Thief Logo |
Today, AMD is adding Mantle support for a new game, Thief and debuting the first public demo of its TrueAudio technology. We’ve taken both out for a spin to kick the tires and check the performance — so where do things stand? We’ve already covered the basics of Mantle performance in previous stories, so we’re going to focus on a hands-on comparison of two platforms using two midrange GPUS — the R9 270 and the R7 260X. AMD has released a great deal of supplementary information regarding its own performance expectations, however, so we’ll refer to that to flesh out our analysis in several respects.
This time, we’ve chosen CPU and GPU configurations that are more likely to represent what people at home are playing with or contemplating for future purchases. Improving the performance of the R9 290 and R9 290X is important for competitive positioning against Nvidia, but the number of people with an A10-7850K and an R9 290X is quite small. The R7 260X and R9 270 (and the HD 7790 and HD 7850 they replaced), on the other hand, are two of AMD’s most common mass-market GPUs. So, what can owners of these products expect?
Mantle under Thief
We tested two configurations: The Intel Core i3-4330 and the AMD A10-7850K using both the R7 260X and the R9 270. Windows 8.1 was installed on both systems with all appropriate patches. AMD’s latest Catalyst 14.3 beta driver was installed. Thief has a built-in benchmark and we used it for testing; programs like Fraps can’t monitor a Mantle application for frame rates yet. Game details were set to Very High Quality, but with two changes: SSAA (Super-sampled antialiasing) was disabled, while anisotropic filtering was bumped to 16x.First, here’s the R7 260X. Note that the performance gains are extremely small in both cases — the Core i3-4330 doesn’t pick up any performance, while the A10-7850K gains just 5%. Clearly at Very High Detail, even with supersampled antialiasing disabled, the GPU is the fundamental bottleneck here. We couldn’t measure frame times in this instance, but the general feel of the game was the same, and color presentation was identical.
Let’s look at the R7 270X, which has significantly more headroom, and see where we’re at.
Here, we see larger improvements. The Core i3-4330 picks up 6%, while the AMD A10-7850K leaps up 23%. That’s an impressive set of gains for a software API update and it puts the two chips on an even footing.
Thief also includes an option to benchmark the game in 32-bit vs. 64-bit mode — we took it for a spin on the Core i3-4330 to see if it would make a difference. Some features are only available in 64-bit mode, including Mantle, but an apples-to-apples benchmark run of the game didn’t reveal any performance advantages to 64-bit. The average frame rate was 42 fps in both cases.
Supplementary data
AMD has released multiple graphs showing where it expects to see the greatest performance gains from using Mantle; we've reproduced several of these below:AMD's Mantle performance |
What this data shows, in aggregate, is that if AMD can continue boosting the attractiveness of Mantle with its various vendors, it can deliver meaningfully better experiences for its customers. Unfortunately, the data we’ve seen suggests that for most mainstream Intel users, the advantages of Mantle are somewhat limited. AMD’s data predicts that the Core i5-4670K can gain over 20% improved performance on the R9 280X and nearly 50% with the R9 290X, while the Core i7-4960X barely breaks the 10% threshold for R9 280X and only hits >30% gains with the R9 290X.
Source: ExtremeTech
The Chief Technomancer
VB Kid
No comments:
Post a Comment