Cheap VS expensive SSD

[torlen11cc]

Hey,
is it true that the average user has no reason to purchase an expensive SSD drive? Because in actuality, if the drive isn't used intensively, there is no meaningful difference between Samsung EVO 850 and Kingston UV400 SSD (for example).
I mean that the drive is used purely for storing programs such as Skype, Chrome, Steam etc., for games such as GTA V and of course the operating system - which is basically how it is used amongst almost all gamers.
What do you think?

 

[WhoBeDaPlaya]

Yes

 

[dave_the_nerd]

I'd tend to agree.

Unless you know you're bottlenecked by storage performance, spending 2x the $/GB for "fast" SSDs isn't very sensible.

There have been exceptions - extremely low-end, small SSDs where performance was outright mediocre, even compared to spinning disks. But you'll have an awfully hard time finding one of those still for sale.

 

[deustroop]

Hey,
is it true that the average user has no reason to purchase an expensive SSD drive? Because in actuality, if the drive isn't used intensively, there is no meaningful difference between Samsung EVO 850 and Kingston UV400 SSD (for example).
I mean that the drive is used purely for storing programs such as Skype, Chrome, Steam etc., for games such as GTA V and of course the operating system - which is basically how it is used amongst almost all gamers.
What do you think?

No

The whole point of a speedier drive is to use it specifically for " games such as GTA V and of course the operating system. " Otherwise its f*** wasted.

 

[dave_the_nerd]

No

The whole point of a speedier drive is to use it specifically for " games such as GTA V and of course the operating system. " Otherwise its f*** wasted.

Right. That's a given. The question is re: the performance difference between budget and pro level SSDs, and whether or not it's worth the money. (Of course the more expensive drives are faster...)

Basically, the more your workload relies on sequential data transfers, the less important a higher-end SSD is. Office use and multitasking workloads rely more on random small I/O, whereas gaming - if it's limited by storage at all - tends to be doing big sequential transfers (loading levels and maps and stuff.)

Nobody's suggesting that we go back to HDDs. But even a UV400 will do sequential reads at ~450MB/s.

 

[mikeymikec]

I disagree with the general opinion on this thread. In the era of HDDs or SSDs, I'd always be happy to pay more for better quality, because a dying storage device is a PITA. From what I understand, dying SSDs tend to be more sudden death types than slow deaths, which makes it even more annoying*. More expensive devices (relative to an amount of storage) tend to have longer warranties, ergo they're probably more reliable.

A customer of mine went from a Samsung 840 (non EVO/PRO) 128GB to a 850 PRO 256GB and said that he thought his system's performance improved as a result. "Feels" is often a poor benchmark however

* - as always, backups are essential, but system downtime is still system downtime, backups or not.

I may be a bit different to many on this forum insofar as I build PCs that have often been used until the customer wanted a replacement 8-12 years later (for performance reasons), and what may be considered to be a perfectly performing but lower-end SSD today may be something entirely different in that kind of timespan.

 

[WhoBeDaPlaya]

A customer of mine went from a Samsung 840 (non EVO/PRO) 128GB to a 850 PRO 256GB and said that he thought his system's performance improved as a result. "Feels" is often a poor benchmark however

Now, now, please don't discount the digital equivalent of the "butt dyno"

 

[dave_the_nerd]

I disagree with the general opinion on this thread. In the era of HDDs or SSDs, I'd always be happy to pay more for better quality, because a dying storage device is a PITA. From what I understand, dying SSDs tend to be more sudden death types than slow deaths, which makes it even more annoying*. More expensive devices (relative to an amount of storage) tend to have longer warranties, ergo they're probably more reliable.

A customer of mine went from a Samsung 840 (non EVO/PRO) 128GB to a 850 PRO 256GB and said that he thought his system's performance improved as a result. "Feels" is often a poor benchmark however

* - as always, backups are essential, but system downtime is still system downtime, backups or not.

I may be a bit different to many on this forum insofar as I build PCs that have often been used until the customer wanted a replacement 8-12 years later (for performance reasons), and what may be considered to be a perfectly performing but lower-end SSD today may be something entirely different in that kind of timespan.

