joakleaf

10 ms·

61.
by joakleaf 4y ago
Apple compared M1 max to RTX 3080 (mobile) which was a stretch. M1 ultra was compared to RTX 3090 which was a larger stretch. The M1 max deliver about 10.5 tflops The M1 ultra about 21 tflops. The desktop RTX 3080 delivers about 30 tflops a
62.
by joakleaf 4y ago
Not sure if you got through clipping, but it was one of those things I had to go through first at some point in the mid 90s myself. I feel your pain, but after having implemented it about 5-10 times in various situations, variants and langu
63.
by joakleaf 4y ago
The problem with the epidemic was/is that you can’t easily decide whether you want to take the risk of getting ill or not. And trans-fats are often used in places where you may not be aware of them. Smoking is a clear choice, and you a
64.
by joakleaf 4y ago
Yes, I also highly recommend the night tour of Colosseum. I've done day tours several times. Last time was was in October last year, and it was insanely crowded so I can't imagine how it is now. Definitely feel like there are more
65.
by joakleaf 4y ago
So... the last 3-4 Apple events have had minute long segments with Johny Srouji presenting the Apple Silicon chips. He is Apple's senior vice president of hardware technologies. There is an interview with him in the Wall Street Journal
66.
by joakleaf 4y ago
Unfortunately, it is not even consistent: • In January 2021 he said end of 2021 [2]. • In May 2022 he said May 2023 [1]. So he moved from around 2 years to 1 year. Looking at the rate of progress from Tesla FSD videos on YouTube, I wouldn&
67.
by joakleaf 4y ago
Apple's 20% was for multi-core.
68.
by joakleaf 4y ago
No they didn't reveal single core performance. It is commonly assumed that M1 shares high performance Firestorm cores with A14. It looks like Geekbench scores for A15 over A14 is about 18% in multithreaded, and 10% in single thread. It
69.
by joakleaf 4y ago
An outlier for m1 ultra was what was reported by parent.
70.
by joakleaf 4y ago
That's kind of disingenuous. Searching on Geekbench for Apple M1 ultra single core scores returns values mostly in the 1770-1780 range. E.g. https://browser.geekbench.com/v5/cpu/14597244 Most 12900K score are
71.
by joakleaf 5y ago
Yes, and Ken Silverman got contracted to develop the engine behind Duke Nukem 3D and the other 3D Realms shooters.
72.
by joakleaf 5y ago
In Harvard Graphics, I am pretty sure it was used with the "Scroll" transition in slide shows. That option is visible in at 15.34 in your video for the "Euromap.cht". As I remember, it only ran smoothly (i.e. used the sm
73.
by joakleaf 5y ago
id Software and Carmack often gets credit for this, but I remember seeing a number of games already using smooth-scrolling in EGA before. Zanygolf from 1988 and Rick Dangerous 1989 comes to mind (albeit vertical scrolling). Harvard Graphics
74.
by joakleaf 5y ago
The Cruise tech presentation from last fall seems very relevant: https://youtu.be/uJWN0K26NxQ (Highly recommended for anyone interested in self-driving cars!)
75.
by joakleaf 5y ago
Cruise actually consider both social dynamics and uncertainty (i.e. what can hide behind an obstacle, or where are pedestrians/bikes/cars likely to move to). If you are interested in self-driving cars, I can highly recommend their
76.
by joakleaf 5y ago
Well, for one thing because we outside China would automatically assume they are showing the same thing inside China as we see. Now we know that is not true. That is deceptive. ... And a a shareholder, I was also kept in the dark -- Which i
77.
by joakleaf 5y ago
Protected mode using 16 bit segment:offset addressing could use more than 1 MB of memory. The segments were 64K, but the base address of each individual segment was defined in a descriptor table and could point to anywhere within 24-bit ad
78.
by joakleaf 5y ago
For perspective correct texture mapping quake did the distance divide every 8 pixels on the FPU, and the affine texture mapping on the integer side of the CPU in between (you can actually see a little bit of “bending”, if you stand right ne
79.
by joakleaf 5y ago
I think, what is often overlooked as to why the PC won the market is exactly its hi-res textmodes; 1) The BIOS was insanely slow yes. But on the other hand it was really fast when you wrote to the screen memory directly (address $B8000). An
80.
by joakleaf 5y ago
For anyone interested the Monto Carlo Tree Search used for the planning (that the tentacle shows) is described here (from the AI day video, at the 1h21m50s): https://youtu.be/j0z4FweCy4M?t=4910
81.
by joakleaf 5y ago
As I understood the Tesla AI presentations the path is determined using a Monte Carlo Beamsearch which looks for a feasible path while optimizing an objective that includes minimizing sideways g-forces and keeping the paths derivatives low
82.
by joakleaf 5y ago
There was a little bit of information on the chip, but not much: A13: 8.5 billion transistors. A14: 11.8 billion transistors. A15: 15 billion transistors. A13 neural engine: 5 trillion operations / sec. A14 neural engine: 11 trillion o
83.
by joakleaf 5y ago
... and dropping prices doesn't necessarily meaning dropping revenue or even profit. Intel's per chip profit may drop, but if they sell more because of lower prices, they may actually increase their overall profit. It is really ha
84.
by joakleaf 5y ago
It is actually not clear to me if the service will become profitable: Elon Musk recently said SpaceX needs $30 billion for Starlink to survive[1] (meaning it isn't profitable now). He also said that they are aiming for 42000 satellites
85.
by joakleaf 5y ago
I understand you, but I suspect that scattering atoms of e.g. rare earth metals across the planet makes them difficult to reuse. Would probably be much easier, if the materials stay together.
86.
by joakleaf 5y ago
While I understand your line of thinking, I think is bizarre that short life span is seen as good thing in a world, where we should probably be careful with how we use every single resource for the sake of the environment and future generat
87.
by joakleaf 5y ago
We are negative, because it seems either Musk doesn't seem to fully respect and appreciate the difficulty and the steps required to build a humanoid robot, or he is lying about what he thinks is possible at the expense of customers, sh
88.
by joakleaf 5y ago
If Musk was really serious about this, he would start with smaller steps: 0) Finish the tentacle charging robot they demoed in 2015 (roll out to superchargers) 1) Build delivery robot on wheels using FSD (like the Starship Technology's
89.
by joakleaf 5y ago
You are right, sorry... Still not cheap though.
90.
by joakleaf 5y ago
No, but at 200 million for 3 days (1440 minutes) it is still $139000/minute, or almost $1.4 million for 10 minutes.

More ›