- Joined
- Dec 25, 2020
- Messages
- 8,215 (5.21/day)
- Location
- São Paulo, Brazil
Processor | 13th Gen Intel Core i9-13900KS |
---|---|
Motherboard | ASUS ROG Maximus Z790 Apex Encore |
Cooling | Pichau Lunara ARGB 360 + Honeywell PTM7950 |
Memory | 32 GB G.Skill Trident Z5 RGB @ 7600 MT/s |
Video Card(s) | Palit GameRock GeForce RTX 5090 32 GB |
Storage | 500 GB WD Black SN750 + 4x 300 GB WD VelociRaptor WD3000HLFS HDDs |
Display(s) | 55-inch LG G3 OLED |
Case | Cooler Master MasterFrame 700 benchtable |
Power Supply | EVGA 1300 G2 1.3kW 80+ Gold |
Mouse | Microsoft Classic IntelliMouse |
Keyboard | IBM Model M type 1391405 |
Software | Windows 10 Pro 22H2 |
You do realize that the MI300 is AMD's fastest ramping product in its entire product history? Their data center GPU's sold so well in Q2 that they changed their market forecast by 500 million dollars. People are buying them and will continue to buy them because they perform really well. Nvidia failing to secure demand is more potential market share for AMD.
It is and it is not enough to dethrone or even dent Nvidia's business. It's huge. Zluda (CUDA interpreter for ROCm/HIP) has no future ahead of it due to CUDA licensing terms blocking it

Nvidia bans using translation layers for CUDA software — previously the prohibition was only listed in the online EULA, now included in installed files [Updated]
Translators in the crosshairs.
I'm not disputing the fact that AMD made some wonderful processors with recent Instinct chips, it's just that the software isn't and won't be there. Not until an open-source solution which is more flexible and performant than OpenCL comes to exist. Some newer AI software for training and inference has been written for Radeon given that Instincts perform so wonderfully, but it remains that CUDA software is still superior, corporations are risk averse, and Nvidia's toolkit is just better.