Is there any evidence to indicate that budget SSDs are less reliable than higher-end models? (Besides the rated number of write cycles and manufacturer warranties?) Like, actual failure or return rates?

There was a french retailer that published some of that a few years ago, during the OCZ/Sandforce debacle. But I haven't seen anything recent. *shrug*

 

[WhoBeDaPlaya]

Oddly enough, the only SSDs I've had die on me were 80GB Intel X-25Ms.
They'd slow down to a crawl, and finally wouldn't register during POST.
All three of them :|

Replaced them with 120GB 3.5" Vertex 2s (original ver), and they're still running fine to this day.

 

[torlen11cc]

I should mention that I currently use the Kingston V300 SSD drive, and it is one of the cheapest SSD drive models. It performs well: the operating system and applications that I use load within seconds, so I'm quite content. The question is what will a more expensive SSD drive provide me with? faster loading times? I understand that the differences are marginal, only several seconds or millisecond, so the value of a more expensive SSD drive seems low compared to the money invested in it.

 

[VirtualLarry]

The effective ("seat of the pants") performance, of the 2nd-gen SandForce controller in the Kingston V300, among others (Mushkin ECO2 as well, and Kingston HyperX Fury), in Windows, for an OS / apps drive, is actually quite good. Regardless of mediocre benchmarks with CDM (probably using incompressable data).

At least, that has been my impression, from using some V300 120GB and 240GB drives, as well as VisionTek GoDrive 120GB drives.

The truth is, that "one little trick" of the 2nd-Gen SandForce controller (compression) does wonders for Windows' OS files and programs, and smaller data files. Maybe not so much for extensive previously-compressed media files.

I've also seen great benchmarks from the UV300 drives, which I think use a different controller. Over 300MB/sec random 4K QD32 read scores, which are really impressive. (Think, PCI-E level random I/Os.)

Too bad that the UV300 drives were so hard to find here in the US. (Got some on ebay.)

 

[Billy Tallis]

There have been exceptions - extremely low-end, small SSDs where performance was outright mediocre, even compared to spinning disks. But you'll have an awfully hard time finding one of those still for sale.

The threshold for "extremely small" is moving up. IMFT's 32-layer 3D TLC is 384Gb (48GB) per die, and this year everybody will be moving to 64+ layer 3D TLC with 512Gb (64GB) per die. At least some manufacturers will also be producing 64+ layer parts with smaller capacities for the mobile market, but I'm not sure if they'll be the cheapest option for small low-end SSDs. Either way, 256GB-class SSDs are increasingly suffering the same limitations that 128GB-class drives had two years ago (though per-die performance has increased some), and 128GB SSDs really don't make much sense any more. (Especially ones like the Intel 600p, where the expensive NVMe controller ends up with an unpopulated channel.)

 

[VirtualLarry]

How does that play out, for example, with the 128GB Adata SU800 Ultimate 2.5" SATA6G SSD?

As I understand it, that drive uses the IMFT 3D TLC NAND. If those come in 48GB die sizes, then how does that fit into 128GB? Loads of spare area, after using 3 dies? Will we start seeing "non-traditional" SSD sizes, like 144GB at the low end? Does the Adata drive in question, actually have 144GiB of NAND in it?

 

[Billy Tallis]

The 128GB SU800 does have only three NAND chips, for a nominal raw capacity of 144GB. Only three of the four channels from the SM2258 controller are used, but having so much spare area does help offset some of the bottlenecks that creates.

The Crucial MX300 uses the same TLC, and comes in capacities of 275GB, 525GB, 750GB, 1050GB and 2050GB. Of those, only the 750GB had a "normal" amount of overprovisioning.

