AMD Strix Point APUs With RDNA 3+ iGPU Should Match RX 6400 With 12 CUs, RTX 3050 With 16 CUs

AMD Strix Point APUs With RDNA 3+ iGPU Should Match RX 6400 With 12 CUs, RTX 3050 With 16 CUs

AMD's Strix Point APUs featuring RDNA 3+ iGPU Should offer performance comparable to entry-level discrete GPUs as indicated in new performance rumors.

AMD is bringing entry-level discrete GPU performance to RDNA 3+ iGPUs featured on next-gen Strix Point APUs.

of AMD Strix Point APUs There is going to be a complete overhaul to bring in brand new core IPs on all fronts. gave The CPU side will be upgraded to Zen 4 cores.the GPU side will be upgraded to RDNA 3+ cores. The NPU side will be upgraded to the latest XDNA 2 “Ryzen AI” architecture.. In addition, the Strix APUs will come in two configurations, a standard monolithic and a premium chiplet design. AMD will designate the monolithic variants as Strix Point 1 and the chiplet variants as Strix Point 2. Both of them will have the same architecture but will come in completely different configurations.

As you can tell from its name, the chiplet design will be one to look out for as it is designed for enthusiast PCs and while it will take the performance of APUs to the next level, the standard monolithic design A good improvement will also be offered. In terms of graphics performance.

Image Source: Wccftech (AI Generated)

This was recently mentioned by @Xinoasassin, Insider with good knowledge of AMD and Intel architectures, that the RDNA 3+ iGPUs featured on Strix Point APUs will feature performance comparable to entry-level discrete GPU solutions. In 3DMark Time Spy the 12 compute unit configuration at a limited TDP of 22-24W is rated at around 3150 points and we can expect around 4000 points in the same benchmark with the full configuration. .

AMD Ryzen 8050 Strix Point Mono Expected Features:

  • Zen 5 (4nm) monolithic design
  • Up to 12 cores in hybrid config (Zen 5 + Zen 5C)
  • 32 MB of shared L3 cache
  • 35% faster CPU vs Phoenix at 50W
  • 16 RDNA 3+ compute units
  • 128-bit LPDDR5X memory controller
  • XDNA 2 engine integrated
  • ~25 tops AI engine
  • 2H 2024 Launch (Expected)

For comparison, the AMD Radeon 780M (RDNA 3) iGPU delivers around 2300-2400 points in the Time Spy benchmark at the same TDP and averages 2800 points at peak TDP on laptops. The same chip scores an average of 3300-3500 points even using full unrestricted desktop TDPs. At the same TDP, that's about a 35% performance increase which is pretty good for what's mostly going to be the same architecture with some tuning and faster clock speeds. AMD Radeon RX 6400 and NVIDIA GeForce RTX 2050 discrete notebook GPUs also average 3500 points in 3DMark Time Spy and both GPUs run at higher TDP.

RTX 3050 (2048 Cores @ 50W)

Strix Point (RDNA 3+ 16 CU @22W)

RTX 2050 (2048 Cores @50W)

Radeon RX 6400 (RDNA 3 12 CU @53W)

Strix Point (RDNA 3+ 12 CU @22W)

Hawk Point (RDNA 3 12 CU @ 55W)

Hawk Point (RDNA 3 12 CU @22W)

A full 16 compute unit configuration of AMD RDNA 3+ iGPU featured on high-end Strix Point monolithic dies should match the GeForce RTX 3050 GPU (35-50W) which should be great for the mainstream gaming segment. AMD's RDNA 3 iGPUs have already performed very strongly in the integrated space, competing with the likes of Intel's Arc Xe-LPG architecture and outperforming them in the majority of gaming titles. They are getting good driver assistance and features like ray tracing, FSR upscaling, and FSR frame generation Great job we can see more improvements and tech innovations supported by the next generation RDNA 3+ graphics architecture.

It was recently reported that AMD's The RDNA 3+ graphics architecture will remain in Ryzen APUs until 2027 So it's going to be around for a while. We may see improvements down the road similar to what we've seen with Vega iGPUs. Now we also reported how The AI ​​PC frenzy may have damaged some of these chips. As expected these include other technologies such as a larger cache that will help boost the performance of both CPUs and GPUs. Having larger iGPUs also comes with the cost of bandwidth scaling.

GPUs require high bandwidth and there's only so much you can do. A possible solution is to implement an on-die cache such as the infinity cache used by discrete GPUs but this is unlikely for Strix Point. A possible way is to ship these premium laptops with faster LPDDR5x memory solutions which will prove to be a big advantage, Apart from this, LPDDR6 is also being discussed in the industry. That can further eliminate some of the bandwidth constraints from these chips. Intel's Lunar Lake CPUs are going with the on-package LPDDR5x memory route. But the latest block diagram still doesn't show any onboard GPU cache, so it looks like we still have some time before we see more innovative chip packaging technologies in SOCs to improve graphics performance on said chips. can help increase

The chips will also mark a major upgrade for the mini PCs and handheld gaming PCs that have taken the gaming industry by storm.

Share this story.



About the Author

Leave a Reply