I'm currently taking stock of our server before we add a new one. I wasn't involved in its purchase (before my time) but the more I learn about its specifications the more concerned I am. Note: Ultimately it will be used as a Database Server in our new configuration (I don't know if this will change the answer to my question or not).
The server currently has a single dual-core Opteron 2212 (2.0Ghz). The problem is that while I've been investigating this processor, I've come across statements like this one: "Dual- and quad-socket set-ups are a must with this chip... It would be pointless to only use one."
Plus, all the benchmarks I've seen for this CPU only show a two-socket configuration.
For example: CPUBenchmarks.net
In fact, I can't even find any benchmarks for the single CPU... everything lists it in a two-socket set-up.
Did we do something wrong here? Should this chip always be used in a dual-socket configuration? If so, why? What's so bad about having only one? And why did Dell allow us to buy a PowerEdge SC1435 from them with only chip installed?
I'm really confused why so many places would seemingly list this as an issue.
No, you did nothing wrong. If servers can be upgraded, then adding a second CPU is simply a feature that you have not exercised. Who cares what other people say? If the system is running fine without any issues, if anything you've saved yourself a few bucks and a little bit on the electric bill. You should be praised assuming that your servers don't need the extra CPU and everything is running smoothly.
All CPUs can be used in anyway they're designed to. If you have just one in place and it works fine, don't sweat it. Both Xeons and Opterons are built for this.
Absolutely nothing. I ran 3 servers for a small office of about 30-40 users all running just one Opteron CPU each. I didn't need to get dual CPUs for each server. The load wasn't that much so why spend more $$$ and consume more electricity when it's not needed.
Rule #1 in business: if logic/reason cannot explain things, it always comes down to money. Actually, in Dell's case, it's always good to offer consumers options. Granted too many options and we become unhappy as they're are too many permutations of purchasing just one thing. This book explains why.
Unless you're running a huge database cluster and need every single ounce of CPU horsepower for ETL, CRM or what have you, there's really no negative points on running a single CPU if that's what's needed. If applications, file servers, email, whatever are all running smoothly and capacity is not an issue, what's the point of adding more muscle when it may not be fully utilized?
Now, if you're looking to vendors for advise on running things, then you're simply getting the sales pitch. Always be weary of the source of information and their goals.
In fact, I'm going to go out on a limb to say simply don't listen, believe or give credit to others when they don't supply any real reason, logic or some explanation as to why they believe what they believe. This rule can be applied anywhere so don't fret on what people advise or say without any reasoning. No reasoning or logic, just dismiss it as noise.
I've run Opteron systems for work and at home. As CPUs they're great as they're usually cheaper then their Xeon counterparts and usually use less electricity. In my eyes, adding another CPU to a system "just because" it would be silly not to is poor logic. Poor logic is never good reason to do (or not do) something. You're doing just fine in questioning this line of thinking. All you need is the confidence to follow through on your logic/reasoning.
One other thing to bring up: If you are going to use it as a database server, you will want to gauge how long your average query is going to be. If you get two long running queries, your database server is going to become unresponsive to the third query (even if it is just a simple single record request). We made this mistake on a database server and the users very quickly became very unhappy. Especially if you are using Stored Procedures with RBAR in them...
After about a week, I placed an order for another dual core chip to fill the second socket. Our database runs our accounting package, which frequently runs queries that feed reports. Some of these queries hit tables with millions of records in them and take 15-30 seconds to run. If your load is different, you might be able to get away with one dual-core socket.
I'm a big fan of your work ;-)
I just took a look at this spec sheet, and it looks like your box can only handle two sockets max. I would guess that it having two cores is perfectly appropriate.
The only reason to think it is underpowered is if it just doesn't hold up in production. We use dual core boxes all the time.
One of the things to consider is the cost from a price/performance ratio. Very often you can get dual processor AMD machines cheaper than the single processor version (and/or the price/performance cost cost down due to the opteron fabric). the price/performance ratio goes up primarily because adding an additional Opteron processor increases memory bandwidth to all processors. In addition, one of the reasons that the Opterons are almost exclusively sold as multi-socket configs is that the selling point for the operton was the fact that it scales multi-processor far better than the xeon equivalent. As far as I know, however, the nehalem architecture removes the issue from the intel side.