Are Intel Core i3/i5/i7 Processors Server-Grade?
Solution 1:
From what I've seen, there are five things to be aware of for i-series processors vs Xeon series processors:
- Xeon processors can typically be used in motherboards that support multiple CPUs; i-Series processors cannot (just like the sales guy told you). This is by far the largest difference between between the processors themselves. Certainly a good i7 processor has the reliability and raw performance available to handle a server load, as long as you're still in the single-cpu range.
- Motherboards for Xeon processors are available with more memory slots. This is important if you need more RAM (or will over the life of the server).
- Motherboards for Xeon processors are available with much higher memory and PCI bandwidth, which can make a huge difference in performance, especially for certain types of workloads common to servers.
- Motherboards that support i-Series processors tend to expect desktop grade RAM. Motherboards that support Xeon processors tend to expect server grade (registered) RAM. Of course, you can find server boards and desktop boards that support both types of both CPU and RAM, but the typical situation with each board is to match the RAM type to the supported CPU type.
- New chips from Intel tend to come out as server processors first. This means that most Core i7 processors have a Xeon processor that is almost an exact match. But at the very top end, there will likely be a Xeon processor that doesn't (yet) have an i7 equivalent. This only matters for top-of-the-line hardware, though. — Note: Since I wrote this, it's become more common to have many-core Xeon models that never see an equivalent i-Series release.
We see here the biggest difference between the two is often the supported motherboard rather than the CPU. That said, often those additional features from the motherboard are not needed, and you can get still acceptable performance out of an i7, or even an i5.
Assuming you do have acceptable performance, it's number four above that still matters to you. If you have both an i-Series processor and desktop grade RAM, you're not really running production-level server hardware anymore. It might work fine for a while — in fact, it will probably work fine — but then again it might not, and that's not the kind of risk a good sysadmin wants to take. The failure rate and average longevity just aren't as good... but for the context of this question it's the supporting cast you're worried about here, not the processor itself.
In this case, given that they offer registered RAM with a server-class motherboard, if you are comfortable with the performance level of the hardware then this might be a good fit. I'll add that at time of the original writing, the best Core i3 I could find on Intel's site was a dual core with hyperthreading. The worst Xeon I could find was a quad core without hyperthreading (that lined up more closely with an i5).
Solution 2:
Yes and No... however, it depends on what you define as "server grade".
If you were to get the Xeon equivalent of a Core I cpu, it will technically be a server grade component.
However, if you are upgrading a P4 era Xeon to a modern i3/5/7 (or even a modern Dual Core Celeron!), you will notice immense speed increases in all areas.
However, there is much more to a computer than a processor on its own.
On a server, in addition to the processor, you want good quality, long life components as you want it to last around 5 years (average replacement cycle), you ideally want on board graphics - but just basic/not high end, you want ECC memory and other things - all of this requires a server grade motherboard, and the majority of server grade motherboards only take server grade processors.
All this being said, it completely depends on the usage pattern. For a very small company, I would rather buy an i3 machine and replace it with a similar specification machine 3 years down the line than buying a very expensive machine and keep it for longer. For hosting, unless you specifically need a high end machine, you may as well take the savings as it should be very good and powerful.
Solution 3:
Let us break this down.
Let us take performance first. A million hits per month is such a low volume that you could use pretty much any old chip to handle that. You can run benchmarks on your laptop or desktop to see if that chip can handle your peak loads. Don't forget to populate your database with a few months worth of data first. Most likely your disk performance will be a bottleneck, not your CPU.
Then you are concerned with reliability, i.e running for months and months. Even with the most expensive systems, there are other factors than the CPU that will decide your reliability. Most important is availability of disk storage, which is the most likely thing to fail. You want to ensure you don't lose any data if disk fails. This is commonly achieved by using RAID, either mirror, rAID10 or RAID 5. Next, you want to prepare for disaster, so that you still have your data if the whole datacenter goes down. If you don't want to lose any data at all, you need to replicate your database to another site. If losing half a day or so is no big deal, just schedule backup over network instead.
If your host goes down, either because teh power supply blows, network card, memory , cpu cooing fans or other components fail, you need to have some sort of failover mechanism. This is generally achieved by running in a cluster, where 2 or more systems are hooked up to the same storage. To set up a cluster, you simply define file systems, ip adresses, and applications start/stop/monitoring scripts, and when your server blows, the cluster will ensure that the end user wont even notice. I'd say investing $1k or more for an expensive Xeon does not give you any more reliability. Use the money on buying a 2nd server for standby instead. Or if the hosting company is providing the HW, familiarize yourself with what High availability solutions they offer. If they are professional, they would have this type of support down to a science, including off-site backups, and quick recovery from a server failure. As long as the system options they offer has sufficient performance, then you should be able to sleep at night. If they don't offer any of these features, but only sell standalone server with internal storage, then you should write your own solution, and back up the full OS and application config once, and grab a copy of your DB frequently, so that you can redeploy at same site or different site, quickly. Spending extra dollars on powerful chips does not give you any of the piece of mind you are seeking. Replacing a chip in a server is done in a few minutes, trying to recover from disk failure if you have no backup/recover strategy would take you days, or knock you out forever, or in the case of a business bring them to the brink of bankrupcy. Either your data and uptime is so valuable that you would invest the extra time and money to ensure continuous uptime, or you can go with cheaper solutions such as hoping for the best, but have frequent backups. n Define your requirements, and how much you want to spend, and if that is not enough, pick what features are most important to you. If you set up everything yourself, and run the servers on your own network, you could be up and running with a 2 used desktops or laptops un a cluster for $2-300, plus whatever you need for storage. You can even use internal storage, and make frequent replication from server 1 to server 2 and not even have to get a NAS or other shared storage.
Solution 4:
Simple answer: No
Long answer: Depends
Your usage of the server should tell you if a desktop processor is what you need. Judging from your last question, however, it sounds like you have a very shady company anyway and I would look for other options for servers if I were you. In reality, desktop processors are fine for little test servers, but any server used in production has no reason to not use server-grade parts.