Hacker News

points by andai 1 day ago | 0 comments

> Kugelblitz shows the Earth and its neighbors as a globe and as an atlas you fly over smoothly like a pilot, even on an M1.

Even on M1? Didn't Google Earth run smoothly when it came out? (Which was apparently 25 years ago, oh my goodness!)

[–] Nekorosu 1 day ago

I was going to highlight the same thing. M1 is very powerful and this kind of application would be more limited by the amount of unified RAM. If it works well on 8Gb it's a good one. Mentioning M1 like this actually makes the app look bad.

[–] kiwibyproxy 1 day ago

did you check out the page? this features a "flying mode" in 3D terrain, streamable to Vision Pro

[–] flohofwoe 1 day ago

It's still doing nothing that a 20 year old flight simulator wouldn't be able to do. The one potential performance problem on a laptop with HighDPI display might be the pixel fillrate to render at full resolution, but the pixel shaders in Google Earth should be extremely cheap, and if that's not enough it could drop down to half resolution with 4x MSAA without much quality loss (but an M1 GPU easily has enough oompfh per pixel on a Retina display for what Google Earth needs to render).

[–] echoangle 1 day ago

FSX from 2006 had 3D terrain you could fly over and streaming to a VR headset isn’t that hard either. An oculus Quest has effectively mobile phone power and can run VR games on device.

[–] someguyiguess 1 day ago

Yeah in 480p

[–] VMG 1 day ago

There is a Netflix miniseries about how two German underdog hackers built the first Google Earth precursor in 1994(!)

https://en.wikipedia.org/wiki/The_Billion_Dollar_Code

(IIRC the premise was that the underdogs invented something like mipmaps, which I found hard to believe. The core tech was probably already invented by 3d shooters at that time)

[–] LukaD 1 day ago

It's not the same mipmapping that games use really, but dividing the map data into a quadtree with progressively finer resolutions.

It's more like turning mipmapping into a paging system.

[–] alex0015 1 day ago

I remember that day, 25 years ago, when my dad showed me Google Earth on a Dell Dimension desktop with a Pentium 4 processor. I actually remember the experience pretty clearly. The globe rendered smoothly (at 1024x768) but our internet connection was so slow that it took about 10 seconds for each new step of image resolution to pop in.

[–] jandrewrogers 1 day ago

In the earliest days it required expensive GPUs to run well.

[–] tomaskafka 1 day ago

For agents, M1s are the ancient machines, made long before they were born, and, the horror, barely enough to provide enough nutrients to even a small LLM.

[–] LoganDark 1 day ago

> even a small LLM

So, a medium language model?

[–] xp84 1 day ago

Right? Bloat is slightly more aggressive than Moore’s Law!

[–] walrus01 1 day ago

There's a ton of single threaded performance problems in Google Earth Pro if you watch its activity closely, when you have something like 150MB-300MB of KML placemark lat/long locations and lines loaded. Moving to a new geographic area will cause it to re-read the entire KML for an area and render the placemarks/lines onto the map using *only one CPU core* no matter the capability of your system. There's tons of other things it does with only one core.

Load the fiber maps for a facilities based ISP that spans a few county size up to a whole state size of scale, with good granular detail (like splice case and pedestal locations, etc) and watch Google Earth Pro slow to an absolute crawl even on a $599 current gen Intel core-whatever or AMD Ryzen single socket workstation CPU.

To be fair to it, people are trying to do all kinds of things these days with KML files that it really wasn't designed for, or ever intended for, and are much better done with professional grade GIS software.

[–] renehsz 1 day ago

That makes the quoted statement in the parent comment even more egregious. The goal should be to outperform Google Earth Pro by a large margin.