Question on Speed/Wear External SSD

[BigMcGuire]

This is a bit of an obsessive compulsive question... I have a 2017 TB MBP 13'. I have an external 500GB Samsung EVO at my disposal.

Question 1: Does running Windows 10 on Parallels via the external USB drive result in significant/extra CPU usage vs just running it on the internal drive? I do compile 2+ mil lines of code for work regularly.

Question 2: Do you think running games & parallels on the external drive is worth it to save read/writes on the Mac's internal drive?

I have a 1TB SSD inside of the MBP. <cough>. My Parallels Windows 10 is about 85GB. Games (Starcraft II, WoW Classic, etc... = ~30GB). So this isn't a space question, more of a CPU Usage vs Wear on Internal SSD question.

What would you do? Utilize the SSD or --- I'm clearly insane for even asking such a question? (It's ok, won't hurt my feelings). Thanks!


Owned MBP for 1.3 years now and I have 19.1TB of writes (100% health according to DriveDx).
500GB Samsung EVO 850 gets about 366MB/sec read/write via either USB 3.0 enclosure or USB C enclosure.
Hoping to get 6+ years of use out of the MBP. I don't mind the inconvenience of an external drive (at least I don't think so).

 

[BigMcGuire]

Leaning towards not using external drive. Seems to be noticeable CPU usage/lag when transferring large number of files or doing CPU operations in Parallels (that may/may not involve heavy IO to external drive) - that I don't see when running on internal disk.

I realize that most 256GB SSDs can do 200TB of writes and still be ok, so I think it isn't worth saving the writes at the cost of performance.

 

[Howard2k]

I bet you could find the MTBF of the drives and figure that out, but then with the way these machines are built there are many points of failure that would be financially catastrophic to the machine.

It's an interesting question, don't get me wrong, but I think there's a good argument for either side and Murphy's Law says it's the side you're least prepared for.

If the compiles are faster when local, and I'm getting paid for them, I'd run them local and get the performance benefit. Does time = money is what I'd ask.