even the RTX Pro 6000 is slightly cut down (only 4SM or 13% of an RTX 5060).tells them they got the "garbage" (i.e. reject dies) with their $2000 mid-range GPU.
even the RTX Pro 6000 is slightly cut down (only 4SM or 13% of an RTX 5060).
I wonder what Nvidia is doing with the occasional lucky chip with all working SM.
I wouldn't be so optimistic. Nvidia supplies 85% of the discrete graphics card market. If they cut production 30% and AMD increases production 30%, we're still short on supply by more than 20% of what it used to be. And if demand hasn't dropped, what happens to prices?I can't wait when AIBs turn their attention to AMD and Intel cards because there's no longer enough supply from Nvidia. I see good days ahead.
Intel and AMD will increase their wafer orders because someone's gotta keep the hungry gamers fed, right? Intel and AMD cards not that in demand for AI because CUDA so what else can they do but order more wafers for consumer cards and try to cause a dent in Nvidia's marketshare?Nvidia supplies 85% of the discrete graphics card market. If they cut production 30% and AMD increases production 30%, we're still short on supply by more than 20% of what it used to be.
AMD would rather sell Zen or MI. Intel would rather sell nothing than lose money. No one has to do crap for gamers.Intel and AMD will increase their wafer orders because someone's gotta keep the hungry gamers fed, right? Intel and AMD cards not that in demand for AI because CUDA so what else can they do but order more wafers for consumer cards and try to cause a dent in Nvidia's marketshare?
But there's a saturation point for both of those. The Zen owners will want a GPU to go with their CPU and not everyone wants an MI even if it is priced better or faster. Nvidia got lucky with AI thanks to their headstart with CUDA development. Those familiar with CUDA and seriously in the midst of their AI research have no time to qualify MI for their workloads. Otherwise, Deepseek would've been done on MI.AMD would rather sell Zen or MI.
$10000 for a prosumer GPU or $20000 for a AI/DC GPU versus $2000 for similar gaming GPU. Yeah, I know where I'd allocate my wafers.But there's a saturation point for both of those. The Zen owners will want a GPU to go with their CPU and not everyone wants an MI even if it is priced better or faster. Nvidia got lucky with AI thanks to their headstart with CUDA development. Those familiar with CUDA and seriously in the midst of their AI research have no time to qualify MI for their workloads. Otherwise, Deepseek would've been done on MI.
And who knows if that is more or less than AMD's 5nm allocation? There is little reason to be happy if Nvidia is cutting production of consumer Blackwell even if you're not interested in it.But there's a saturation point for both of those