CES is the time of year when computing companies of every type show off their next-generation hardware, and AMD is no exception. Hot Hardware sat down with the Sunnyvale-based company to talk about its major plans for 2015 in both gaming and the CPU side of the business.
In gaming, the big news of the event is the imminent deployment of displays based on the so-called FreeSync standard. FreeSync is the brand name created for what the VESA (Video Electronics Standard Association) refers to as Adaptive Sync (not to be confused with NVIDIA's "Adaptive V-Sync."
Both FreeSync and NVIDIA's proprietary G-Sync technology are designed to solve the same problem. The images below are drawn from NVIDIA's initial G-Sync presentation, but FreeSync functions in the same way.
Turning V-Sync on prevents screen tearing, but can introduce lag and stutter if new frames don't synchronize with the display controller. This is where the "Sync" in both G-Sync and Free-Sync comes from.
The monitors AMD was showing off included the LG 29UM67 (29-inch, 2560x1080, 30-60Hz refresh rate), the Samsung UE590 (28-inch, 4K, 40-60Hz refresh rate) and a BenQ XL2730Z (27-inch, 2560x1440, 144Hz refresh rate).
AMD expects there to be a total of 11 FreeSync displays available by March at varying refresh rates, resolutions, and panel sizes, including IPS panel options and the aforementioned 144Hz gaming panels. Obviously a full comparison between G-Sync and FreeSync will have to wait for head-to-head hardware, but our team reports that the two standards appeared to perform identically.
Assuming that continues to be true, AMD could have an advantage with this feature -- FreeSync reportedly doesn't add any additional cost to display panels, whereas the ASIC hardware required for G-Sync reportedly increases panel retail cost by ~$150. Of course, it's ultimately up to the manufacturers themselves whether or not to charge a premium for FreeSync monitors -- there's just no baked-in cost increase from specialized hardware.
AMD's Carrizo: Bulldozer's Last Hurrah:
The other major topic of the day was AMD's upcoming Carrizo APU. This will be the last version of the much-maligned Bulldozer architecture, but AMD isn't putting much emphasis on the final version of the CPU core, codenamed Excavator. Instead, the company is emphasizing how Carrizo is designed to minimize OEM development costs and offer an efficient path to market for a wide variety of products.
Carrizo's Excavator core has been tweaked in some as-yet unspecified ways to improve performance-per-watt, and the chip is rumored to include 256-bit AVX registers, as well as some increased power throttling abilities and other minor adjustments.
One change that AMD was willing to talk about is that the southbridge is no longer on a separate die connected via MCM. The new southbridge is directly attached to the SoC. The advantages of this are twofold: First, AMD can likely save some power by controlling the entire die directly (and in this game, a savings of 0.2W at idle can be significant). Second, this further simplifies OEM cooling and system design.
AMD does expect Carrizo to offer improved performance compared to Kaveri, but part of that increase is going to come courtesy of longer Turbo times and better power management. There's no word on the GPU side of the equation, but it seems logical to expect either a tweaked version of GCN, possibly updated to the same IP as AMD's R9 285 tapped back in August. Tonga included several features that might be useful in mobile contexts given the bandwidth constraints intrinsic to the form factor.
We'll have to wait for shipping hardware to judge how much the company has improved year on year, but the games we had a chance to test played smoothly, and AMD's APUs have long had an advantage in this field over Intel's competition. Mobile Kaveri delivered strong performance improvements compared to the previous generation of Richland chips and if AMD can pull that off again, Carrizo should further improve the competitive situation.