- Joined
- Dec 29, 2010
- Messages
- 3,770 (0.75/day)
Processor | AMD 5900x |
---|---|
Motherboard | Asus x570 Strix-E |
Cooling | Hardware Labs |
Memory | G.Skill 4000c17 2x16gb |
Video Card(s) | RTX 3090 |
Storage | Sabrent |
Display(s) | Samsung G9 |
Case | Phanteks 719 |
Audio Device(s) | Fiio K5 Pro |
Power Supply | EVGA 1000 P2 |
Mouse | Logitech G600 |
Keyboard | Corsair K95 |
The partners were throwing out numbers all during the release stream. They've already gotten to parity to CUDA with the 200/250. And software optimization is a given not a question. And in case you missed the point with referring to H100's uplift from A100 of 30% but it took 4 years at 4 times the cost is that Nvidia isn't really trying here. The only thing they are really pushing hard is pricing. And thus to have a real player at parity for a much lower price point, it's a positive for the market. And it's clear that even as these major players drill into their own chip IP's so as to free themselves from a suppliers ridiculous profit margin, if someone comes along with a competing product at a fair price they will still gobble them up.When did I say either of those things? Make your point without putting words in my mouth.
Semianalysis conclusion was that MI300X should be well positioned against both H100 and (maybe) H200. Just inferring some things about MI300X itself, and how AMD is choosing to present its performance in the slides.
There's some software optimizations left, and like I said their customer base this time gives them good reason to keep working on that front.
Last edited: