AMD’s NDA (non-disclosure agreement) for the R9 290X lifts on 15 October but that hasn’t stopped new slides from leaking out detailing what kind of performance the company expects from its new flagship, the R9 290X. Preliminary testing done by AMD suggests that it will allow most games to be easily playable on UltraHD 4K monitors.<\/p>\n
AMD’s R9 290X is based on a new GPU family called Hawaii and it is the biggest GCN-based chip to date. Its been billed as a Titan-killer and will be priced competitively with the Geforce GTX780.<\/p>\n
The new slides show that AMD expects the R9 290X to run most games around the 40fps mark with all settings at maximum and a native resolution of 3840 x 2160. All of the games tested run higher than 30fps, although AMD makes no mention if this was the minimum or maximum framerate being measured. Notably, Battlefield 3, Far Cry 3 and The Elder Scrolls V: Skyrim all run at 60fps or higher.<\/p>\n
The second leaked slide additionally shows the performance boosts AMD expects users will see when adding two or more Radeon R9 290X cards together. Adding in a second card boosts the 3D Mark Firestrike score by 87%, while a third card improves the score by 160% compared to a single card.<\/p>\n
The third slide details the specifications for the Radeon R9 290X and the R9 290. Both cards come standard with 4GB of video memory on a 512-bit bus, both require only a single 6 and 8-pin PEG power connector with a possible maximum TDP (thermal design power) of 300W and both support DirectX 11.2, OpenGL 4.3 and AMD’s Mantle API.<\/p>\n
To illustrate the changes compared to the high-end Radeon HD7970 and HD7950, the table below compares specifications and also notes how much the hardware has changed in terms of percentages.<\/p>\n
| Specification \n<\/span><\/strong><\/td>\n Radeon HD7970 GHz Edition<\/strong><\/td>\n | Radeon R9 290X<\/strong><\/td>\n | Change %<\/strong><\/td>\n | Radeon HD7950 Boost Edition | \n<\/strong><\/td>\n Radeon R9 290<\/strong><\/td>\n | Change %<\/strong><\/td>\n<\/tr>\n | Compute performance | \n<\/strong><\/td>\n 4.096 TFLOPS<\/td>\n | 5.6 TFLOPS<\/td>\n | \u00a0+26.85%<\/td>\n | 3.046 TFLOPS<\/td>\n | 4.9 TFLOPS<\/td>\n | \u00a0+37.84%<\/td>\n<\/tr>\n | Stream processors | \n<\/strong><\/td>\n 2048<\/td>\n | 2816<\/td>\n | \u00a0+27.27%<\/td>\n | 1792<\/td>\n | 2560<\/td>\n | \u00a0+30%<\/td>\n<\/tr>\n | Raster Operators | \n<\/strong><\/td>\n 32<\/td>\n | 64<\/td>\n | \u00a0+100%<\/td>\n | 32<\/td>\n | 64<\/td>\n | \u00a0+100%<\/td>\n<\/tr>\n | Clock speed<\/strong><\/td>\n | 1000MHz<\/td>\n | —<\/td>\n | \u00a0—<\/td>\n | 850MHz<\/td>\n | —<\/td>\n | \u00a0—<\/td>\n<\/tr>\n | Boost speed<\/strong><\/td>\n | 1050MHz<\/td>\n | 1000MHz<\/td>\n | \u00a0-4.76%<\/td>\n | 925MHz<\/td>\n | 947MHz<\/td>\n | \u00a0+2.32%<\/td>\n<\/tr>\n | Memory speed<\/strong><\/td>\n | 1500MHz<\/td>\n | 1125MHz<\/td>\n | \u00a0-25%<\/td>\n | 1250MHz<\/td>\n | 1125MHz<\/td>\n | \u00a0-25%<\/td>\n<\/tr>\n | VRAM buffer<\/strong><\/td>\n | 3GB<\/td>\n | 4GB<\/td>\n | \u00a0+25%<\/td>\n | 3GB<\/td>\n | 4GB<\/td>\n | \u00a0+25%<\/td>\n<\/tr>\n | Bus width<\/strong><\/td>\n | 384-bit<\/td>\n | 512-bit<\/td>\n | \u00a0+25%<\/td>\n | 384-bit<\/td>\n | 512-bit<\/td>\n | \u00a0+25%<\/td>\n<\/tr>\n | AMD TrueAudio support | \n<\/strong><\/td>\n No<\/td>\n | Yes<\/td>\n | \u00a0—<\/td>\n | No<\/td>\n | Yes<\/td>\n | \u00a0—<\/td>\n<\/tr>\n | TDP<\/strong><\/td>\n | 250W<\/td>\n | 250W<\/td>\n | \u00a0—<\/td>\n | 200W<\/td>\n | 250W<\/td>\n | \u00a0—<\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n | Several figures are omitted because AMD either provides no specifications or the changes present no performance boosts or deficits. TDP is the only real unknown at this point because AMD has not revealed their expectations for power consumption.<\/p>\n The jumps between GPUs are significant, increasing hardware specs by 34% on average. The changes in the number of stream processors available as well as the memory size and bus may hold significant performance increases in real-world conditions. Clock speeds will also be a big factor in performance as these GPUs are much larger than their Tahiti cousins which now make up the R9 280X and R9 280.<\/p>\n The R9 290X and the R9 290 also do away with Crossfire dongles, running the technology instead over the PCI-Express 3.0 bus and in the Catalyst software and drivers. AMD says this has no performance penalties compared to using the older physical bridge. This could not only make frame pacing issues a thing of the past, it could also enable AMD to allow more than four Radeon R-290X cards in Crossfire as long as there’s enough bandwidth between the cards for communication.<\/p>\n\n\t\t |