5

Time ago here was a question about speed comparison between SSD and RAM. So, RAM was much faster. It's ok.

But Intel announced huge speed update with Optane SSD's. So, and probably we should try to use it with RAID0 option.

What is faster now? Is it time to forget about RAM or not? May SSD swap is lightning fast and enough?..

Improve this question

3 Answers 3

Reset to default
2

The answer is pretty simple, RAM is always going to be faster because its first in the chain.

Whenever something needs to be stored temporarily, it first is going to be stored in RAM. If this is not possible, and virtual ram has to be used, a SWAP file will be used, which can be on your SSD. Windows will try to keep using the RAM though, so it will swap background stuff out of RAM to the swap file. If you want to use one of those programs, it is first being swapped back into RAM before accessible again.

Now, when something never leaves RAM its always faster. The speeds may be marginal, but its faster nevertheless.

If you have slow RAM and super fast SSD that outperforms the RAM by a million, it will simply appear that the SSD is not that fast, because its still waiting for the RAM. Note that slow ram is still very fast and it is unlikely to notice. But really fast RAM which is an option too, is still by far the best choice, and the more you need, the more RAM you should get.

Improve this answer
6
  • Thanks. But look, I posted a mention not about typical SSD, but Optane with 3D XPoint, I mean this one: wccftech.com/… I understand what you mean, but my question is about the more practical use. And I see not enough info to compare Optane SSD with modern RAM. Ok, RAM is faster by default, but isn't speed the same or very very close each other?..
    – Arseniy Kamyshev
    Jun 5, 2016 at 5:04
  • My point is, that an SSD is ALWAYS going to be 2nd in the chain. You ask if this SSD could essentially replace the RAM, and it cannot.
    – LPChip
    Jun 5, 2016 at 10:11
  • If any of these answers helped you, feel free to mark one of them as answer so we know you aren't looking for help anymore. You can do this by pressing the grey checkmark on the left of an answer. It'll turn green, and it rewards the poster. Also, if more answers are helpful, give them a thumbs up. :)
    – LPChip
    Jun 6, 2016 at 11:54
  • 3
    For current technologies, SSD will be 2nd in the chain. However, I believe Intel has stated that it will also be releasing 3D XPoint DIMMs (not sure if it's still branded as Optane), so they are apparently comparable enough that Intel thinks this is a legitimate use case.
    – rationalis
    Jun 28, 2016 at 20:50
  • 3
    @LPChip: I guess he is talking about abandoning disk-ram duality someday. And no, SSD is not ALWAYS going to be 2nd in the chain. Avoid making future predictions.
    – Rok Kralj
    Jul 12, 2016 at 0:43
15

1 s(second) = 1000 ms(millisecond)
1 ms(millisecond) = 1000 μs(microsecond)
1 μs(microsecond) = 1000 ns(nanosecond)

  • DRAM: 200 ns
  • 3D Xpoint: 7 μs (7000 ns)
  • enterprise NVMe PCIe SSD: 30 μs (30,000 ns)
  • consumer NVMe PCIe SSD: 110 μs (110,000 ns)
  • Hard disk seek: 10 ms or 10,000 μs or 10,000,000 ns

DRAM is still the fastest !

Source: https://www.theregister.co.uk/2016/04/21/storage_approaches_memory_speed_with_xpoint_and_storageclass_memory/

Improve this answer
2
  • Access times don't tell the whole story of course - transfer speed is as important. However, you only have to take a look at the DRAM throughput compared to the NVMe throughput (usually x2 or x4 for M.2 drives) to see that DRAM still easily beats XPoint / Optane. E.g. 12,800 MB/s for dual channel DRAM vs 2500 MB/s for a high end consumer level Optane drive. It comes closer if you use one DRAM DIMM / channel though.
    – Maarten Bodewes
    May 6, 2019 at 14:09
  • Hmm, seems like I used PC-800 memory bandwidth for that, we're currently at PC-2400 ~ raw bandwidth of memory means ~30,000 MB/s for PC-2400. Sorry, could not edit anymore. Raw bandwidth values were amazingly hard to find.
    – Maarten Bodewes
    May 6, 2019 at 14:26
1

As a general rule: CPU cache is faster than video card RAM which is faster than system RAM which is faster than SSDs which (ragarding access time) are faster than HDDs.

As a direct answer, RAM is hands down faster. A typical SSD has a random access time of ~100 micros (seek time between 30 - 160 us.), while typical RAM has access times of a few ns.

Improve this answer
2
  • So, yes, thank you. Don't you know how the chain of latencies looks in the typical platform now in 2016? I mean CPU -> (which latency?) -> video card RAM -> (which latency?) -> system RAM -> (which latency?) -> SSD Optane -> latency?..
    – Arseniy Kamyshev
    Jun 5, 2016 at 5:08
  • You can find relatively easy information on all of that. On memory: en.wikipedia.org/wiki/GDDR5_SDRAM ; on cache: extremetech.com/extreme/… . SSDs got to as low as 30us. But don't be miss-lead, for the general performance there are other factors that matter too.
    – Overmind
    Jun 6, 2016 at 8:30

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .