I would counter that the SSD's aren't entirely the same. The most comparable consumer NVME M.2 would be the 960 pro and it's still around $300. Definitely cheaper than Apple, but it's also not in the $150 range for the same performance.
You are definitely right, an NVMe SSD is going to spank my MX500 in bulk throughput. That is still 3-5X market -- 4X specifically :) in today's market there's no way you can justify a $1200 uncharge from a 500GB to a 1000GB SSD. Of course, Apple doesn't have to, though.
970 Pro is faster than the Apple Macbook SSDs and about 1/3 the price. $349.99 direct from Samsung for 1TB with 3,500MB/s Seq. Read and 2,700MB/s Seq. Write
> performance between different SSDs is basically unnoticeable
Spinning rust drives will likely give you up to ~110 MB/sec.
A SATA SSD will likely give you to up ~550 MB/sec, a 5x increase.
An NVMe SSD will double the write and quadruple the read speed of a SATA SSD, in 1/3 the size.
Yes, any SSD is likely good enough for your average user to feel like a computer is fast. If anything you do is even remotely heavy on I/O, more speed = better.
Modern HDDs easily go beyond 110MB/s, but nonetheless stay well below 200MB/s afaik. It'll probably be somewhere in between. Otherwise your figures are fairly correct, if not on the conservative side for NVMe drives (a Samsung 970 Pro will sustain sequential speeds above 3GB/s read and 2GB/s write).
But focusing on sequential read speed misses the main improvement that SSDs give you over HDDs: random reads and writes ("seek speed"). Reading or writing lots of small files from an HDD will ruin your transfer speeds unless they're written sequentially, whereas a SATA SSD will do quite well and an NVMe SSD will do really well. This is the cause of the noticeable speedup when you upgrade to an SSD.
Generally, yeah, but the exceptions are pretty significant.
Whether or not these use cases are common enough to disprove the "99% of the time" claim is debatable (these kinds of use cases may be far less than 1%) but regardless, check these out:
- 10x latency improvement in SQL Server stress tests between slowest and fastest SSD
- 3x rendering time improvement from slowest to fastest SSD
Now, for anything I do in my daily life as a software engineer? Yeah, SSDs were fast enough years ago. Only time I even scratched the surface of their performance was when I copied a VM from one drive to another.
Relative to each other what really matters are your 4K random read/write IOPS and some measure of reliability. Back in the day there were in fact huge differences between certain SSDs (Samsung controllers vs. Indilinx Barefoot [1] back in the early '10s). Today it's a non-issue.
SSDs years ago saturated the SATA III bus with respect to peak transfers. We're talking an order of magnitude these days, with SATA capping out at 550-600MB/s burst (my MX500) vs NVMe's 3GB/sec (970 PRO). That doesn't paint the real picture to a large extent because 4K random IO is still <<100MB/s in either case.
Large DRAM buffers vs. unbuffered can make a big difference too. Reliability these days from any of the big players is usually very good, since their NAND only comes from a handful of big players -- ditto for the controllers. I guess if you buy shady "Kingston" knockoff SSDs with recycled NAND that's a different matter.
tl:dr; yep, you're right. Any modern SSD from a big player is a good bet, beyond that it's just gamesmanship.
