New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Vultr High Freq vs. Dual Xeon (older CPU) server?
This is to be used for a Wordpress/WooCommerce site, cPanel on AlamaLinux 8.
Here are the options:
Dual Xeon E5-2630v4 (20 Cores / 40 Threads) / 128GB RAM
vs.
Vultr High Frequency 4 vCPUs / 16GB RAM
I get that the dedicated server has 8x more RAM, but the the Xeon CPUs are a bit dated as they were released in 2016. (https://www.intel.com/content/www/us/en/products/sku/92981/intel-xeon-processor-e52630-v4-25m-cache-2-20-ghz/specifications.html)
Will the Dual Xeon beat out Vultr's High Freq 4 vCPUs?
Believe it or not, these are both at the same price point. I'm tempted to pick the Dual Xeon dedicated server, but figured I might as well throw it out there for advice.
Thanks in advance for any input/advice!
Comments
The dual xeon almost certainly. Unless vultr gives you 4x 9950x cores which I doubt, and even then they're just a thread and still shared. Vultr pricing is high, and bandwidth is very expensive.
At this point I would use any LET provider who have Ryzen in their offers as it would be way faster in your usecase than these propositions.
I'll go with dedi. Vultr HF is shared, you might get into trouble if you utilize high resources.
Let's assume.
Say you have 4 Dedicated Cores. Let's just say each core runs 3x faster than a single Xeon E5-2630v4 core. Does your workload need to have that much more speed? Does it even benefit from it? Is your workload going have a lot of connections at once that can be processed all at once (so muti threaded. Not only taking on connection at a time.)
If your website can handle multiple connections at once (isn't single threaded), than you can really only handle 4 connections at once with 4 cores. Even if it take a little longer to load a page, it would take 20/40 people 300ms to load a page with the dedi rather than 4 people 100ms to load a page with the VPS than 16 other people waiting 100ms for each person in front of them.
Oh want to scale? Need more ram? Want to run more applications? Well your going to need more cores and ram anyway.
Also that dedi all the way. Same price than yeah take the dedi. You get 100% of everything. That VPS you might have 3x the speed, but might only have 10-20% of CPU usage average. If you where hosting game servers than that dedi probably wouldn't cut it.
Definitely take the Xeon at that price point. Vultr HF is a shared environment, so even if the single threaded performance is faster, you're not guaranteed to be able to use 100% of it all the time. We do have VMs based on Ryzen 7000 series CPUs if that interests you. These are 100% dedicated, with pinned cores on the hypervisor and dedicated disk i/o.
Was going to say this but you beat me to it.
Regarding Vultr, I’m not sure what CPUs they use, but I ran YABS on some of my high frequency instances for the memes and got a GB6 score of well over 2000. Unfortunately I don’t remember the exact number but it was shockingly high. Even the South Africa server had a GB6 of I want to say around 2300. Note that this was over a year ago so the situation may be different now.
My use case was to run game servers, and Vultr still is my preferred provider for exotic regions that are scarce on LET, like South America and South Africa. But if you don’t need that much computing power, go for a dedi and set it and forget it without worrying about “fair use.”
For the OP’s specific case, take the dedi and don’t think about it. A website will benefit way more from all that RAM and CPU cores than 4 high frequency cores will.
depends on the traffic. for a low traffic website, a beefy vps is fine.
Reliability?
Dedicated hardware may require maintenance with hardware compared to "Cloud VPS" which won't ever go down for hardware at-least
Dedis are very reliable too. I have been using hetzner dedis with soft raid hdd for several years, never have an issue, never down. Even the one provider dedis with single hdd or old ks1s with single 2tb hdds, no issue at all for several years.
Server is running AlamaLinux 8, cPanel, Apache, WooCommerce w/ WP Rocket.
So yes it can take advantage of multiple cores, but like you said, even with Vultr's high frequency 4 vCPUs, that's just 4 shared CPUs.
Google page ranking is important, and the site the server is used for has great rankings, so I wouldn't want to mess that up by moving to a dedi only to find out the pages are loading slower.
