Almost immediately after our press conference with nVidia concluded -- the one directly challenging the validity of Mantle -- AMD contacted us for a discussion on their R9 295X2 video card. The R9 295X2 has been spoiled for quite a while in traditional AMD marketing fashion, namely by sending extremely flattering photos to some of the major tech outlets. As of last week's call, we were able to get the full Radeon R9 295X2 specs, including TDP, fab process, memory & buses, and details on the difficulty with PSU support.
Let's get right to it with this one.
In my many years working on the journalism side of this industry, I've never seen nVidia put forth such an aggressive stance as exhibited during last week's press conference. We'll start this post with some rapid-fire catching up from the last few months.
The past months have been very AMD-intensive. AMD's Mantle API fronted momentous marketing outreach, touting a bypass to DirectX's performance overhead that has historically been a drain on CPU and GPU output. The hardware has been held back by the API, we were (somewhat accurately) told by AMD, and Mantle was the proposed solution to put developers "closer to the metal" -- or closer to the hardware-level -- similar to console development. This news came to be during a period of silence for Microsoft's DirectX API, which hadn't seen noteworthy development since 2012 with Dx11.1 (at that point). It was the ideal opportunity for an emergent API to make a big splash without significant, refreshed competition.
After offering reddit's computer hardware & buildapc sub-reddits the opportunity to ask us about our nVidia GTC keynote coverage, an astute reader ("asome132") noticed that the new Pascal roadmap had a key change: Maxwell's "unified virtual memory" line-item had been replaced with a very simple, vague "DirectX 12" item. We investigated the change while at GTC, speaking to a couple of CUDA programmers and Maxwell architecture experts; I sent GN's own CUDA programmer and 30+ year programming veteran, Jim Vincent, to ask nVidia engineers about the change in the slide deck. Below includes the official stance along with our between-the-lines interpretation and analysis.
In this article, we'll look at the disappearance of "Unified Virtual Memory" from nVidia's roadmap, discuss an ARM/nVidia future that challenges existing platforms, and look at NVLink's intentions and compatible platforms.
(This article has significant contributions from GN Staff Writer & CUDA programmer Jim Vincent).
In a somewhat tricksy move today, AMD hosted a press conference a couple of miles from nVidia’s active GTC event going on down the road. In yesterday’s keynote by nVidia CEO Jen-Hsun Huang, we saw the introduction of the new Titan Z video card, Pascal architecture, machine learning, and other upcoming GPU technologies. Now, less than 24 hours later, AMD has invited us by to look at their new high-end workstation solution – the W9100 FirePro GPU.
The presentation was pretty quick compared to what we got with nVidia, but the primary focus was on computationally-intensive OpenCL tasks, real-time color correction and editing playback in full 4K resolution, and “enabling content creation.”
Let’s start with the obvious.
AMD announced yesterday that it is releasing the AM1 platform. The AM1 socket type and accompanying motherboards are scheduled to be released globally on April 9, the day after Windows XP officially ceases support. This will feature the AMD quad-core and dual-core “Kabini” APU with socketed FS1b motherboards in micro-ATX and mini-ITX sizes. Currently, ASRock, ASUS, BIOSTAR, ECS, Gigabyte, and MSI are planning to have something ready to go by the launch date. The Kabini APU consists of the Graphics Core Next (GCN) architecture-equipped HD 8000 series GPU and AMD's “Jaguar” CPU cores, as found in the PS4 and Xbox One.
The name is still unknown, but what will eventually become DirectX 12 should be shown off at GDC shortly; we'll be in attendance to report on the new announcements and will also be attending the GPU Technology Conference the following week, so check back for deeper analysis as we are exposed to information. In the meantime, Microsoft's new iteration of DirectX has some between-the-lines reading for AMD's Mantle.
It might seem like just a few weeks ago that we posted about AMD's Catalyst 14.1 beta drivers --and that's because it was. The 14.1 beta drivers saw the delivery of Mantle to the general public, which would later be applied in a large (>1GB) Battlefield 4 patch. Of course, I've yet to discover a software application that can release patches without introducing new issues. Catalyst 14.2 attempts to address some of the leftovers from 14.1 and prior.
14.2's major changes bring TrueAudio and Mantle support for the impending launch of the new Thief game; TrueAudio, if you're unaware, is AMD's way of driving surround-style audio through the video card (offloading some processing). The experience as a whole is pretty unique and promising, though we've only tested the "studio version" at CES, where the environment and equipment were fairly optimized.
NOTE: This has been updated with the launch version of the game!
During our hands-on press preview with Titanfall's PC deployment last night, we put the game through its paces on numerous GPU configurations atop our standardized test bench. Initial test attempts resulted in some frustration and hurdles, but with enough research and troubleshooting, we managed to develop a stable, reliable test bed for Titanfall's PC debut.
If you're yet unfamiliar with Titanfall, check out our (now-outdated) Titanfall Analysis.
In this Titanfall benchmark & analysis, we look at the best video cards for Titanfall, framerates (FPS), performance of APUs, SLI configs, & CrossFire, and more; the graphics devices we tested on Titanfall include AMD's 7850 1GB (+ CrossFire), the A10-5800K Trinity APU (7660D), NVIDIA's GTX 650 Ti Boost 2GB (+ SLI), GTX 760 2GB, and GTX 580 1.5GB (for reference), and Intel's HD4000 integrated graphics processor (IGP) on the 3rd-Gen Ivy Bridge CPUs. This IGP is also found in modern Haswell CPUs.
Note: Titanfall is presently in early beta, so it is highly likely that these numbers will improve as the game nears launch and optimization patches are released. It is also likely that nVidia and AMD will release updated drivers with profiles for Titanfall shortly, at which point we will re-test the game appropriately.
Although it doesn't compete directly with Thunderbolt's ability to push PCI-e signals, AMD's new DockPort standard (renamed from code name "Lightning Bolt") has promising features and pricing expectations.
We recently discussed that DisplayPort's primary adoption advantage over HDMI is its lack of a licensing fee (HDMI costs $10,000 per year for adopters to utilize). DockPort continues the free legacy of DisplayPort and other AMD technologies, making itself available for integration without licensing from manufacturers. AMD released a new video showcasing the tech's use case scenarios.