15 years of FP64 segmentation, and why the Blackwell Ultra breaks the pattern • Hacker News Board

Re: 15 years of FP64 segmentation, and why the Blackwell Ultra breaks the pattern

#91

Post by Paul_Clayton » Fri, Feb 27, 2026, 11:24 PM UTC

kbolino wrote: post #73

thesz wrote: post #32

Earlier quoted context omitted.

> Assume FP64 units are ~2-4x bigger. This is wrong assumption. FP64 usually uses the same circuitry as two FP32, adding not that much ((de)normalization, mostly). From the top of my head, overhead is around 10% or so. > Why would gamers want to pay for any features they don't use? https://www.youtube.com/watch?v=lEBQveBCtKY Apparently FP80, which is even wider than FP64, is beneficial for pathfinding algorithms in g…

Has FP80 ever existed anywhere other than x87?

The Motorola 88k and 68k both supported (eventually) extended precision, and, of course, Itanium supported it for x87 compatibility.

https://en.wikipedia.org/wiki/Motorola_88000 (under "Registers": "32 80-bit (88110 only)")

https://en.wikipedia.org/wiki/Extended_precision (see section titled "IEEE 754 extended-precision formats")

Re: 15 years of FP64 segmentation, and why the Blackwell Ultra breaks the pattern

#92

Post by fooblaster » Sat, Feb 28, 2026, 7:39 AM UTC

gdiamos wrote: post #88

fooblaster wrote: post #82

Earlier quoted context omitted.

CUDA was profitable very early because of oil and gas code, like reverse time migration and the like. There was no act of incredible foresight from jensen. In fact, I recall him threatening to kill the program if large projects that made it not profitable failed, like the Titan super computer at oak ridge.

I remember it being less profitable than graphics for a long time. It did make money that would be interesting to a startup, but not to a public company.

Again, it wasn't exactly a huge sink of resources. There was no genius gamble from jensen like you are suggesting. I suspect your view here is intrinsically tied to your need to feel like you and others who are in your position are responsible for your own success, when in fact it's mostly about luck.

Re: 15 years of FP64 segmentation, and why the Blackwell Ultra breaks the pattern

#93

Post by throwaway81523 » Sat, Feb 28, 2026, 10:03 PM UTC

adrian_b wrote: post #49

throwaway81523 wrote: post #7

No mention of the Radeon VII from 2019 where for some unfathomable reason AMD forgot about the segmentation scam and put real FP64 into a gaming GPU. From this 2023 list, it's still faster at FP64 than any other consumer GPU by a wide margin (enterprise GPU's aren't in the list). Scroll all the way to the end. https://www.eatyourbytes.com/list-of-gpus-by-processing-powe...

They did a mild segmentation with that one, by reducing the throughput from 1:2 to 1:4 in the consumer variant, with the hope of forcing people to buy the "professional" version. Even with the throughput reduction, Radeon VII had a performance somewhat better than the previous best FP64 product, AMD Hawaii, due to the large and fast memory. Most later consumer GPUs from NVIDIA and AMD have never approached again such…

Lol I just looked up the Battlemage B580. Seems it was introduced in 2024 at $250 and I looked on Newegg expecting for it to be less by now. Instead it's $425. Damn AI.