My only worry is that if pages take longer to load, the site will lose its already great google rankings. Server is running AlmaLinux 8, Apache, cPanel, WooCommerce w/ WP Rocket. All of those can take advantage of multiple cores, but from what I understand, the speed of a single page loading for a single visitor will be based on the performance of a single core.
At least for Vultr (rather than AWS EC2 etc.) the VM will definitely go down if the host node down.
What about upcloud ?
If your site heavily uses cache and it's not like a dynamic content site that almost changes on every page load, just go with faster core. Otherwise, anything that gives you more cores then.
well it is WooCommerce so I consider it "dynamic content" but from what our developer tells me, unless the price of an item changes, or it goes out of stock, or there is a change to the description, etc., it will be cached and even then only the first visitor on a changed page will receive the uncached page.
I'm not in-depth on how WP Rocket generates its cache. However, in the end, after everything is cached, core speed (with only a fractional difference), network latency, and I/O are the only things that matter in how fast a server can render a cached page. For Ajax, I consider a maximum wait time of 1 to 2 seconds to be acceptable since they're uncacheable. This might even be less if your server location is close to your customers.
Vultr is very expensive, yet it is a shared environment. At this price range, it is better to pick a dedicated server and you get 100% out of it. I don't trust vCPUs if speed and consistent performance are important, especially when you are only hosting one or two sites. If you care about reliability, Vultr, might provide you the best compared to others, but idk, it's been many years since the last time I used them, but downtime is inevitable for any provider.
If your current dedicated server candidate doesn't satisfy you, find something better. I see a lot of offers that have higher clock speeds with fewer cores, especially from Ryzen.
That makes a lot of sense. I am looking at some Ryzen servers which are actually cheaper than the Vultr High Freq 4 vCPU plan and provide the same base clock speed and more cores, and anywhere from 64GB to 128GB of RAM.
I've been told to run Geekbench on Vultr and on the dedicated server to make sure, so I'll definitely do that as well. Not sure if there is some other benchmark that can emulate dynamic sites.
Thanks for the help and advice!
Realistically for services like Vultr, the benefit is the hourly billing - that is, you can scale up or down whenever you want. This is also why I use them since I sometimes need a bunch of servers for a few days and don’t want to pay for a full month term, in addition to needing servers in exotic locations like Tel Aviv, Santiago, and South Africa. Most providers here don’t have a footprint in these areas due to their high cost.
To provide an example, in 2023 I needed 400 vCPUs and 400 GB RAM (split across 200 instances scattered all around the planet) for a very busy weekend and was able to fulfill that demand with Vultr for an invoice of approximately $150. The hourly billing, worldwide presence, and on-demand scaling up/down was well worth their fee schedule. Any other arrangement would have cost over $1k since most services want at least a 1 month term, especially if you needed a CPU with a GB6 of 2000+.
If you’re going to run something for 30 straight days and can run it in a common region (US, EU, Singapore), then Vultr (and other similar cloud services) is extremely expensive and you don’t benefit from their perks such as hourly billing or exotic locations. LET providers. Hetzner, or Kimsurfi from OVH are a much better value for your money in this case (which is also the case for most people). I just wanted to share a niche example where Vultr was the most affordable option in a unique circumstance.
As with everything, your mileage will vary based on your specific needs. I still think for what the OP wants, a dedi is better, since CPU clock is unlikely to be the main bottleneck for a website.
what kind of services you deploy?
Vultr's support is also garbage, go with anyone else
Yeah, they definitely are very "hands off" compared to other providers I've worked with. I recall even to help with something where you are having trouble adding an IP to a VPS and it might be something not configured correctly on their side.
It was hosting game servers for a special April Fools charity event. Each server can hold only 6 players and we had at least 1000 concurrent players that weekend so we needed a lot of vCPU and RAM very quickly, yet for only 3 days (Fri-Sun).
That’s not a typical hosting need for most people on here, but it also shows not every client has the same need so different providers are great at different things.