A Chinese web-site that leaks unofficial information about future central processing units and graphics processing units has published what it claims to be the first benchmark results of AMD’s next-generation Radeon R9 300-series graphics card code-named “Fiji XT”. If the benchmark results are correct, then the graphics adapters will offer rather high performance in all existing games.
Chiphell, which has a reputation of leaking both accurate and inaccurate information, has published two charts representing average performance and power consumption of AMD Radeon R9 300-series graphics card – which was called “Captain Jack” graphics adapter to emphasize that it belongs to the Pirate Islands family of products – across a number of benchmarks versus current-generation graphics boards.
The list of benchmarks that were used to get “average” FPS [frames per second] number features a long list of games, including Alien Isolation, Assassins Creed: Unity, Battlefield 4. Crysis 3, Dragon Age: Inquisition, Far Cry 4, Metro Last Light Redux and many others. All titles were benchmarked in 2560*1440 resolution with all eye candy (antialiasing, level of details, etc.) set at maximum. The testbed used was powered by Intel Core i7-4790 (3.60GHz) microprocessor. The testers claim that they utilised AMD Catalyst 14.11.2 beta and Nvidia GeForce drivers 344.75.
While the results clearly show that the Radeon R9 300-series “Fiji XT” graphics card is a clear winner that provides around 10fps performance advantage – or a solid 16 per cent – on average versus a factory-overclocked GeForce GTX 980 graphics board, it should be noted that the test conditions are not clear and thus the result cannot by verified anyhow. Moreover, even if the result is correct, then the Radeon R9 300-series “Fiji XT” is only 30.9 per cent faster than the Radeon R9 290X “Tahiti XT”, a not very high generation-to-generation improvement.
At present the Radeon R9 300-series “Fiji XT” (or Captain Jack) graphics cards are probably already available in test labs of AMD’s partners in various southeast Asian countries, therefore, it is possible that an enthusiast employed by a partner could benchmark one of them. What should be noted is that at this time nobody has final specifications of the product that will be sold as the Radeon R9 390X. The GPU configuration is unclear: it may feature 4096 stream processors and 256 texture units, but AMD can well disable two compute units inside its chip for redundancy reasons and the final chip will sport 3968 SPs and 248 TUs. Even if the configuration of the chip itself is locked (i.e., will not be changed), both AMD and Nvidia finalize clock-rates weeks, not months, before commercial launch. As a result, whatever it was used to get benchmark results, it was not a product that will be known as the Radeon R9 390X. Finally, while the Radeon R9 300-series “Fiji XT” is believed to be based on the GCN 1.2 architecture and thus should have decent software support now, the drivers used by the benchmarkers were re not final drivers for a shipping product.
It is interesting to note that if the performance and power consumption measurements are correct, then the Radeon R9 300-series “Fiji” graphics cards should be rather power efficient. Keeping in mind that AMD’s Radeon R9 285 “Tonga Pro” consumes lower amount of power than AMD’s Radeon R9 280 “Tahiti Pro” (both chips are made using 28nm process technology and feature similar amount of transistors), increased power efficiency of GCN 1.2-based “Fiji” GPU may be true.
AMD did not comment on the news-story.
Discuss on our Facebook page, HERE.
KitGuru Says: We have no idea whether the benchmark results are correct or not, but even if they are, “average” fps across multiple games is a metric that is not good. Moreover, assuming that the results are correct, if the Radeon R9 300-series “Fiji XT” graphics card (we assume that the anonymous tester used a higher-end model) is only 15.9 per cent faster than a factory-overclocked GeForce GTX 980, then it will hardly be a performance breakthrough, despite of massive amount of stream processors and HBM DRAM memory with incredible bandwidth.