Upgrade from 120 GB Crucial M4 to 480 GB Crucial M500 SSD

I didn’t really need to upgrade my solid-state drive. The Crucial M4 is a fast and reliable drive, and even at only 120 GB, I still had plenty of free space. We had another machine that needed a replacement drive, though, so I decided to use my Crucial M4 for that machine and buy a new drive for my own workstation.

I had my eye on the new Crucial MX100 drives. I even ordered one from Amazon that was available with Prime shipping. I didn’t notice until the order was complete that the drive was back ordered and wouldn’t ship for almost three weeks. I canceled that order and decided to go with the 480 GB Crucial M500.

The 512 GB Crucial MX100 is a faster, a little bigger, and costs a bit less than the 480 GB M500 drive, but none of that mattered to me at the time. I needed to start moving drives around sooner rather than later.

Better performance just doesn’t matter

I ran some benchmarks on this new drive, and the new SSD is quite a bit faster than the old Crucial M4, but that doesn’t really matter. If it weren’t for the benchmarks, I wouldn’t be able to tell you that the Crucial M500 is any faster. My computer booted up fast before, and it still does. Applications used to open very, very quickly, and they still do.

I’ve been saying for a long time that the biggest value a solid-state drive brings over a traditional hard drive is the much-improved random access. Spinning hard disks have been stuck at around 100 to 200 seeks per second since the late 1990s. Even an SSD from a few years ago can manage several thousand seeks per second.

It is those random I/O operations that count. The maximum sequential throughput that the drive can sustain is almost meaningless. You won’t notice it unless all you’re doing is copying large files around your local machine.

Size matters

In my computer, I have one solid-state drive and a mirrored pair of 1 TB 7200 RPM disks. This has served me quite well. The only thing I didn’t have room for on my SSD were some of my larger games, but it was no big deal keeping those on the 1 TB RAID 1 array.

Upgrading to an SSD that is four times bigger than the old one has been a very pleasant experience. I was able to move a whole bunch of stuff from the slow RAID array over to the SSD, including the “huge” 90GB Steam library. I still have over 150 GB free. That’s five or six times more free space than I had on the old solid-state drive.

It really is wonderful that spacious solid-state drives are finally becoming affordable.

Benchmark caveats

I ran my obligatory Bonnie++ benchmarks on the Crucial M500. There are plenty of other benchmarks available these days, but I’ve been running Bonnie since the late nineties. Somewhere here in my archives I have Bonnie benchmarks of old IDE disks, and benchmarks of my old 8-port 3ware IDE RAID card.

So far, I figure it is best to keep running benchmarks that I can compare to my ancient machines.

I did something very different this time. Bonnie defaults to using a scratch file that is twice the size of your available RAM. I usually reboot and tell the kernel to only use two or four gigabytes of RAM. This helps make sure the benchmark doesn’t take too long to run, that I don’t needlessly waste write cycles of the flash chips, and that I even have enough room to store a 64-gigabyte file.

I just ran the benchmark without rebooting this time. I had almost 400 GB free on the drive, and this drive is pretty fast, so it didn’t take long to run at all!

I am also using full disk encryption, so I am artificially limiting these drives. They’re still much, much faster than they need to be, but don’t be surprised if my numbers come up lower than you expect.

And finally, the obligatory benchmarks

Version  1.97             ------Sequential Output-------- --Sequential Input-- --Random-
Concurrency   1           -Per Chr- --Block--- -Rewrite-- -Per Chr- --Block--- --Seeks--
Machine              Size K/sec %CP K/sec  %CP K/sec  %CP  K/sec %CP K/sec %CP  /sec %CP
FX-8350 480 GB M500   63G   483  99 406072  55 150005  34  1362  99 309761  38 +++++ +++
FX-8350 120 GB M4      7G   632  99  96865  10  77523   7  3124  99 300500  15  4234 104
FX-8350 120 GB M4      7G   642  99 178581  28  82598  25  3065  97 200485  32  2439 179
Laptop 80 GB X25-M     8G 56415  89  87157  11  39827   9 69707  98 298590  29 16150  45

I included two sets of results for the Crucial M4. I know those benchmarks were both run shortly after I built my new desktop machine, but my notes aren’t giving me a hint as to why the block input and output results are so wildly different between the two. I thought it best to include them both.

I also included the results of one of my old benchmarks of my Intel X25-M G2. That one is cheating a little bit, because it isn’t encrypted. You might also notice that the X25’s “per character” results are orders of magnitude faster. I’m not sure why that is, but all of the Bonnie benchmarks that I’ve run in recent years have resulted in abysmally slow “per character” throughput. It doesn’t matter which drive I test or which machine it’s in.

Analyzing the results

In my tests, all three drives are capable of sustaining about 300 MB/s sequential read speeds. Other people have seen higher read speeds, so I’m assuming encryption is causing this bottleneck for me. I’m not worried. That is plenty fast enough for my purposes.

I was most impressed by the write speeds of the Crucial M500. It is more than twice as fast as my old Crucial M4. What’s more surprising to me is that the encryption isn’t limiting it to 300 MB/s. I’m guessing that writes are easier to encrypt in parallel, so my workstation’s 8 cores are being better utilized during writes.

I had to check the Bonnie++ documentation to find out what it means when the result of a test is a row of plus signs instead of a number. It does this when a test completes in less than 500ms. I have to assume that means the Crucial M500 is beating my old X25’s 16,150 seeks per second.

The drive I chose not to buy

I also looked at the 500 GB Samsung EVO 840. The EVO is comparable in performance to the Crucial MX100, and both are a little faster than my Crucial M500. However, the Samsung EVO was $40 or $50 more. I didn’t see the need to spend more for performance I wouldn’t even notice.

Which drive should you buy?

I combed through all sorts of benchmarks before buying the new drive. That’s what I do. I enjoy reading benchmarks and specifications. It is really a waste of time. Most of the solid-state drives on the market today are extremely fast, especially the ones from all the major makes and models.

You’d have a hard time buying a “bad” drive from any of the popular SSD manufacturers. Intel, Crucial, and Samsung have been making fast and reliable drives for several generations now.

If you’re spending extra money to pick an SSD that makes higher numbers in a benchmark, you’re probably not going to be able to notice your investment without a stopwatch. My wife’s computer still has the hand-me-down 80 GB Intel X25-M, and it still boots up fast and is as quick and snappy as my computer.

That said, had the Crucial MX100 been available, I would have chosen it instead. It is faster, bigger, and costs less than the Crucial M500. I’m not unhappy that I had to choose the lesser drive, because I’ll never notice the difference.