You sir owe me a new keyboard
I seem to have drooled into mine !
Those guys in white coats at the Serial ATA International Organization (SATA-IO) must have been really proud when they unleashed the SATA 6Gb/s specifications on the world. You can imagine them thinking: this will keep us ahead of the game when it comes to hard drive performance. Certainly they were right when it comes to the …
Dear Sir Runcible
It is worse than that. The first page link to the Kingspec review is dead; link rot I suppose. Wouldn't be bothered, but sometimes older Kingspec parts can be found cheaply. I might not like the means by which older Kingspec parts come to be available cheaply, though...
A tadge? Is that bigger than a tad?
The PCMark8 graph is a graph of 'GB/$' against something. Now, I'll admit, at first, to misreading that as 'GB/s', which is an easy mistake to make, and could have easily been clarified with a bit of text, but the other axis has nothing on it to hint what it might be. You can get between 2500 and 5000 of them, or maybe pay between 2500 and 5000 of them, if that's any help (but I probably can't afford them, whatever they are, unless they are Drachmas, and then I'll take all you've got, based on the contents of a wallet left over from a holiday some time ago, and which is of no other use).
"The theoretical bandwidth speed..." What is the 'bandwidth speed'? Is it similar to 'bandwidth'?
Then, on page 5, there is the CrystalDiskMark 3 benchmark. Well, there probably isn't, 'it' is probably 'they'; there are two bar graphs, the first is 'Sequential Read' and 'Sequential Write', the second is 'Default 4k Read' and 'Default 4k Write'. Now, if these results are all from a CrystalDiskMark benchmark suite, then you'd go for the plural, and the title really is only attached to the upper set, where it should be attached to both. If it isn't, then it is a bit of a mystery what the lower one is from, and what the 'Default' is about. Is it the default of some testing suite that is never mentioned? Is the assertion that somehow these drives default to reading and writing in 4k blocks (which seems unlikely, but would have been quite plausible with spinning rustware, and even some SSDs have a sort of 'emulation mode' which would do this, but on these (roughly) enterprise drives? I'd be surprised, but that has happened before. And why is it helpful for this test be in 'old, not very good' drive emulation mode, rather than 'spiffiest, most performant' mode - I mean, there are probably reasons that this mode is relevant to somebody, but what are they and who is it)?
But, the real shame here is that the Author had the chance to make it easier for readers trying to work out which interface is 'better' (more suitable), by including, eg, 'PCIe NVMe' on the labels on the graphs to make it clear. Now, the information is available a bit further down that page, but some won't have noticed that (yet) and will be flipping back to the first page to find out which drive is on which interface.
All-in-all, there is a good article in here, but the presentation prevents it from being better.
Wrong units there. SATA III is 6Gbps (Gigabits, not bytes) which yields somewhere less than 600 MBps when serial bits are deserialized. Now my OCZ RevoDrive 3 has no problem delivering 1.2 GBps (GigaBytes per second) on reads which is why it's in my bloody workstation! Short of renting something better on AWS or investing thousands in a machine that will rarely be used (32 GB 2400 MHz RAM is usually enough here), the OCZ easily loads the hybrid SLI (Quadro, Tesla) at speed and delivers a solid TFLOP of double precision compute. Not TOP-500, but damn! I still remember the Cray Y-MP.
Agreed - a one tenth of a second difference in the time to load a game is not worth hundreds of pounds.
The only point where the price premium may be worthwhile is on a database server as local storage.
(For storage accessed over a network, 6Gbps SATA is more than fast enough as a pair of such drives can saturate a 10Gbps network link.)
I got 2 180GB SSDs in my system, along with 3 2TB HDDs.
The HDDs are used for data storage, the SSDs are used for the Windows OS - one to boot on, the other with the pagefile and one game that takes ages to load.
Ever since I installed them, I experience very fast boot-up time (once a day, so don't really care) and nice general performance from my Windows system.
That one game also benefits immensely from being housed on an SSD since map switching times are divided by ten compared to what they used to be.
That kind of performance is worth every penny to me.
I don't think that will ever quite happen simply for the fact that you are always going to desire more permanence in your non-volatile memory while it's usually more desirable to allow faster (less permanence) in your state memory (RAM). That's why I don't see parity going away for any of the RAM equivalents while yon cosmic ray is going to have a hard time altering the equivalent of flash, PCM, magnetic, spintronic, ... your intentionally more permanent memory.
It wouldn't be that hard now with 3D everything to stack your RAM-equivalent in the processor module (we've been here before!) while it sits in a literal Lego-set of permanent memory around it (done that too!), whatever we end up with at that cycle of technology.
Simon, you consistently mangle PCIe width and PCIe speed. Obviously, anytime anyone says "x2" or "x4", etc. they are referring to width. For example, the M25 may have four lanes in each direction, but the speed is still 5 mph... Yes, I know you managed a nod to (actual) PCIe speed towards the end (Gen1, etc) but at the end of the day the average Joe won't care if you have a Gen2 x2 or a Gen1 x4.
The important question to ask when buying flash-based storage is what happens when the spare area is used up.
Intel rather obnoxiously brick the drive ensuring that data recovery is impossible. Why they wouldn't fail to read-only I don't know. At least that way you could recover any data off the thing.
I think this is one of those questions that has enormous potential for 'YMMV'.
Before saying this you have to accept that I may well have had a very unusual experience, but I am still running an OCZ Petrol 64GB and OCZ Vertex 4 256GB, with no problems (firmware always updated). Do I trust them? That's another question, given the terrible stuff I've read about them. The small one, not bothered, it's in an old duffer laptop that is very much a secondary use machine. The bigger Vertex? Hmmm. I carried it forward in a new machine I built last year and actually, now it's over three years old and my daily runner, I've made the choice to replace with a Samsung 850 EVO.
I've installed many of the EVOs in engineering scenarios and they have been very well received (granted, the 840 have some slowdown issues allegedly addressed with updated firmware and a 'restoration tool'). However, the Samsungs have easily been the quickest and best supported I've encountered. I've also installed quite a few Crucial MX series. Solid enough, bar some irritating firmware bugs played down by Crucial IMHO (MU01, fixed in MU02) and piss-poor support with firmware updates if you're using a Mac.
The only one that's gone squiffy on me was an mSATA - can't remember the make now, but could have been a Samsung PM851 or Lite-On. Eh, sorry. Can't recall now.
All in, I'd say go for it - I haven't seen a higher number of equivalent failures compared to magnetic (anecdotal) and the benefits far outweigh the worries.
Biting the hand that feeds IT © 1998–2019