AMD’s newest Ryzen 7000-series processors for desktops primarily based on the Zen 4 microarchitecture characteristic a tiny RDNA-based graphics processing unit inside their I/O die (IOD). This GPU is hardly meant for gaming, however as SkatterBencher found, it will probably nonetheless be overclocked, which improves its efficiency by round 42% in case of AMD’s Ryzen 7900 processor (through VideoCardz). This doesn’t imply the GPU abruptly turns into a viable gaming answer, however not less than it can acquire efficiency.
AMD’s Ryzen 7000-series ‘Raphael’ desktop CPUs combine an RDNA 2-based GPU with 128 stream processors and a couple of CUs inside their IOD. This graphics answer operates at as much as 2.20 GHz and delivers about 0.563 FP32 TFLOPS of compute throughput, which is corresponding to a high-end graphics card from 2007 (e.g., ATI Radeon HD 2900 XT). That is by far not sufficient to get first rate body charges in a contemporary sport. Nonetheless this GPU can show graphics and playback some movies. Moreover, it may be overclocked.
Overclocking of the built-in GPU shouldn’t be a trivial activity although. On the one hand, AMD’s Agesa 184.108.40.206 helps overclocking the iGPU, however to squeeze all of the attainable efficiency of this graphics processor, one must play with AMD’s Precision Enhance Overdrive settings in addition to improve voltages and energy limits.
In the meantime, even a considerably elevated Precision Enhance Overdrive energy funds by itself won’t get a big enchancment for the built-in GPU since it’s a tiny low-end graphics answer that is not meant to carry out. Subsequently, it makes extra sense to manually improve VDDCR_SOC voltage in addition to VDDCR_GFX voltage to get extra energy to the built-in GPU.
By default, the GPU operates at 2.20 GHz at 0.997V, however with the voltage elevated to 1.2V, the graphics processor can hit 2949 MHz. Additional changes of voltage/frequency curvature to 1.395V assist to convey GPU frequency all the best way to three.10 GHz. In the meantime, this will increase SoC energy to 60.7W, up from round 38.5W by default, which implies elevated cooling necessities.
From the efficiency standpoint, this brings an as much as 42.5% of extra efficiency in video games like Tomb Raider, however at 17 frames per second the title stays unplayable. Moreover, it additionally permits to run a raytracing benchmark that doesn’t work on the built-in GPU by default. But, even such excessive overclocking doesn’t convey a considerable efficiency enchancment in various artificial benchmarks, presumably due to different limitations.
Whereas overclocking of the built-in RDNA 2-based GPU doesn’t appear to convey any real-world fruits, it’s a doable. Moreover, now that AMD doesn’t permit adjusting the voltage/frequency curve on discrete RDNA 3 graphics playing cards, it seems like overclocking of its built-in GPUs is simpler than overclocking of discrete GPUs.