Thursday, January 9, 2014

AMD claims it can offer the benefits of Nvidia’s G-Sync with a free driver update.


At CES this week, AMD made an unusual announcement about Nvidia’s new G-Sync technology. According to the company’s senior engineers, they can replicate much of the advantages of Nvidia’s G-Sync tech through the use of what are called dynamic refresh rates. Multiple generations of AMD video cards have the ability to alter refresh rates on the fly, with the goal of saving power on mobile displays. Some panel makers offer support for this option, though the implementation isn’t standardized. AMD engineers demoed their own implementation, dubbed “FreeSync,” on a laptop at the show.


AMD’s windmill application FreeSync demo. Unfortunately, it’s impossible to find video demos on YouTube that don’t ruin the G-Sync or FreeSync effect.

Dynamic refresh rates would theoretically work like G-Sync by specifying how long the display remained blank on a frame-by-frame basis, providing for smoother total movement. AMD has stated that the reason the feature didn’t catch on was a lack of demand — but if gamers want to see G-Sync-like technology, AMD believes it can offer an equivalent. AMD also told Tech Report that it believes triple buffering can offer a solution to many of the same problems G-Sync addresses. AMD’s theory as to why Nvidia built an expensive hardware solution for this problem is that Nvidia wasn’t capable of supporting G-Sync in any other fashion.

Nvidia rebuts

Nvidia, unsurprisingly, has a different view of the situation. Tech Report spoke to Tom Peterson, who stated the difference between a laptop and a desktop running a software equivalent to G-Sync is that laptop displays are typically connected using embedded DisplayPort or the older LVDS standard. Standalone monitors, in contrast, have their own internal scaling solutions and these chips typically don’t support a variable refresh rate.

I think Nvidia is probably being honest on that score. The G-Sync FPGA is fairly hefty, with 768MB of onboard memory and a limited number of compatible monitors. Nvidia has a long interest in keeping its technology proprietary, but it also has reasons to extend G-Sync as widely as possible for as little up-front cost as possible. A G-Sync upgrade kit for $50 that fits any modern monitor would sell more units than a $100 or $150 kit that only fits a limited number of displays or that requires a new LCD purchase.

Nvidia’s G-Sync includes a 768MB buffer combined with a custom FPGA.

It’s entirely possible that both companies are telling the truth on this one. AMD may be able to implement a G-Sync-like technology on supported panels, and it could work with the manufacturers of scalar ASICs if G-Sync starts catching on for Nvidia. Nvidia, meanwhile, is probably telling the truth when it says it had to build its own hardware solution because existing chips for desktop displays weren’t doing the job.

Whether this works out to a significant halo for Nvidia in the long run or not will come down to price and time-to-market. In the past, Nvidia took the lead on computing initiatives like PhysX and CUDA, getting out in front on technical capability, while industry-wide standards followed along at a slower pace. The impact on the consumer market has been mixed — PhysX definitely delivered some special effects that AMD didn’t match, but CUDA’s impact on the consumer space has been small (its HPC success is another story altogether).

The difference between these technologies and G-Sync is that monitors are fairly long-lived. Buy a G-Sync monitor today, and you have the benefits for five years or more. Some games benefit from G-Sync more than others, but once Nvidia smoothes out the development pipeline, we should see a consistent stream of titles that run better in that mode. It’s not like hardware PhysX, which was never supported by more than a handful of major games in any given year. In the long run, if panel makers start building variable refresh rates into their own displays, than the need for Nvidia-specific G-Sync technology may fade out — but that doesn’t mean the company can’t make a pretty penny off the concept while it lasts. And since it’ll take time for panel manufacturers to adopt the capability if they choose to do so, it means Nvidia has a definite window of opportunity on the technology.

No comments:

Post a Comment