marcusarmstrong

Challenging in an apartment setting, but the state of the art here is an Energy Recovery Ventilator (ERV) installed in your home to continuously refresh the air.

Rim brakes need frame reinforcement at their mount points too, albeit not adapters, so that’s 20g. Hubs are 20g heavier each… I wouldn’t really go for a 20 spoke wheel myself but if you did I suppose that would save 20g. 50g for thru axles. 120g on the brifters.

So to add it all up you’re looking at 770g in components/additions, offset by 75g skewers, 310g calipers, and an added 25g or so per rim for the brake track. So discs are naively something like +335g. I obviously agree it’s a heavier system, but I don’t think ballparking the penalty as triple what it actually is is reasonable.

In 2016 maybe.

On a disc brake bike, calipers are ~100g, rotors ~120g, and then you’ve got ~150g of tubing/fluid. They’re certainly heavier, but the components don’t even add up to 1000g, nevermind when you offset the rim brake component weights.

I’d be particularly interested in something along these lines in a 1U form factor if anybody is familiar with such a thing. Obviously the actual fan-less-ness is less important in that use case but I have a hard time justifying high wattage hardware for my home server operation, and racking provides a lot of convenience.

A TDP comparison is certainly reasonable, though we don't actually have real power consumption metrics so it's a challenging thing to really evaluate. But more to your question, it depends on whether you're trying to evaluate _this chip_ or the _chip family_. Personally I'm not going to be buying anything with the M1 in it because they aren't machines that fit what I need, but I'm extremely interested in benchmarks of the M1 because of what they tell us about the hypothetical M1X (or whatever it ends up being called) in a body that I would, in fact, purchase.

Worth qualifying the comparisons here with core counts... these are mostly 8 core/16 thread parts, with a few 6 core/12 thread parts, vs the M1 4+4.

A great question for the folks who actually do CPU design! As somebody who writes JS rather frequently I’m not complaining.

That's not true at all. HTTP/2 is hugely impactful without using push at all—multiplexing beyond 4/8 streams enables radically different, performant bundling strategies, all "for free" with zero configuration, just by the act of using the protocol.

Wikipedia claims MTr/mm^2s of:

  Intel 14: 37.5
  Intel 10: 100.8
  TSMC 16/14/12: 28.88
  TSMC 10: 52.51
  TSMC 7: 96.5 to 113.9, depending on iteration.
  TSMC 5: 173

150% vs Nashorn is an important caveat here. At {employer} we started our SSRing via Nashorn and later moved to a separately deployed Node service for performance reasons... and saw a 4x drop in time per request for rendering tasks.

Can’t feature detect during SSR, so making an educated guess and falling back gracefully (when possible—many times it just isn’t, as with ES6 syntax) is important to get initial page load right by default.

IMO it’s explained better in the Client Hints specification.