looking to upgrade to an hedt setup. Specifically any of the new gen3 threadripper’s, vs intel’s 10980xe or whatever they have for competition. (xeon’s?)
My specific use case is web development though, and particularly with wamp/lamp stack. So php/mysql running on apache or nginx. This is for local development purposes, and im trying to speed up my time spent compiling scripts, waiting for sites to render, (inefficient) queries to complete, etc
Again this is for local dev purposes only, so im not worried about “traffic” per se. More after raw performance
My typical workloads:
Will sometimes run 1-2 VM’s (docker)
Compiling scripts for css/js, using npm/gulp/yarn
testing/viewing sites on multiple browser window sometimes
Typical database queries + processing that data, on any given site
Is web development (or specifically php/lamp stack) more favorable towards single thread/core performance? And what hedt CPU would you suggest for my specific single person use case?
I got the 3900x about a month ago and tested it against my i7 8700k (stock and @5ghz), and the i7 did better in every single test. I also tested the 3900x against the i9 9900k (oc’ed to 5ghz and stock), and it did even worse.
i was super bummed about that and as a result, am probably the only person in the world that has returned their 3900x.
sorry but… you’re not satisfied with the performance you get from an i7 8700k?
…
…
…
How much ram do you have?
Cause I think anything you do will just improve marginally… web development is not that resource heavy… most of us get along with freaking dual-cores on laptops
For the longest time, I was doing fullstack development on a 2.2GHz (turbo) 5200u with 8GB of ram.
Sounds like you need to improve the efficiency of your code or adjust your expectations.
Code takes time to compile. Queries take time to run.
All of this is heavily impacted by your disk speed though.
What sort of wait times are you experiencing? If you’re concerned about that for local dev and you’re looking to go up to 24 cores, (TR3), you’ve got a problem, because if this application ever sees the light of day, you’re going to need a hell of a lot of infrastructure to run it.
Each and every task you just listed here is IO bound.
Exactly this. Spend your money on your storage performance.
Also, this is a reminder to tackle some of your technical debt. If you think you need speed because you’re getting slow Docker/Vagrant/Jenkins performance, tightening up your builds can be orders of magnitude better than a hardware upgrade.
64gb ram @ 3200mhz
500gb 970 evo
(again, currently: i9 9900k for cpu)
so not too throttled there i hope? (ram/ssd) ram could be 3600mhz i guess
When i run apps and websites on actual web hosts (AWS for example), that run xeon processors, etc. There is a huge noticeable improvement in speed vs my local desktop. (For a further example, even like the AWS a1.large instances outperforms my local)
A great example some people could relate to maybe… a shitty plugin-heavy Wordpress site. Compare those load times locally on any consumer setup to when its running on AWS/WPEngine or whatever.
As for code optimizations, etc… well in an ideal world, of course that would be super awesome. But that’s not my point here. If you code, you know there’s a good chance you’ll inherit some complete shit that you can’t do anything about.
I know i can code in notepad on a laptop from 2010. But i can assure you the seconds/minutes spent waiting for it to compile/then refresh the browser/test your changes definitely add up…
CPUs aint your bottleneck son. Cheap out on the CPU do some bottom tear threadripper for the pcie lanes and then get all large NVME drives for your IO problemo solved. Shoot just go first TR its aight and its for the lanes. Also its hella cheap
RAM speed doesn’t matter unless your developing stuff on a RAM disk or something and you just have God tier amounts of memory lol
I’m running 8-10 windows server virtual machines on a i7-6700 with ssd and 64 gigs of ram. I’m usually still not cpu bound unless they all do windows update or something dumb at the same time
Agree with above… ssd and ram is likely what you need. Not cores so much
A couple of docker instances you won’t even get close to maxing a quad core for web development.
If you’re serious about figuring this out, you might consider reaching out to the folks at Puget Systems to see what they have to say about your situation. They’re kinda the gold standard in designing systems for professionals that just work. They’re just not inexpensive.
I’ll just add that my workload includes an instance of SCCM and SQL server as I’m using the box for an SCCM test environment across multiple virtual AD sites (pfsense doing WAN simulation between them).
Another one of the VMs is my Windows+office “workstation” for shit I need windows for.
So it’s not like my VMs are idle either.
A Ryzen R7 or r9 with sufficient RAM and SSD would be even better (which is what I’m using at home)
Edit:
When I say waste of money… well if it was my money I wouldn’t buy into HEDT for that. Well. Maybe I would anyway. Because I like hardware.
But I guess what I’m trying to say is that you’re likely aiming far beyond what is required.
just wanted to update, just in case anybody runs into this problem.
it was basically a setting that was causing all the slowdown.
apparently Bitnami (wamp) adds a bind-address=127.0.0.1 (by default) into mysql’s my.ini config file. I don’t think mysql originally includes this setting in their install
i removed this and ended up getting waaaaay more realistic page load speeds of a few seconds vs like 10-20s. jfc bitnami…
If you think you need speed because you’re getting slow Docker/Vagrant/Jenkins performance, tightening up your builds can be orders of magnitude better than a hardware upgrade.