Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Even if it may still take some years, I have really high hopes for chiplet GPUs.



Would the GPU have its own RAM or simply share address space with a CPU?


It's own RAM. The memory bandwidth of my GPU (616 GB/s) is about 23 times higher than that of my system memory (26.5 GB/s). And I have applications that utilize 90% of that GPU memory bandwidth.


Thanks! I've written code for discrete GPUs but not integrated, and so I'm unsure what the advantages would be. Would there be lower latency for transfers, then, if not higher throughput?


For real 3D performance, it would need dedicated gddr. Regular ram won't cut it for a Gpu at the high end. If all they want is mid-tier, DDR4 would work.


DDR4 isn't even adequate for mid-tier. NVIDIA's lowest end desktop GPU (GT1030) comes in DDR4 and GDDR5 variants and the DDR4 version has roughly half the performance.


I believe chiplet gpu is meant to be two or three chiplets in one graphics card, not an apu design.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: