Slow and inconsistent write speeds from internal Kingston A400 SATA SSD in 2012 27" iMac (TRIM enabled)

[ElCani]

I recently (maybe 2-3 months ago) upgraded my partner's 2012 27" iMac, replacing the stock HDD with a 1TB Kingston A400 2.5" SATA SSD. After installation, I ran Blackmagic Disk Speed Test and, from memory, got roughly 350-450 MB/s read and write speeds. Not fantastic, but much better than an HDD. The internal 128GB blade SATA SSD which was part of the Fusion drive was little slower when tested.

Anyway, yesterday I realised I hadn't enabled TRIM on the new SSD, so I did so using Terminal. Once the machine rebooted I ran BDST again I was surprised to see very slow write speeds, ranging from 10-60MB/s. I tried disabling TRIM but that made no difference. Further testing has revealed write speeds to be very inconsistent, sometimes at 400MB/s and sometimes 10-20MB/s. The attached screenshots are an example of this, the tests being run within seconds of each other and with no other apps open on the iMac. OS is Big Sur via Open Core Legacy Patcher.

Anybody have any idea what might be wrong? I realise the A400 is a cheap SSD but surely it shouldn't be this slow or inconsistent? The SSD is APFS formatted and 63% full. Drive DX shows 100% health on all parameters. My 2011 iMac with a 2TB Sandisk SATA SSD is much more consistent and reads are over 400MB/s.

Thanks in advance for any assistance.

 

[EugW]

That is normal behaviour for that model SSD. At least for Disk Mark. Not sure how it shows up in Black Magic.

Too late now, but I generally would not recommend a DRAM-less model like the A400 for installing an iMac unless you're on a strict budget, since it's difficult swapping drives in those iMacs. If you're buying a new drive, you're usually better off spending a bit more to get one with DRAM if performance matters to you.

That said, if all you're wanting is fast boot times, fast wake from sleep, and fast general simple business computing, then the A400 should be fine despite what the benchmarks show.

 

[ElCani]

Ok, thanks. You’re right in that I should have done more research before buying the SSD, but the computer is only really used for MS Office and web browsing, so maybe not a disaster. I guess what concerned me was the massive degradation from the first speed test I did, and also why each test I run seems to produce vastly different results - any idea why that might be?

 

[EugW]

At this point I wouldn’t worry about it. For your described usage it will be ok.

You should be aware that sequential disk transfer speeds are pretty much irrelevant for your type of usage so such benchmarks are not informative. I know everyone likes seeing those 500+ MB/s transfer rates but it’s not as useful a gauge as the random transfer speeds. Unfortunately this is where the cheap DRAM-less are often significantly slower. The read/write speeds can drop below 25 MB/s in many cases, whereas the better DRAM-endowed drives will be several times as fast for this metric.

However, for most people doing relatively straightforward stuff, that isn’t actually a major problem, cuz that’s still an order of magnitude faster than any hard drive.

Put it this way, I recently put an old Samsung 850 EVO in my 2010 iMac. I was sure to put in a drive of reasonable size (250 GB is much more than we need in that machine) and was a DRAM-endowed model since I never want to swap out that drive again. However, for my ancient 2007 Mac Pro I only put in a 120 GB DRAM-less Crucial BX500 because it was dirt cheap, the machine wasn’t used much at the time, it has several drive bays, and it’s dead easy to install drives in the Mac Pro.

But when it comes down to it, the DRAM-less is just fine for light business type usage and surfing. The Samsung EVO with DRAM is sometimes faster, but the Crucial BX500 without DRAM is still very decent, and hugely faster than any hard drive.

BTW, one last thing. 60% capacity utilization is fine but I get nervous when the drive is more than about 80% or so full. If there isn’t much space left, this can affect performance.