lee_mcr
Member
OP
- Joined:
- Sep 2024
- Posts:
- 242
- From:
- Manchester, UK
Right then, been benchmarking OPCache across half a dozen shared hosts for a client project and the variance is *proper* wild. Some advertise it enabled but `opcache_get_status()` returns nada, others throttle you into oblivion the moment your hit ratio crests 85%.
My current setup on Hostinger (shameless plug, I work there ): 96% hit ratio, 128MB allocated, but I'll be honest—we're on aging E5s and neighbor noise is real. Saw a bloke's WooCommerce spike because the chap in the next container was running a cron from hell.
Contabo looked promising: Ampere Altra, hit 99.2% in my tests. OVHcloud? Claimed OPCache, turned out it was "available on request" which meant ticket roulette.
What I actually need: working status, no hidden throttling, and enough isolation that someone's rogue WordPress doesn't tank my compilation cache. Budget's flexible if the silicon's modern. Anyone found ARM-based shared hosting that doesn't lie about specs? Cheers lads.
Honey badger don't care... about downtime
armstrongvds
Member
ARM Enjoyer
- Joined:
- Jun 2024
- Posts:
- 199
- From:
- Seoul, KR
Contabo's Ampere stuff is insane value! I benchmarked their Altra Max against some EPYC box and it ran circles around for PHP workloads, like 40% faster opcode hits! I used https://github.com/masonr/yet-another-bench-script for that.
Pro tip: check `opcache.memory_consumption` in your phpinfo. Contabo advertises 256MB but I got 512MB on one node, probably config drift lol. Their neighbor noise is way lower than x86 hosts I've tried, something about the core density.
Have you looked at HostHatch? Not ARM but their Ryzen 9 nodes with LVE limits actually respect the cache isolation. 98.7% hit ratio sustained over a month for me.
one small ping for man...
olya
Member
Always Free
- Joined:
- Jun 2024
- Posts:
- 159
- From:
- Warsaw, PL
If cost is the primary constraint, Oracle Cloud free tier deserves a look. Not shared hosting, but the Ampere A1 gives you 4 OCPUs and 24GB RAM at zero cost.
Pro tip: configure OPCache properly or you'll waste it.
1. `opcache.enable=1` in php.ini
2. `opcache.memory_consumption=256` (I use 512 for Magento-likes)
3. `opcache.max_accelerated_files=100000` — default 10k is too low for modern frameworks
4. `opcache.revalidate_freq=2` in production, 0 only for dev
Heads up: the free tier has no SLA. I run monitoring against a paid GreenCloudVPS instance as hot standby. The A1 chips are genuine Ampere, not rebadged server rejects. Hit ratios consistently 97-99% if your code isn't doing something silly like `opcache_reset()` every request.
licensing is a suggestion
armstrongvds
Member
ARM Enjoyer
- Joined:
- Jun 2024
- Posts:
- 199
- From:
- Seoul, KR
Lol olya with the numbered steps again. Respect tbh Debian on a 5 dollar RackNerd box. Opcache hit 94% but I was sharing with like forty wordpress installs. One neighbor got pwned and the load went to 80 Moved to arch on CloudCone vps. 512MB cache all mine. No neighbors no noise. VPS hosting is a trap
one small ping for man...
wendy
Member
- Joined:
- Jun 2024
- Posts:
- 164
- From:
- Manchester, UK
Have you checked your own code? I moved a client's PHP app from IIS to a Linux shared host, supposedly better OPCache. Hit ratio started at 60%. Turned out their autoloader was generating unique filenames per request through a timestamped cache-bust parameter. The host was fine. The code was poisoning the cache. Windows Server with IIS and WinCache has similar issues, but at least the management GUI makes it obvious when your allocation is exhausted. Most Linux shared hosts hide that behind `phpinfo()` archaeology.
Have you tried restarting it?