When it comes to PC components, it’s all about performance. But the meaning of some performance measures can be very unclear including today’s topic Input/Output Operations Per Second (IOPs). What are IOPs, and are they measurements worth your attention?
Storage Drive Performance Explained
There are two primary types of storage drives available for your PC: hard disk drives (HDDs) and solid-state drives (SSDs). The former are the well-known, traditional PC drives that feature spinning platters and a read/write head that positions itself over the platters to retrieve data or add new data. SSDs, meanwhile, have no moving parts making the retrieval and storage of data much faster.
Due to this difference, we can notice a performance increase just by swapping out an HDD for an SSD. But if we wanted a better understanding of those speed differences or the differences between two drives of the same type, then we need numbers, and that’s where IOPs come in.
PC storage drive performance is typically expressed in two ways: sequential read/write performance and random read/write performance. Sequential read/write measures how quickly the drive can access large chunks of data that are located right next to each other on the drive such as a large video file. Random read/write, on the other hand, is the opposite of sequential. This is when the system grabs smaller files that can be located in disparate parts of the drive such as opening multiple files and programs at the same time.
Sequential reads and writes are typically expressed in terms of how many megabytes per second of throughput the drive can achieve. Random performance, meanwhile, is typically expressed in IOPs.
What Are IOPs?
We’ve already said that IOPs stands for Input/Output Operations Per Second, but what does that mean? It’s a measure of how many tasks (reading and writing data) a drive can carry out every second. In simplistic terms, the higher the IOPs number, the better the drive performs, but it’s never that easy. An IOPs result can be affected by several factors such as the size of the data blocks for the test, and the queue depth (how many data requests are waiting to be processed during the test). There are also other factors to consider such as whether the IOPs number you’re looking at expresses a random read operation, a random write operation, or a mix of the two.
In storage drive reviews you might see a sentence like this: “Random 4K IOPS read and write are rated at 1.5 million and up to 1.8 million IOPS on mixed 70/30 random IOPs.” That sentence was taken from a PCWorld article about a data center SSD, the Intel P5800X. What it means is that the block of test data being written or read was 4 kilobytes, and then the test checked how many times per second that amount of data could be written or read. Read and write tests for the P5800X both topped out at 1.5 million operations per second (for the larger capacity version of this drive), while a mix of 70 percent read operations and 30 percent write operations boosted the maximum IOPs performance to 1.8 million.
The Complications With IOPs
In a perfect world, you’d be able to take a look at whatever the IOPs number is on a spec sheet and easily compare one drive to another. That is not the case, however. First, we need to know the data size being used during the IOPs test. Usually, published IOPs numbers use 4K (4 kilobytes), but they can also use larger sizes. Be sure you’re comparing the same size of test data since that can change the performance numbers.
Another issue with IOPs is the queue depth. A lot of published tests have a queue depth of 32 meaning there are 32 data requests waiting to be written. Manufacturers like tests with larger queue depths, because drives get more efficient the more data there is to be read.
If your drive commonly has a queue depth of 32 then that is a fine measure since it gives you a realistic understanding of performance for your machine. The trouble is a home PC would struggle to get that amount of data in the queue even when under load. Meaning that home users are unlikely to see that higher efficiency quoted in the QD32 tests.
When looking at the IOPs performance of a drive, the more consequential test would be one that has a queue depth of 1. Whether it’s easy to find a QD1 test depends on the manufacturer. Take a look at this webpage for the Samsung 980 Pro, for example, you can find random read and write 4KB, QD1 tests that max out at 60,000 IOPs. The datasheet for Seagate’s FireCuda 530, however, only shows 4KB QD32 tests.
So should you bother with comparing drives if all you can find are QD32 tests? No. Seagate’s results show an IOPs of 1 million on the 4KB QD 32 tests, as did Samsung’s 980 Pro. That makes it a wash on paper for higher-end performance. If you consult third-party reviews, however, they usually come down on the side of the FireCuda 530 as the better performing drive for home users.
So what do we do with all this information? Are IOPs measurements essentially useless for the home PC buyer? If you can find 4KB QD1 or, at the most, QD8 tests from manufacturers, then not necessarily. But as with everything else in the PC universe if you want to make an informed purchase about an essential component you need to consult multiple third-party reviews with similar IOPs tests, and then take their conclusions in aggregate to figure out the best performing drive you can get for your budget.