🕵️‍♂️ Engage in discussions ANONYMOUSLY. NO REGISTRATION is required. 💬

✅ Registered accounts earn up to $25 per thread via PayPal and BTC. 🤑 Read more in News and Announcments! Unlock the ability to chat, share, send private messages and earn money per post in our community. Just register an account. It's Free!

SignUp Now!
  • 🕵️‍♂️ Engage in discussions ANONYMOUSLY. NO REGISTRATION is required. 💬 ✅ Registered accounts earn up to $25 per thread via PayPal and BTC. 🤑 Read more... Unlock the ability to chat, share, send private messages and earn money per post in our community. Just register an account. It's Free!

How a $95 AMD 4600G/5600G APU became a 16GB VRAM CPU for Running Stable Diffusion?

  • Thread Author
In a surprising twist of ingenuity, a Reddit user has showcased an impressive feat – transforming an affordable AMD Ryzen 5 4600G APU into a functional 16GB VRAM GPU, capable of handling AI applications. This unconventional experiment sheds light on the adaptability of hardware and the exciting possibilities that arise from creative exploration. In this article, we delve into the details of this endeavor, highlighting its significance and potential applications.

How he did it?​

Our main protagonist is the Ryzen 5 4600G APU, introduced in 2020 has a hexa-core, 12-thread configuration built on Zen 2 architecture. Running at a base and boost clock of 3.7 GHz and 4.2 GHz. This 65W APU also features a Radeon Vega iGPU with seven compute units clocked up to 1.9 GHz. APUs like this share system memory rather than having dedicated memory, allowing them to tap into the system's resources for optimized performance.

The brilliance of this lies in the Redditor's clever move! – Allocating 16GB of DDR4 system memory to the Ryzen 5 4600G, essentially turning it into a makeshift 16GB "graphics card." This creative workaround grants more memory than some Nvidia GeForce RTX 40-series models, like the RTX 4070 and RTX 4070 Ti, capped at 12GB. Although the APU doesn't match the power of premium GPUs, it effortlessly avoids memory constraints during AI tasks, making it a cost-effective solution for less demanding endeavors.

Unleashing the Full Language Model Potential​

The magic happens because the APU's have compatibility with third-party ROCm packages, including those offered by BruhnBruhn Holding. By integrating these experimental packages, the APU gains the ability to collaborate with popular AI frameworks like PyTorch and TensorFlow. This opens doors to a diverse array of AI applications, ranging from Stable Diffusion and FastChat to MiniGPT-4 and Whisper.

The Redditor's YouTube video showcases the Ryzen 5 4600G's capabilities across various AI tasks, highlighting its prowess with Stable Diffusion – an AI-driven image generator fueled by textual input.

o_OImpressively, the APU creates a 512 x 512-pixel image in about one minute and 50 seconds!

A commendable feat considering the modest investment of $95. Notably, the choice of DDR4 memory plays a role here, as faster memory options could potentially elevate the APU's AI performance further.

Challenges and Considerations​

Amid the excitement, it's crucial to acknowledge the practical limitations and considerations. While the Ryzen 5 4600G offers an affordable and adaptable solution for AI enthusiasts, it doesn't replace top-tier GPUs designed for intensive AI workloads. Its true strength lies in its adaptability and affordability, making it an appealing choice for those exploring AI applications on a budget.

One key point to address is the absence of detailed setup instructions in the initial demonstration video. The promised comprehensive setup guide holds the potential to empower enthusiasts to replicate this feat and embark on their AI experiments.

Exploring Future Prospects​

The potential implications of emerging technologies in this field is enormous. With the emergence of AMD's latest mobile Ryzen chips like Phoenix, which harness DDR5 memory, the prospect of enhanced AI performance becomes even more exciting. This suggests the potential for even more accessible and affordable AI exploration, enriching the collective knowledge of the community.

Transforming a $95 AMD Ryzen 5 4600G APU into a 16GB VRAM "graphics card" is a game changing experiment! While this experiment may not replace high-end GPUs in professional AI settings, it undoubtedly highlights the versatility and untapped potential of existing hardware. As the administrator of the Big Hardware forum, I'm genuinely impressed by the user's creativity and resourcefulness, and I eagerly anticipate the promised in-depth video guide that will shed light on the setup process, enabling enthusiasts to embark on their AI journeys with newfound inspiration.

Additional Resources:
For those interested in delving deeper into the discussion and details, you can find the original Reddit thread that sparked this journey:
You must have 10 posts to see the links. Currently you had 0 posts.
.

Feel free to share your thoughts, questions, and insights about this remarkable experiment in the comments section below!
 
New member
Joined
Aug 19, 2023
Messages
3
Credits
57
Integrated graphics don't come with their own VRAM, so you'll need to allocate some of your RAM for them.

It's unclear whether assigning 4GB of RAM to a 5600G is feasible; I've usually seen motherboards allow up to 2GB. However, here's a simple approach:

  • Restart your computer.
  • Press Delete repeatedly during the restart to access the BIOS.
  • Hunt for an option like "Integrated graphics dedicated memory" and check its maximum setting.
  • If 4GB is possible, choose it, then Save and Exit the BIOS.

I don't recommend this for two reasons:

a) Your computer will automatically share RAM with the integrated graphics, exceeding the typical allocated amount.

b) The RAM you dedicate to integrated graphics won't be usable for other tasks. For instance, assigning 4GB of RAM on a system with 8GB total means you'll be left with only 4GB for everything else.
 
Moderator
Joined
Aug 18, 2023
Messages
25
Credits
302
Hey, I just wanted to share some insights on the performance differences between APUs and high-end graphics cards, particularly in handling LLM workloads and memory allocation.

The APU may not match the performance of those drool-worthy high-end GPUs. Nevertheless, there's a silver lining – it doesn't struggle with memory bottlenecks during resource-intensive tasks. Sporting a generous 16GB of RAM, it has ample breathing room, making it a great fit for less demanding tasks.
 
New member
Joined
Aug 19, 2023
Messages
1
Credits
26
Just wanted to chime in on the discussion about those APUs. In my experience, these APUs are getting pretty close to GTX 1050 levels when it comes to their integrated graphics. But here's the thing: to really make them run quick, you've got to pair them up with some dual-channel RAM in the 3200 to 3600MT range. That's where you'll squeeze out the most bandwidth and performance.

And you know what? Overclocking the iGPU can give you a nice little boost. I've seen some solid performance gains when I pushed the integrated graphics a bit.

I can see APUs giving a real punch considering what they are. Just make sure you're giving them the right RAM and maybe a little overclock, and you'll be pleasantly surprised!
 
New member
Joined
Aug 19, 2023
Messages
4
Credits
27
Considering substantial VRAM with ✨2 shared 3090 GPUs, mindful of power draw and 4000 series' efficiency gains. Undervolted 2080 Ti to 160W maintains performance. Awaiting 4080's price drop, 2080 Tis now is budget-friendly on the secondhand market. 2x 2080 TIs outpace a 4080 in FP16, providing 22GB VRAM. Eyeing 7900 XTX's stable diffusion potential and 4080's price decrease to ~$800.
 
New member
Joined
Aug 19, 2023
Messages
1
Credits
5
Could this be also running on a 5500U or 5700U as well?
 
New member
Joined
Aug 19, 2023
Messages
3
Credits
57
Requires motherboard compatibility, but mobile devices often lock BIOS access. Double-check with your brand and consider reaching out to customer support if needed.
 
Top