Our recent Fury X driver comparison took rumors of a disparate relationship between press and launch drivers to task, ultimately finding that no real difference existed. This testing procedure exposed us to the currently discussed “coil whine” and “pump whine” of the new R9 Fury X. Today's test seeks to determine with objectivity and confidence whether the whine is detrimental in a real-world use case.
AMD's R9 Fury X video card emits a high frequency whine when under load. We have located this noise on both of our retail units – sold under Sapphire's banner, but effectively identical to all Fury X cards – and reviewers with press samples have cited the same noise. The existence of a sound does not inherently point toward an unusably loud product, though, and must be tested in a sterile environment to determine impact to the user experience. The noise resembles coil whine, for those familiar with the irritating hum, but is actually an emission from the high-speed pump on the Fury X. This relegates the noise to what is ultimately a mechanical flaw in the engineering rather than something electrical, as coil whine would suggest.
Our R9 Fury X analysis is still forthcoming, but we interrupted other tests to quickly analyze driver performance between the pre-release press drivers and launch day consumer drivers.
All testing was conducted using a retail Fury X, as we were unable to obtain press sampling. This benchmark specifically tests performance of the R9 Fury X using the B8, B9, and release (15.15.1004) drivers against one another.
The purpose for this test is to demystify some rumors that the Fury X would exhibit improved performance with the launch day drivers (15.15.1004), with some speculation indicating that the press drivers were less performant.
Regardless of how its mechanics pan-out, Star Citizen is slated to claim the throne as one of the most graphically intense PC games in recent history. This is something we discussed with CIG's Chris Roberts back when the Kickstarter was still running, diving into the graphics technology and the team's intent to fully utilize all tools available to them.
We've been trying to perform frequent benchmarks of Star Citizen as the game progresses. This progress monitor comes with a massive disclaimer, though, and is something we'll revisit shortly: The game isn't finished.
The recent launch of the GTX 980 Ti, R9 Fury X, and AMD 300 series cards almost demands a revisit to Star Citizen's video card performance. This graphics benchmark looks at GPU performance in Star Citizen's 1.1.3 build, testing framerates at various settings and resolutions.
Following our initial review of AMD's new R9 390 ($330) and R9 380 ($220) video cards, we took the final opportunity prior to loaner returns to overclock the devices. Overclocking the AMD 300 series graphics cards is a slightly different experience from nVidia overclocking, but remains methodologically the same in approach: We tune the clockrate, power, and memory speeds, then test for stability.
The R9 390 and R9 380 are already pushed pretty close to their limits. The architectural refresh added about 50MHz to the operating frequency of each card, with some power changes and memory clock changes tacked-on. The end result is that the GPU is nearly maxed-out as it is, but there's still a small amount of room for overclocking play. This overclocking guide and benchmark for the R9 390 & R9 380 looks at the maximum clockrate achievable through tweaking.
All these tests were performed with Sapphire's “Nitro” series of AMD 300 cards, specifically using the Sapphire Nitro R9 390 Tri-X and Sapphire Nitro R9 380 Dual-X cards. Results will be different for other hardware.
AMD's most recent video card launch was September of 2014, introducing the R9 285 ($243) on the slightly updated Tonga GPU. Tonga was laterally imposed to take the place of the Tahiti products, namely the HD 7970 and its refresh, the R9 280. The Radeon 7970 video card shipped in late 2011 on the Tahiti GPU, a die using TSMC's still-fabbed 28nm process, and was refreshed as the R9 280, then updated, improved, and refreshed again as the Tonga-equipped R9 285. At its core, the 285 would offer effectively identical on-paper specs (with some changes, like a 256-bit memory bus against the 384-bit predecessor), but introduced a suite of optimization that yielded marginally improved performance over the R9 280.
All of this is to say that it's been a number of years since AMD has introduced truly new architecture. Tahiti's been around four years now, Hawaii shipped in 2013 and was a node refresh of Tahiti (more CUs, ROPs, and geometry / rasterizer processors), and Fiji – the anticipated new GPU – won't ship for a short bit longer. Filling that space is another refresher line, the Radeon 300 series of video cards.
AMD's lull in technological advancement on the hardware side has allowed competitor nVidia to increase competition in some unchallenged market segments, like the high-end with the GTX 980 Ti ($650) and mid-range with the GTX 960 ($200). The long-awaited R9 300 series video cards have finally arrived, though, and while they aren't hosting new GPUs or deploying a smaller fab process, the cards do offer marginally increased clockrates and other small changes.
This review benchmarks the AMD R9 390 and AMD R9 380 graphics cards against the preceding R9 280, R9 290(X), GTX 960, and other devices. The R7 370 and R7 360 also launch today, but won't be reviewed here.
In our GTX 980 Ti overclocking endeavors, it was quickly discovered that the card encountered thermal bounds at higher clockrates. Driver failures and device instability were exhibited at frequencies exceeding ~1444MHz, and although a 40% boost in clockrate is admirable, it's not what we wanted. The outcome of our modest overclocking effort was an approximate ~19% performance gain (measured in FPS) for a selection of our benchmark titles, enough to propel the 980 Ti beyond the Titan X in gaming performance. Most games cared more about raw clock speed of the lower CUDA-count 980 Ti than the memory capacity of the TiX.
In a recent update to the site for AMD-exclusive manufacturer XFX, leaked images of the company's impending R9 390X card were posted for public review. The images were likely posted in error and have since been removed.
The product render shows XFX's R9 390X marketing, resembling Hawaii rebranding from the R9 290X as an 8GB device. The R9 390X will lack HBM and other architectural updates to Hawaii – items that will only be found on the company's Fury and Fury X GPUs.
Over the course of our recent GTX 980 Ti review, we encountered a curious issue with our primary PCI Express port. When connecting graphics cards to the first PCI-e slot, the card wouldn't detect and resolution would be stunted to lower values. Using one of the other slots bypassed this issue, but was unacceptable for multi-GPU configurations – something we eventually tested.
Multi-GPU configurations have grown in reliability over the past few generations. Today's benchmark tests the new GeForce GTX 980 Ti in two-way SLI, pitting the card against the GTX 980 in SLI, Titan X, and other options on the bench. At the time of writing, a 295X2 is not present for testing, though it is something we hope to test once provided.
SLI and CrossFire have both seen a redoubled effort to improve compatibility and performance in modern games. There are still times when multi-GPU configurations won't execute properly, something we discovered when testing the Titan X against 2x GTX 980s in SLI, but it's improved tremendously with each driver update.
Our initial review of the $650 GTX 980 Ti, published just over twelve hours prior to this post, mentioned an additional posting focusing on the card's overclocking headroom. The GTX 980 Ti runs GM200, the same GPU found in nVidia's Titan X video card, and is driven by Maxwell's new overclocking ruleset.
Maxwell, as we've written in a how-to guide before, overclocks differently from other architectures. NVidia's newest design institutes a power percent target (“Power % Target”) that increments power provisioning to the die to grant OC headroom. Unfortunately, this metric can't be exceeded beyond what the BIOS natively allows (without a hack, anyway), and means that we're sharing watts between the core clock, memory clock, and voltage increase. Overclocking on Maxwell offers some granularity without making things too complicated, though it's not until we get hands-on with board partner video cards that we'll know the true OC ceiling of the 980 Ti.
This post showcases our GTX 980 Ti initial overclock on the reference cooler, yielding a considerable framerate gain in game benchmarks.