Daily Tech News 5 November 2026

• The CEO of Team Group - a maker of memory modules and SSDs but not of the chips themselves - says that the shortage will get worse next year before probably getting worse still in 2028. (Tom's Hardware)
  
  He also notes that the Big Three memory manufacturers whose profits now need to be calculated using the alphabets of dead languages rather than regular numbers are in no hurry to reverse this trend.
  
  Of course this means they'll get their lunch eaten in the long term by Chinese manufacturers like CXMT and YTMC, and good riddance.
  
  
  
• China meanwhile is still well behind the curve on complicated logic chips like GPUs, and has turned to Nvidia with a $70 billion order for export-friendly RTX Pro 5500 models. They'll get 1% of that if they're lucky, because Nvidia can't get the GPU chips or the memory chips. (Tom's Hardware)
  
  A pox on every house.

• The PHP maintainers knew of the recent security bug for two years before it was patched. (Tom's Hardware)
  
  Patched only for certain version, leaving many people - me, for one - scrambling to make somebody else's outdated PHP code run on a new, patched version of PHP.
  
  Thanks, guys.
  
  
  
• AMD is forecasting revenue of $17 billion this quarter - more than Intel - after booking a year-on-year revenue growth of 54% for Q3. (Tom's Hardware)
  
  And they don't even make memory chips.
  
  Well, not big ones.
  
  
  
• The hackers who destroyed food in European supermarket freezers for the second time recently also hit blood banks. (Yahoo)
  
  Fortunately it seems that the blood banks aren't run by idiots, and when the temperature alarms went off, staff yanked the threatened supply from the stricken coolers and put it in unstricken ones not connected to the internet.
  
  
  
• Insects don't swarm at LED lights. (Birds Outside My Window)
  
  Horseshit.
  
  
  
• Ripple-65B Preview is an open-weight (like open-source) AI model for coding. (Ricursive)
  
  What is interesting about it is that on common tests it outperforms models like Meta's Muse Spark Open 816B. Which may not mean much unless you know that the numbers represent approximately how many gigabytes of RAM you need to run them.
  
  65GB is far more reasonable than 816GB. It is a dense model, so either you'd want a huge and horribly expensive graphics card, or something like AMD's Radeon Max, Nvidia's Spark, or Apple's anything, which have unified memory so a 128GB system can dedicate most or all of that to the LLM alone.
  
  Or, say, a Ryzen 255 mini-PC with 64GB of RAM. Not the fastest but gets the job done. And has an added bonus in that I already have a couple of them.
  
  
  
• Speaking of Nvidia's Spark here's a review of an Asus 14" laptop with one. (Notebook Check)
  
  Seems like general-purpose performance is actually pretty good. Damn expensive though.
  
  
  
• Not everything is awful. My local supermarket has gluten-free dino nuggies in stock. Never thought I'd see the day. (I have celiac disease so my diet is pretty restrictive, and it's surprising sometimes to see what finds its way out to my mid-sized country town.)

posted by Pixy Misa at 04:00 AM

| Access Comments