Furiosa: 3.5x efficiency over H100s

Post by SoftTalker » Thu, Jan 15, 2026, 4:32 AM UTC

hadlock wrote: post #45

wahnfrieden wrote: post #32

Earlier quoted context omitted.

GPT 5.2 is an incredible leap over 5.1 / 5

5.2 is great if you ask it engineering questions, or questions an engineer might ask. It is extremely mid, and actually worse than the o3/o4 era models if you start asking it trivia like if the I-80 tunnel on the bay bridge (yerba buena island) is the largest bore in the world. Don't even get me started on whatever model is wired up to the voice chat button. But yes it will write you a flawless, physics accurate flig…

Post by SoftTalker » Thu, Jan 15, 2026, 4:37 AM UTC

linuxftw wrote: post #39

roughly wrote: post #7

I am of the opinion that Nvidia's hit the wall with their current architecture in the same way that Intel has historically with its various architectures - their current generation's power and cooling requirements are requiring the construction of entirely new datacenters with different architectures, which is going to blow out the economics on inference (GPU + datacenter + power plant + nuclear fusion research divis…

Based on conversations I've had with some people managing GPU's at scale in the datacenters, inference is an after thought. There is a gold rush for training right now, and that's where these massive clusters are being used. LLM's are probably a small fraction of the overall GPU compute in use right now. I suspect in the next 5 years we'll have full Hollywood movies being completely generated (at least the specialfx)…

Post by ethbr1 » Thu, Jan 15, 2026, 5:01 AM UTC

roughly wrote: post #7

I am of the opinion that Nvidia's hit the wall with their current architecture in the same way that Intel has historically with its various architectures - their current generation's power and cooling requirements are requiring the construction of entirely new datacenters with different architectures, which is going to blow out the economics on inference (GPU + datacenter + power plant + nuclear fusion research divis…

Post by KiwiJohnno » Thu, Jan 15, 2026, 5:09 AM UTC

ethbr1 wrote: post #55

roughly wrote: post #7

I am of the opinion that Nvidia's hit the wall with their current architecture in the same way that Intel has historically with its various architectures - their current generation's power and cooling requirements are requiring the construction of entirely new datacenters with different architectures, which is going to blow out the economics on inference (GPU + datacenter + power plant + nuclear fusion research divis…

> (at which point Intel typically fired their chip design group, hired everyone from AMD or whoever, and came out with Core or whatever) Didn't the Core architecture come from the Intel Pentium M Israeli team? https://en.wikipedia.org/wiki/Intel_Core_(microarchitecture)...

Post by TurdF3rguson » Thu, Jan 15, 2026, 5:11 AM UTC

jsheard wrote: post #29

wmf wrote: post #28

Earlier quoted context omitted.

That may not be a good example because everyone is saying Groq isn't worth $20B.

They were valued at $6.9B just three months before Nvidia bought them for $20B, triple the valuation. That figure seems to have been pulled out of thin air.