://www.nixeus.com/nxedg27240

Please note that this display was provided free of charge by the manufacturer for the purpose of this comparison.

  • The monitor does not have a toggle for adaptive-sync. It is always on and available to the GPU.
  • AMD’s drivers have freesync on by default and reflect a range of 35-240hz. Nvidia’s drivers do not show a range, and must have G-Sync enabled manually for the selected display.
  • Both GPUs performed identically within the stated range. The monitor’s OSD allows you to look at the currently displayed refresh rate, which should closely match the in-game frame rate if within 35-240 (it may not be an exact match due to difference in polling rates).
  • When below 35fps, the monitor’s refresh rate was a multiple of the currently displayed frame rate, confirming that LFC worked fine on both GPUs.

Conclusion: Performance is identical for both GPUs. As with before, Nvidia requires one minor, additional step for adaptive-sync to work.


G-Sync Compatibile Certification

Nvidia has a strict set of requirements for G-Sync Compatible certification. The exact requirements are not made public, but we can guess a few based on their statements and prior announcement at CES.

https://www.nvidia.com/en-us/geforce/news/g-sync-ces-2019-announcements/

G-SYNC Compatible testing validates that the monitor does not show blanking, pulsing, flickering, ghosting or other artifacts during VRR gaming. They also validate that the monitor can operate in VRR at any game frame rate by supporting a VRR range of at least 2.4:1 (e.g. 60Hz-144Hz), and offer the gamer a seamless experience by enabling VRR by default.

Right off the bat we can use that to explain why the HP Pavilion 27q would not pass certification. Its range of 48-75hz doesn’t meet the 2.4:1 ratio specified by Nvidia. Additionally, adaptive-sync is a toggle in the monitor’s OSD that is off by default.

But the two Nixeus monitors tested don’t seem to have any glaring weaknesses when matched up against that rubric. Adaptive-sync is always on and not a toggle. The ranges greatly exceed the 2.4:1 ratio. There are no visual artifacts and ghosting is held in check. So, I reached out to Peter Trinh, Director of Product Development at Nixeus Technology for a comment.

On whether the Nixeus displays were tested by Nvidia:

I’m not sure as Nvidia never reached out to us and we never provided them with our monitors to the best of my knowledge. They never reached out to us with some questions or confirmations, too.

And on the ghosting, specifically, was adaptive-overdrive employed on the NX-EDG27240?

No, it does not due to our FreeSync Certification testing - the NX-EDG27240/NX-EDG27240S and NX-VUE24A/NX-VUE24B displayed ghosting in our testing with FreeSync ON with the monitor’s Over Drive set to OFF so we decided to market them without Adaptive-Over Drive. With the monitor’s Over Drive set to Low or Med, ghosting was eliminated or minimized with FreeSync ON. However when we tested the same monitors with Nvidia’s G-Sync compatibility drivers, there were little to no ghosting when those monitor’s Over Drive setting was set to OFF with our internal test tools.


Conclusion

It’s tough to draw a definite conclusion from a sample of only three monitors. However, care was taken in choosing these monitors. We have monitors with a narrow range and a wide range. We have LFC and no LFC. High quality scalers and a low-quality scaler running in a dual mode. Adaptive overdrive and standard overdrive. All that we’re missing is a Freesync 2 w/HDR sample, and a sample that runs the Freesync range in 2 modes (typically LG or Samsung), but I was unable to obtain any of those in time for testing.

Based on the testing here it’s fair to say that Nvidia GPUs SHOULD behave virtually identically to AMD GPUs with any display that uses the DisplayPort Adaptive-Sync Open Standard. The key exception is that Nvidia’s implementation doesn’t use the ported version over HDMI that AMD has made use of.

My takeaway from my testing and the testing I’m seeing elsewhere is this - users should feel safe in buying any adaptive-sync display (that works over DisplayPort), provided there’s no review or discussion showing that the monitor in question absolutely does not work with Nvidia hardware. Because both Nvidia’s driver and the adaptive-sync monitors are using the open standard, they should work together with the exceptions of shoddy implementations from the monitor manufacturer, or potentially a bug in Nvidia’s driver-based implementation.



from Daily Technology News http://bit.ly/2TeOL7u
Click Here For More News!

No comments:

Post a Comment