TR Forums

I've been in search of picking up some 16gb microSD cards for use with my Raspberry Pi 2. Im going to pretty much install a few different OS on each and swap them out as needed. In my search I've noticed that SD cards rated for say 90mb/s don't reach anywhere near those speeds according to user reviews. Some of these reviewers are still happy, while others are angry and return the cards.

This leads me to asking whether this is shady marketing, or simply some form of "peak" speed being reported. Obviously there are other metrics being left out of the mix like 4k IOPS, ect. Also, how does one determine a quality SD card that will reach speeds that you expect? If these were SSDs underperforming like this people would be crying afoul for refunds or class action lawsuits.

PS - I realize the RPi 2 won't be able to take advantage of a card that can do 90mb/s and from the foundations documentation the IOPS are more important.

That peak speed will only ever be achieved in a linear run.

USB sticks and SD cards are horribly slow when you're messing with smaller files. For example, I have a PNY USB 3.0 128GB pen that can hit 100+MB/s. But as soon as I copy a bunch of Word documents or, heaven forbid, a directory with code files, the speed drops precipitously.

The problem is compounded by the fact that 99.99% of people simply don't know how to benchmark. Some will copy a movie or run HDTach and claim that the drive is really fast all the time, when it's not. On the other hand, others will copy 250MB of 100-500KB Word and Excel files and then say that the drive is terrible - in an unfair scenario.

Linear for sure, I explained why that was the case to another tech (in his early 60's) who thought a larger number of smaller files would be quicker.

I guess the best test enviroment for these cards would be to put them through some of the same testing that SSD drives go through. In theory they are similar enough that it should at least give us a frame of reference.

Yeah.....Amazon still has a Sandisk 16GB for $5.00 as an add-on item, but it's Class 4. I read somewhere on the RasPi site that they recommend a Class 6 card to be safe, so I spent $3.00 more and got a Class 10. The "class" is supposed to refer to the minimum write speed in MB/s. Exactly what methodology they use, I'm not sure, but there should be plenty of reading available.

morphine wrote:

The problem is compounded by the fact that 99.99% of people simply don't know how to benchmark. Some will copy a movie or run HDTach and claim that the drive is really fast all the time, when it's not. On the other hand, others will copy 250MB of 100-500KB Word and Excel files and then say that the drive is terrible - in an unfair scenario.

Indeed. Or they haven't verified the bottleneck isn't somewhere else (I haven't looked at SD card reviews lately but I recall seeing one where the tests were done via a USB 2.0 reader, which of course tops out at ~30MB/s).

Claims on the RPi forums (Q&A) claim the SD slot can benefit from up to Class 6 and no higher. Bear in mind that the class ratings are extremely broad and don't define things like IOPS and its listing an average speed of read/write.

Otherwise the Q&A claims something like 34MB/s to be the cap on the SD slot which uses the same USB controller (480mbps USB 2.0) as the USB slots and NIC. USB 2.0 can actually handle a max theoretical speed of 60MB/s. But IIRC the controller on the RPi has a sort of reservation whereas the NIC actually has all 100mbps set aside, i'd imagine they did something similar for the SD card slot... perhaps 100mbps and any additional unused it can take....?