Saturday, June 25, 2022

Saturday, June 18, 2022

Puzzle Leaderboard - Chess Rankings - Chess.com

https://www.chess.com/leaderboard/tactics?page=171

I'm #8540 on the chess.com puzzle ratings.  I was expecting that the top ratings would not go too high, but I was wrong.   The top three ratings are all 65540, which for the reasons I give below, I suspect is the highest possible rating.


I find this 65540 number suspicious because the top three ratings are this number.  The maximum value that can be stored by a 16-bit number is 65536.   If you want to save storage space, why use a 32-bit or a 64-bit number to store ratings when a 16-bit number would do?  The 65540 number almost fits.  You can make it fit by making the lowest possible rating the number 5.  Why would you set a lower limit on the rating?  To not accidentally run into a divide by zero problem, which can crash computer code, or other mathematical oddities from having a low or negative number in your equation.



Tuesday, June 14, 2022

Guide: What to do AFTER building your computer...

Re: Watch "Apple just killed M1 - WWDC 2022" on YouTube

I have a big side note here.  Microprocessor instructions are either 8-bit or more likely 16 bits long.  Each instruction represents a particular action to be taken by the microprocessor in machine code.  A typical machine code instruction might be to add one register's value to another and put the result in a specified location.  This would be the equivalent of the "add to memory" button on a calculator.

x86 processors are Complex Instruction Set Computers (CISC), where Arm processors are Reduced Instruction Set Computers.  RISC tries to be more efficient by doing fewer things faster.  There is a 30-year-old battle of architecture between RISC and CISC, and right now RISC is winning.

I wonder if there might be a better way.  I have long imagined a 16-bit or longer instruction where each bit represents a particular action to be taken by the microprocessor.  For example, one bit might be to fetch data from memory, and another bit might be to fetch data from a register, and a third bit might be to add them together.  On a 64-bit processor with a 64-bit data bus, there would be no problem having 64 individual bits each presenting a particular action to be performed.  Such a system might allow the combining of actions in novel ways or the performing of some actions in parallel, increasing efficiency.

On Tue, Jun 14, 2022 at 6:05 PM John Coffey <john2001plus@gmail.com> wrote:
I've seen parts of this before.  M1 is alive and well.  M2 is a more powerful design, but only marginally better than M1, and only will be available initially on a couple of platforms.  I would choose M2 over M1, but I would rather have a 5 nano-meter x86 chip, likely an AMD APU.

The current generation AMD 5700G would check most of my boxes.  It has a PassMark score of 20K compared to 6K of the  i5-7500 on my 2017 iMac, and the roughly 14.6K Pasmark Score of the M1.  BTW, your laptop scored around 10K, but that is still noticeably better than my desktop.   My first priority was to get a more powerful processor and the 5700G certainly fits the bill.  However, the graphics capability is around 900 gigaflops, compared to about 3.6 Teraflops on my iMac and 12 Teraflops on the Xbox series X which uses a custom AMD APU.  In other words, it is not a gaming powerhouse.

I could buy a 5950x with no graphics, which gets a PassMark score of around 40,000.  Then I would have to buy a graphics card.

So there are very strong rumors, which AMD has mostly confirmed, that they are coming out with a "Phoenix" line of APU processors that have much more graphics capability comparable to or surpassing some lower-end graphics cards.  This is what I am waiting for.


--

On Mon, Jun 13, 2022 at 11:19 PM Alberwrote:
This is a good overview of the Apple M2 chip and other Apple upgrades. The new chip is very impressive and will probably have Intel and AMD trying to do something similar in the future.