Computex 2014: AMD Demonstrates First FreeSync Monitor Prototype
by Ryan Smith & Anand Lal Shimpi on June 5, 2014 6:00 AM EST- Posted in
- Displays
- AMD
- GPUs
- FreeSync
- Computex 2014
Our very own Anand Shimpi just got off of the Computex showfloor for a bit after paying a visit to AMD’s booth. Among the items AMD is showing at Computex is the current status of their FreeSync project, whose base feature of Adaptive Sync was recently added to the DisplayPort 1.2a standard as an extension.
First announced at CES 2014, FreeSync is AMD’s take on variable refresh monitors, utilizing some variable refresh functionality first designed for embedded DisplayPort (eDP). At the time AMD was showing off the concept on laptops (due to the need for eDP) but are back at Computex with an update on the project.
Here at Computex AMD is showing off the first prototype monitor that is FreeSync capable, which interestingly enough is based on a retail monitor that was hardware capable and could be converted with updated firmware. AMD’s actual demo hasn’t changed – they’re still running the fan blade demo we saw at CES 2014 – but it’s now running on external monitors. The monitor in question operates with a fairly narrow range of rates of just 40Hz to 60Hz, which for the purposes of a prototype is good enough to showcase that the technology works, though it is narrower than the refresh ranges AMD is expecting for retail monitors.
At this point AMD is emphasizing that while they were able to get FreeSync up and running on existing hardware, owners shouldn’t be expecting firmware updates as this is very unlikely to happen (though this is ultimately up to monitor manufacturers). Instead AMD is using it to demonstrate that existing panels and scalers already exist that are capable of variable refresh, and that retail monitors should not require significant/expensive technology upgrades. Meanwhile AMD is also holding to their earlier projection of 6-12 months for retail monitor availability with retail prototypes expected around September, which puts final retail availability potentially as early as very late this year, but more likely into the first half of 2015.
Finally we have a video interview of the FreeSync demo in action. It bears mentioning that YouTube is limited to 30fps, so while we can give you some idea of what FreeSync performs like it’s not a fully capable representation. That will have to wait for closer to release when we can sit down with a high speed camera.
Facebook
Twitter
RSS
44 Comments
View All Comments
Soulwager - Thursday, June 5, 2014 - link
Did they let you play with the tech demo and verify it's demonstrating what they say it's capable of? There was a report that freesync mode was running at a fixed framerate between 47 and 48 fps, which sounds very similar to what they had to show at CES.r3loaded - Thursday, June 5, 2014 - link
If AMD pulled off variable refresh rates so easily, then what does Nvidia's G-Sync do or offer that requires expensive hardware (particularly that FPGA board)? Does G-Sync have some innate technical advantage over FreeSync?nathanddrews - Thursday, June 5, 2014 - link
There's a lot of misinformation surrounding these technologies, but as far as I've been able to decipher any monitor that wants to variable refresh needs to have *some kind* of hardware inside to coordinate that process. NVIDIA saw an opportunity to provide a solution (proprietary, but possibly licenseable) via dedicated hardware upgrades to existing and future DP monitors. In response, AMD/VESA sought out a method that used existing technology (eDP) in order to bring it to future DP1.2a monitors. Either way, the hardware has to be there alongside software/firmware.I just hope we see variable sync cover all refresh/frame rates at all resolutions. Confining it to sub-60fps is nice, but tearing and stutter is also an issue at higher refresh/frame rates.
Alexvrb - Thursday, June 5, 2014 - link
I believe the primary point of these technologies is to deal with sub-<display refresh rate> framerates. If the demo monitor in question is a 60hz display, then it isn't going to display more than 60 FPS. If you're already getting >60 FPS on a 60hz display, then you can just cap the framerate or enable v-sync (unless you don't mind tearing).However, if you are either consistently below 60 FPS, or are dipping below it periodically, then this kicks in. Being able to handle sub-60 FPS on a 60hz display fluidly is a huge deal. Previously you'd get nasty artifacts or you'd have to enable v-sync, which is often even less desirable. In either case they seem to feel future retail units will have a larger range than 40-60 without tearing. How well this functions (like most things) might also vary from display to display.
Now, just because they're only demoing the technology on a 60hz display doesn't mean it won't work on a 120hz display. A 120hz display with Freesync might be able to smoothly handle a much larger range of framerates, too. The best part? It won't be a costly addition and leverages existing technology, thus every monitor with DP support will have it eventually.
Samus - Monday, June 9, 2014 - link
The difference is GSYNC can run your refresh rate in 1hz increments between say, 30-144hz, 144 times a second. So every frame is in sync with every refresh.Stobing and/or high refresh rates don't do that. They allow higher refresh rates (which prevents tearing as "V-SYNC" will work beyond 60FPS) they don't prevent stutter since the refresh isn't variable.
There are utilities that allow you to run 4 refresh rates "simultaneously" on compatible hardware with any video card that supports the maximum refresh rate you specify, but they are crude implementations and still don't completely prevent tearing, but they help.
Soulwager - Monday, June 9, 2014 - link
G-sync doesn't do 1hz increments like that, it doesn't even make sense to think of those increments in terms of Hz. Better to think of it in terms of how long you wait before checking for a new frame. The first check might be made assuming maximum refresh rate, so 6.9ms. After that you might just let your loop spin until it hits a new frame. Obviously there's some polling rate here, but the exact number could be anything from a few kHz to a few MHz. In any case, you'd probably measure the increment between possible g-sync refresh intervals in microseconds.Soulwager - Thursday, June 5, 2014 - link
My guess is AMD is having a harder time of it than they're letting on, and that the FPGA+onboard RAM was necessary to get the desired performance in a short amount of time.Long term, there may be advantages to keeping the RAM in the display, that have to do with reducing memory usage and overhead in the GPU, but the FPGA will certainly be replaced by an ASIC at some point.
Kevin G - Thursday, June 5, 2014 - link
RAM in the display could also be used for panel self refresh where the display tells itself to refresh when it does not receive an update. Enabling variable timing with panel self refresh sounds like a natural extension.Soulwager - Thursday, June 5, 2014 - link
Well, you either have to do panel self refresh, or you have to re-send frames before the panel fades. G-sync does panel self refresh, but it's unclear how freesync will work. It's possible to use an extra framebuffer on the GPU to shift cost from the display to the video card, but there may be some performance and total system price considerations that make RAM inside the monitor a better option.Death666Angel - Thursday, June 5, 2014 - link
Not really. Current LCDs can hold their image for long enough that you don't have to redraw it at low Hz settings. I can set my displays to 24Hz and they work fine,The reason I heard for why nVidia needs proprietary tech and AMD can use existing stuff is that AMD supported the blanking in eDP while nVidia didn't and had to install a board inside the monitor that could communicate with the graphics card to arrange the blanking interval.
If this makes it inside next gen 4k panels I might be interested. In the mean time, 110Hz 1440p IPS fine. :D