Why are smaller capacity SSDs slower?

I was reading this Tom's Hardware article on testing SSDs and came across the following claim:

With SSDs, performance varies by capacity point. Smaller drives tend to be slower than larger ones, even in the same family.

However, the article doesn't backup this claim or explain why.

It doesn't seem intuitive to me that smaller drives would be slower. I would think it would be the other way around since a larger drive has a wider "area" to access via the same bandwidth.

So is it true that smaller (capacity) SSDs are slower? And if so, why?

PGmath

Posted 2016-04-02T16:26:55.120

Reputation: 1 430

9'Smaller' might just be 'older design' – Aganju – 2016-04-02T16:44:37.877

7By "small", do you mean in physical size or in capacity? – JesseTG – 2016-04-03T04:04:57.110

1@JesseTG Capacity. I'll make that clearer. – PGmath – 2016-04-03T04:05:23.683

2The crux of the issue is that you're reasoning about throughput using a latency argument. – user541686 – 2016-04-03T08:37:38.543

Answers

107

Larger SSDs from the same SSD model/family (comparing different models makes no sense) are faster when writing data because they use more Channels in parallel:

while smaller ones only use few channel (4 instead of 8) because they have less NAND modules on it.

magicandre1981

Posted 2016-04-02T16:26:55.120

Reputation: 86 560

13I'd add "for a given family/controller" if I wanted to be pedantic. – Journeyman Geek – 2016-04-03T07:50:08.870

8Worth pointing out that the reason for this is that flash chips only tend to be available at a reasonable price in one or two capacities at a time. For instance, currently MLC chips are available in 32, 64 and 128 gigabit varieties, and as the 64 gigabit chips only cost slightly more than 32 gigabit ones but less than half the 128s, anyone designing a new SSD range today will probably design it based on 64Gb chips, regardless of the capacity they're aiming for. – Jules – 2016-04-04T12:40:18.237

Now that's a good simple answer! – Raystafarian – 2016-04-07T08:24:57.297

@magicandre1981 Would you be willing to write a bit more about this as an article for the SU Blog? – rahuldottech – 2016-04-10T17:58:45.087

@RahulBasu not really. – magicandre1981 – 2016-04-10T18:07:08.777

Also note that once all channels are filled, adding more nand dies generally does NOT speed things up more, at least not nearly as much. In your example, all of bank 2 could go away without affecting speed much, because every channel is still fully utilized, but once you remove any more dies you start losing parallelism and therefore speed. – Nate S. – 2018-08-16T00:01:12.037

The higher capacity variants of an SSD model often get their higher capacity from simply having more NAND flash chips of the same type as the lower capacity variants.

Having more flash chips allows for a design where the controller on the SSD can access more data in parallel, allowing for higher speeds.

Håkan Lindqvist

Posted 2016-04-02T16:26:55.120

Reputation: 916

39Sort of like a built-in RAID-0. – kasperd – 2016-04-03T09:36:28.813

Let's pretend that SSD is using 32GB NAND chips for storage, so an 128G SSD need 4 of those chips to add up the space, and an 256G need 8 of the identical chips to make up the space, which means for read/write, on the 256G SSD it can R/W to 8 chips together rather than 128G's 4 chips, which make it much faster.

But since there are other bottle necks, the final speed won't show as double on both R/W, however in real life, there are a lot of examples of how 128G's only have half of the write speed of the 256G's(i.e retina Macbook Pro's PCIe SSD benchmark, Sandisk Extreme CZ80 USB Drive).

The description was not 100% accurate, but it gives the basic idea of why small SSD are slower.

John Li

Posted 2016-04-02T16:26:55.120

Reputation: 51