Hello ladies and gentlemen, in this article I will tell you about the difference between server RAM and regular RAM. In fact, there is no fundamental difference between server RAM and consumer RAM based on the same architecture except for some details, but server RAM has a few key features that make it preferable in the corporate segment.
Consumer RAM vs server RAM
So, let’s break down the differences. I won’t go on for half an hour about what RAM is, but just to summarize, RAM is volatile. That is, as long as the memory cards are energized, it will store data, but if it is de-energized, all data from the memory will be automatically cleared.
RAM is fast memory, due to its architecture it provides faster access to data, accordingly, no swap file on a hard disk or SSD can compare in speed.
But such memory has a small disadvantage, in the process of operation it accumulates errors, which in time can become a serious problem.
Of course, not in one day, but ordinary RAM will accumulate a lot of errors in the process of long-term operation, for example, in a month without shutting down the computer. And the only way to get rid of the errors is to de-energize the device.
And for the corporate segment, such an option is not suitable. Servers should run without shutting down for as long as possible, months, even years. Accordingly, a slightly different approach to memory was needed.
Also, desktop RAM has another disadvantage: a limit on the number of slats. Because of the way RAM works, you simply can’t put a lot of strips, because the controller won’t be able to process them all.
First, let’s look at the types of memory. There are a lot of them, if I try to describe everything in detail, the article will turn into a book, so I’ll try to keep it brief.
For example, SO-DIMM memory is often used in notebook computers. SO-DIMM memory modules are more compact and power efficient.
In personal computers, a popular option would be UDIMM . This is a non-registered and unbuffered memory that simply acts as a medium of exchange between disk memory and the CPU. Some critical data is loaded into RAM and stored there to speed up data exchange with the processor.
This memory is characterized by a memory controller that performs memory addressing as well as data transfer. This method does not include the ability to correct the memory without de-energizing.
And also does not allow you to increase the number of RAM strips, because the data controller simply can not work with a large number of UDIMM strips.
And for consumer systems, that’s quite enough.
But server memory is a bit different, register modules are used there.
For example, RDIMM is a type of memory with addressing and buffering. In this type of modules, a special chip takes care of memory addressing. It is usually located between the memory boards on the RAM strip.
Data transfer, of course, continues to be performed by the memory controller, but by not having to address the controller, enough resources are freed up so that the data transfer limit can be expanded.
Because of this, more RAM slots can be installed on server hardware than in consumer systems.
Disadvantages include a slight slowdown in RAM speed.
LRDIMM is also commonly used. This type of memory allows more memory to be installed per server due to its architecture, as the data bus considers these boards to be dual-peer, but they are actually quad-peer.
There are noticeably more memory modules on each RAM board.
Of course, with small amounts of RAM this memory shows a slight decrease in speed, but with large amounts the speed is on par with UDIMM . And LRDIMM is needed just to get the maximum amount of RAM, in other scenarios this memory will be an unnecessary waste of money.
Also, like RDIMM, it buffers memory and addresses memory using its own controller, freeing the data controller from this task.
Of course, there are other options, such as using NVDIMM, which combines volatile and non-volatile memory.
And there is also a very unusual type of memory, which is no longer called RAM – DCPMM in Intel Optane DC Persistent Memory Module. In fact, they are drives that can be up to 512GB and are placed in the RAM slot.
Also, server memory allows you to correct errors without de-energizing RAM, as there is support for ECC.
ECC (Error-correcting code memory or error -correcting memory) is a technology that allows you to analyze memory for errors and to automatically correct them. This means that errors will not accumulate in the RAM and the system will not need to be de-energized for a long time.
Of course, modern RAM is pretty reliable, but if you work for a long time, for example, a few months, the memory will accumulate enough errors , so ECC significantly increases the reliability of server hardware.
Based on the above, we can deduce the key differences between server RAM and conventional RAM:
Using register memory instead of other types of RAM. Accordingly, buffering and memory addressing falls on the RAM itself, not on the memory controller.
Use of non-standard memory types: NVDIMM, DCPMM, which are not in demand in the consumer segment of RAM at all.
Ability to install more RAM slots and with more capacity.
ECC. Automatic error correction allows you to get rid of errors without having to de-energize the device.
There are no other major differences, but what is described in the article makes the difference between server RAM and regular RAM quite significant.
This is where I say goodbye to you, I wish you success and stable RAM, have a nice day!
