According to hardware leaks reported by Kepler_L2, Advanced Micro Devices (AMD) is targeting clock speeds between 3.1 GHz and 3.4 GHz for its upcoming graphics processing units built on the RDNA 5 architecture. This next-generation hardware is projected for a release window around the turn of years 2027 and 2028, following delays initially caused by industry-wide chip and memory constraints.
Clock Speeds and Architecture Targets
Clock speeds for the upcoming RDNA 5 architecture represent a notable step up from current generation products. According to available data, current hardware like the Radeon RX 9070 XT features a boost clock of 2.97 GHz, while the Radeon RX 9060 XT reaches 3.13 GHz. The reported 3.1 GHz to 3.4 GHz target for RDNA 5 represents an increase of roughly 200 MHz to 300 MHz. Furthermore, leaked specifications suggest that associated accelerated processing units (APUs) could approach the 4 GHz threshold.
Did you know? Current desktop graphics cards like the Radeon RX 9060 XT operate at a boost clock of 3.13 GHz, setting the baseline for the performance scaling expected in the upcoming RDNA 5 generation.
Memory Configurations and VRAM Discrepancies
The flagship silicon is reportedly designated as Navi 5X, featuring 96 compute units (CUs) and a total of 12,288 stream processors, utilizing a ratio of 128 stream processors per compute unit. By comparison, the Radeon RX 9070 XT utilizes 64 compute units with 4,096 stream processors at 64 stream processors per compute unit. Sources point to a 384-bit memory bus paired with 32 GB of VRAM, though technical analysts note a discrepancy in these figures. A 384-bit bus requires 12 memory chips, which naturally yields either 24 GB using 2GB chips or 36 GB using 3GB chips, barring a complex clamshell configuration.
With 36 GB of VRAM, such a configuration would exceed the memory capacity of Nvidia’s RTX 5090, potentially offering a cost-effective hardware solution for running local artificial intelligence models. However, software support and third-party compatibility remain areas where Nvidia maintains an established lead.
Mid-Range and Entry-Tier Specifications
Moving down the product stack, the mid-range segment is projected to feature 40 compute units packing 5,120 stream processors, alongside a 384-bit memory bus and 24 GB of VRAM. Lower mainstream models are expected to scale down to 24 compute units and 3,072 stream processors, utilizing either 256-bit or 128-bit memory buses paired with 8 GB or 16 GB of VRAM. At the entry level, low-end cards are slated for 12 compute units, 1,536 stream processors, and a narrow 64-bit memory bus.
Industry sources discussing the low-end 64-bit configurations point to 8 GB and 16 GB VRAM capacities, which present similar architectural puzzles. A standard 64-bit bus connects to two physical memory chips, yielding 4 GB or 6 GB using GDDR7 technology. Reaching 16 GB on a 64-bit bus would necessitate an eight-chip array utilizing an 8-bit configuration per chip.
Pro Tip: When evaluating hardware leaks for unreleased GPU generations scheduled several years out, architectural specifications such as memory bus widths and chip counts frequently undergo revision before final mass production begins.
Frequently Asked Questions
When will AMD release RDNA 5 graphics cards?
Current leaks suggest AMD is aiming for a release window around the turn of years 2027 and 2028, though timelines remain subject to change due to market conditions.
What clock speeds are expected for RDNA 5?
According to leaker Kepler_L2, AMD is targeting frequencies between 3.1 GHz and 3.4 GHz for the upcoming architecture.
How does the Navi 5X flagship compare to current hardware?
The reported Navi 5X flagship features 96 compute units and 12,288 stream processors, scaling significantly beyond the 64 compute units and 4,096 stream processors found in the Radeon RX 9070 XT.
Join the Discussion
What are your thoughts on the projected memory configurations and clock speeds for the upcoming RDNA 5 architecture? Share your perspective in the comments below, explore our related hardware analyses, or subscribe to our newsletter for the latest component updates.