AMD’s integrated graphics are generally better than Intel’s. The discrepancy has existed since at least as far back as the nForce 2 chipset, and while the particulars have varied (Intel’s Crystal Well on-die EDRAM gave it better absolute performance at times, albeit at much higher prices), the end result rarely has. Historically, AMD has offered higher integrated GPU performance at the same price point compared with Intel. But that could change with Intel’s next-generation on-die GPUs, set to debut with Intel’s new Sunny Cove CPU architecture.
New leaked benchmarks show Intel’s Gen 11 GT2 solution benchmarking significantly higher than the old, Skylake-era Gen 9 core. Some of these gains are to be expected — the Skylake GPU core is essentially four years old already — but the overall level of improvement is quite good in its own right.
Reddit user Dylan522p compiled leaked benchmarks into charts, shown below. The first chart compares Intel’s i5-8250U with UHD Graphics (GT2) against an unidentified Intel CPU in a GT2 configuration with Iris Plus Graphics 940. Note that since we don’t know anything about the thermal limits imposed on this test CPU, we also don’t know if it’s fair to compare the 8250U with this chip as opposed to a desktop-socketed part. This may overstate the performance improvement slightly.
The gains here are very strong. Performance more than doubles in the Aztec Ruins tests (a video of the Aztec Ruins benchmark is embedded below, though this doesn’t show the benchmark run off this specific hardware, only the test itself). Older scenes still see significant performance improvements, and we’re far from launch day, with silicon and drivers both in active development.
The second graph compares the Intel solution to the AMD Ryzen platform, with data from a 2700U (15W) and a Ryzen 2400G (65W). In the absence of information to confirm that the Intel CPU is a laptop chip actually running within its specified TDP, I’m going to assume that the 2400G is actually the better point of comparison for the core. Even here, it’s clear that the Gen 11 Intel GPU could be potent competition for Ryzen 2400G, at least if these tests are accurate.
Of course, there are questions about how well GFXBench performance will translate into shipping titles and how effectively Intel will compete on price. We’re hearing rumors that Ice Lake systems could actually be on store shelves by this summer rather than being delayed into the holiday season. AMD’s third-generation Ryzen 3000 APUs are still 12nm products, not 7nm, and they don’t offer much more than small speed bumps and some utilization improvements over previous parts. We don’t know yet how much additional performance to expect, but most estimates are on the modest side. AMD undoubtedly has a 7nm APU in the works, but the company has said it won’t use the literal Matisse design with a GPU chiplet onboard instead of a second CPU die to launch the product.
While we won’t draw conclusions about Gen 11 performance until we have silicon and drivers to test, it ultimately isn’t surprising to see Intel closing the gap with AMD. If Chipzilla is serious about using its own architecture to build a GPU, it’s going to need to improve the performance of its own architecture, period. Since Gen 11 is positioned as a stepping stone to Xe, Intel’s discrete GPU hardware, we should expect to see substantial performance and performance-per-watt uplift. Nvidia doesn’t compete in integrated consumer graphics any longer, but both it and AMD scale their mainstream graphics architectures down into embedded products. Intel may have decided the best way to develop its own new architecture was to take a bottom-up approach and use it for relatively simple solutions first before scaling up into datacenters and gaming PCs.
- Intel Shows Off New Gen11 Graphics, Teases Xe Discrete GPU
- Intel at CES 2019: Ice Lake, 10nm Parts, New Network SoCs
- A Matter of Process: Intel Architecture Day 2018 and the Future of Silicon