The Intel Xe DG1 GPU has appeared in several benchmark databases, even showcasing its graphics potential in 3DMark where it handily outperforms the new 7nm Vega GPUs featured on AMD's Ryzen 4000 processors.
Intel Xe DG1 GPU Outperforms AMD's 7nm Vega GPUs By Up To 40% - Up To 1.5 GHz Clocks In Early Tests
The Intel Xe DG1 GPU is the first publicly announced chip that will be making its way into the consumer market later this year. An SDV (Software Development Vehicle) graphics card based on Intel's DG1 GPU has already started sampling to ISV's worldwide so that they can optimize their software for the Xe graphics architecture. But it looks like the first entries of the DG1 have already started appearing on various benchmark databases.
There are various entries for Intel's Xe GPUs (Gen 12) which include tests performed on both Tiger Lake (Integrated GPUs) and Coffee Lake (Discrete GPUs). The Tiger Lake family will be the first to feature the integrated Xe graphics architecture, while the entries on Coffee Lake CPUs feature discrete graphics boards which could very well be the DG1 GPU based SDV mentioned above.
Intel Xe DG1 (Discrete) GPU OpenCL Benchmarks in Geekbench
Coming to the benchmarks, the Xe DG1 GPU is tested in the Geekbench 5 OpenCL benchmark. The Tiger Lake-U variant of the chip reportedly features 96 execution units with a reported clock speed of 1.50 GHz. The OpenCL score for this platform is 12444. The Tiger Lake-U chip itself is a 4 core and 8 thread variant which clocks in at a base frequency of 2.30 GHz. On closer inspection, the chip isn't using its integrated Gen 12 Xe GPU, but rather a discrete Xe DG1 graphics card.
Intel Xe DG1 GPU appears on the Geekbench 5 OpenCL benchmark database. (Image Credits: _Rogame)
The Coffee Lake R-S variants are desktop discrete graphics cards and not integrated parts like the one featured on the Tiger Lake platform. Both entries, the Core i5-9600K and the Core i9-9900K deliver a similar score of 11990 and 12053, respectively. While both platforms are configured with the 96 EUs DG1 GPU, they feature slightly variable clock speeds of 1.00 & 1.05 GHz, respectively.
Intel Xe DG1 (Integrated & Discrete) GPU Graphics Benchmarks in 3DMark
More surprising are the 3DMark benchmarks which are once again spotted by _rogame. The benchmarks compare the discrete DG1 graphics cards to the integrated variant on Tiger Lake-U processors and also AMD's Renoir 'Ryzen 4000' APU lineup. In the overall graphics score, the DG1 GPU is faster than all other chips tested however, the Ryzen 7 4800U manages to come very close. The Tiger Lake-U's integrated variant is however around 18% slower than the Ryzen 7 4700U.
But at the same time, both the discrete and integrated variant features 96 EUs, but the desktop GPUs tend to offer higher sustainable clocks compared to integrated chips. Still, Tiger Lake CPUs and DG1 discrete GPUs are being handled by two different teams at Intel and it is possible that Tiger Lake integrated GPUs may not be as well optimized as the discrete DG1 graphics board right now.
Now the more interesting part is in three of the four graphics tests, the DG1 discrete GPU manages up to 40% better performance than the Ryzen 7 4800U. The Ryzen 7 4800U features the enhanced 7nm Vega GPU with 512 stream processors packed in 8 compute units and each CU being up to 59 percent faster than the 14nm Vega architecture.
There's only one graphics test where the AMD GPU leads and that is the first graphics test which focuses more towards volumetric illumination and the shadows. The rest of the tests focus more on tessellated geometry which may hint that the Xe GPU architecture is currently more optimized towards geometry and texture handling with shadows and lighting specific optimizations heading later on. This doesn't mean that we won't see further GPU performance improvements to geometry by the time the GPU is ready for launch as ISV's just got their hand on Intel's latest GPU architecture for optimization work.
There's also another benchmark in 3DMark TimeSpy where the discrete Xe DG1 graphics card is compared to the integrated GPU on a Tiger Lake-U processor. The specific clock speeds are not mentioned but the discrete graphics card is up to 9% faster than the same 96 EU GPU but on an integrated design. Another thing to consider here is that the current clock speeds are not finalized and we can see even higher clock speeds in the final variants as development boards are set to much lower clock speeds around 1.00 GHz, as we are seeing in these tests.
There have also been revelations of ray-tracing features that were found in the latest Intel GPU driver code. The code refers to several ray-tracing specific features of the DXR API and also makes a mention of hardware-accelerated ray-tracing coming to Intel GPUs. We don't know if these refer to the first iteration of Xe GPUs or the second generation, but considering how everyone is jumping onboard the ray tracing bandwagon this year, Intel might not want to miss out on the opportunity with their Xe GPUs.
"Ray Trace HW Accelerator" pic.twitter.com/JOKAuwosbC
— _rogame (@_rogame) January 21, 2020
The laptop segment is crucial for Intel as AMD has almost entirely disrupted the desktop segment, whether it be mainstream or HEDT. But with Tiger Lake, we can now tell why Intel is focused on making it a mobility-first family since they want to take the fight to AMD by offering a well-rounded chip design, featuring their latest core architectures on both CPU and GPU fronts. With that said, Intel is all set to unveil Xe GPU architecture details and powerful new features of their brand new graphics architecture at GDC 2020 next month so stay tuned for more info.