Do AMD cards still hitch and stutter more than Nvidia?

[I hate you]

The one thing that always kept me from buying AMD cards is that they stuttered more than nvidia cards. I don't mean low framerates, I mean intermittent short freezes in-game, or more technically frametime spikes. I always knew about it from youtube comparison videos even though AMD themselves claimed they didn't (they probably only pretended they didn't).

Here is an article from 2013:
http://www.anandtech.com/show/6857/amd-stuttering-issues-driver-roadmap-fraps/5
It explains how AMD finally discovered their cards had more stutters than Nvidia, so they started releasing new drivers with fixes and workarounds for the issue.

Is the problem still there? Who stutters more at the present? AMD or Nvidia?

 

[tamz_msc]

Almost all reputable review sites nowadays publish the frame times during their benchmark runs in their GPU reviews.
You can have a look at them and come to your own conclusions. One fact that has been established is that frame time variances are much higher in the newer APIs like DX12 and Vulkan.

 

[MajinCry]

There was a big hubbub about it, back when the 7000 series was new. Didn't AMD then release a few drivers, that fixed it? Dunno if the older (<6000 series) cards got the same lick of paint, tho'.

 

[nurturedhate]

That issue was fixed a long, long time ago. Frankly I've had more issues with drivers for my Nvidia cards (780ti, 1070) than I had with my AMD cards (7870, 390x). Geforce Experience causes issues in some games, primarily Starcraft 2. Frame drops in WoW. Frostbite engine games and random crashing. They aren't massive issues and a newer or older driver install would always fix it but just having the thought of "what is this new driver going to break" is never a good thing.

 

[IEC]

AMD cards actually got better frametimes for a while... Now, outside of anomalies, both sides generally do a very good job with frame times. Single GPU configurations don't stutter.

That said, certain combinations such as dual core processors (e.g. $65 Pentium G4560) combined with games like Rise of the Tomb Raider can and will lead to stutter / low minimum FPS. Provided you don't have a bottleneck elsewhere, even a Radeon RX470 is perfectly adequate for smooth 1080p gaming.

Hardware Unboxed did a comparison of GTX 1060 / RX 470 / GTX 1050 Ti. You can see for yourself the smooth performance of the AMD card:
https://www.youtube.com/watch?v=dichjs9HXTg

Even with the dual core bottleneck, the RX470 is the king of budget gaming, coming close to the GTX 1060's performance while costing as little as $115 AR (that's literally less than half price).

 

[greatnoob]

AMD cards are much better than Nvidia counterparts when it comes to frame times and overall tick consistency:

https://youtu.be/RdxaG87XcCs

 

[I hate you]

Almost all reputable review sites nowadays publish the frame times during their benchmark runs in their GPU reviews.
You can have a look at them and come to your own conclusions. One fact that has been established is that frame time variances are much higher in the newer APIs like DX12 and Vulkan.

Interdasting.
First result in google search is guru3d:
In most games they found frametime spikes to be identical, but in the few ones that were different Nvidia always did better than AMD:

 

[tamz_msc]

You came to that conclusion from those two graphs?

 

[nurturedhate]

You came to that conclusion form those two graphs?

Nor did he provide a link for context. We can make data points say about anything we want without context.

 

[Headfoot]

The new APIs still seem to have some teething issues around frametimes, pretty much across the board except Doom. That leads me to believe its a developer issue primarily

 

[I hate you]

It's all I could find on such short notice. You can add input too you know?

 

[I hate you]

Nor did he provide a link for context. We can make data points say about anything we want without context.

http://www.guru3d.com/articles-pages/radeon-rx-480-vs-gtx-1060-fcat-frametime-analysis-review,1.html

 

[Yakk]

The new APIs still seem to have some teething issues around frametimes, pretty much across the board except Doom. That leads me to believe its a developer issue primarily

Yup, usually is developer/engine related. In DX11 & below AMD & nvidia employ driver hacks after the fact to bypass a lot of developer unoptimized/garbage code.

DX12 & Vulkan change this dynamic a lot and game engine developers need to be a lot more knowledgeable and involved upfront. AMD & nvidia can fix and/or bypass a lot less bugs in their driver releases, so now bugs need the developer to issue updates themselves. As we saw in the first wave of games using the news APIs some developers are having quite a bit of trouble adjusting to this new dynamic.

 

[tamz_msc]

On average, across all sites, if you can find one game where frame time variances are higher in AMD cards, then it is probably equally likely that you will find a game where the NVIDIA card has higher frame time variances.

All benchmarking results are indicative in nature, because they are almost impossible to replicate independently. In those graphs you posted, you can never guarantee that those spikes you get will be there in another benchmark run.

I mean, you can catalogue all the graphs you want to your heart's desire, and that won't help you decide in favor of either NVIDIA or AMD.

 

[Flapdrol1337]

http://techreport.com/review/30812/nvidia-geforce-gtx-1060-graphics-card-reviewed

Seems pretty similar comparing 1060 6GB and 480.

The 470 and 3GB 1060 seem to hitch a bit now and then, I guess having lots of memory is more impactful than brand/driver.

 

[SlickR12345]

Its pretty much the same, with Nvidia being slightly favored, though AMD seems to have higher max frames in some games. So essentially you can't really go wrong, though Nvidia does tend to have a little better frame times, while AMD a bit more, with AMD getting more max frames per second on average, Nvidia getting a bit better minimums on average.