 

[cbn]

The threshold for "extremely small" is moving up. IMFT's 32-layer 3D TLC is 384Gb (48GB) per die, and this year everybody will be moving to 64+ layer 3D TLC with 512Gb (64GB) per die. At least some manufacturers will also be producing 64+ layer parts with smaller capacities for the mobile market, but I'm not sure if they'll be the cheapest option for small low-end SSDs.

http://www.anandtech.com/show/11100...ce-roadmap-updates-forecasts-and-ceo-retiring

With the second generation 3D NAND, Micron is shifting their strategy slightly by offering at least two different die sizes. We've previously heard about the 512Gb 64-layer 3D TLC part, but Micron will also be making a smaller 256Gb 3D TLC part. This die is planned to be the smallest 256Gb NAND flash die available from any vendor, at 59 mm^2 or 4.3Gb/mm^2. The smaller die is intended for the mobile market where the 512Gb part will be physically too large.

Parrallelism would be the same on those 64 layer 512Gb parts compared to the small die 64 layer 256Gb part?

If it is, then I think it could pretty interesting to see how eMMC and UFS 2.0 would compare to budget low capacity SSDs that would use the larger dies.

 

[cbn]

How does that play out, for example, with the 128GB Adata SU800 Ultimate 2.5" SATA6G SSD?

As I understand it, that drive uses the IMFT 3D TLC NAND. If those come in 48GB die sizes, then how does that fit into 128GB? Loads of spare area, after using 3 dies? Will we start seeing "non-traditional" SSD sizes, like 144GB at the low end? Does the Adata drive in question, actually have 144GiB of NAND in it?

The 128GB SU800 does have only three NAND chips, for a nominal raw capacity of 144GB. Only three of the four channels from the SM2258 controller are used, but having so much spare area does help offset some of the bottlenecks that creates.

The Crucial MX300 uses the same TLC, and comes in capacities of 275GB, 525GB, 750GB, 1050GB and 2050GB. Of those, only the 750GB had a "normal" amount of overprovisioning.

I'm surprised we haven't seen any 180GB/192GB SSD that use four of those 48GB dies?

 

[dave_the_nerd]

I'm surprised we haven't seen any 180GB/192GB SSD that use four of those 48GB dies?

Are most people buying SSDs that small? I mean, with the proliferation of autoupdate and Windows.old directories everywhere (and hell, my Windows directory by itself is 56GB) is < 240GB really usable for "typical users" who aren't actively monitoring their disk use?

Sales numbers for a given capacity would be interesting. Probably nobody publishes that though.

 

[cbn]

Are most people buying SSDs that small?

Well, there must be some demand because the ADATA SU800 comes in 128GB capacity.

And that one uses 3 384Gbit (48GB) dies.... so that is why I am wondering why not make one with 4 384Gbit (48GB) dies for 180GB or 192GB capacity?

For that matter why not also have a 360GB or 384GB? This would use eight 384Gbit (48GB) dies.

Quad channel SSD controller with an even number of NAND dies.

P.S. My guess on why we haven't seen this happen and instead we see 128GB and 256GB (275GB for MX300) based on an odd number of dies is because the lower capacities are more popular than the higher capacities. Thus 128GB (three dies) and 256GB (six dies) wins over 180GB/192GB (four dies) and 360GB/384GB (eight dies).

 

[cbn]

Looking ahead to the future of 120GB/128GB, I think that could still exist...... even with the large die second generation 3D NAND (ie, 64 layer, 512Gbit/64GB)..... via dual channel controllers like the Phison S11 (DRAM-less SATA 6 Gbps), Marvell 88NV1120 (DRAM-less SATA 6 Gbps) and Marvell 88NV1140 (DRAM-less NVMe PCIe 3.0 x1).

 

[mikeymikec]

Is there any evidence to indicate that budget SSDs are less reliable than higher-end models? (Besides the rated number of write cycles and manufacturer warranties?) Like, actual failure or return rates?

There was a french retailer that published some of that a few years ago, during the OCZ/Sandforce debacle. But I haven't seen anything recent. *shrug*

Not that I'm aware of.

 

[Hellhammer]

I would say this topic is similar to buying a known-brand GPU vs buying a no-name one. The hardware is the same because many of these small brands use contract manufacturers anyway, making them practically marketing and logistics companies. The real difference is in support (and possibly add-on software) because the smaller companies don't usually have the necessary infrastructure to provide proper customer support and RMA in case of any problems. For most drives sold this won't make any difference, but on the other hand if the drive is going to a system that is used on a daily basis it might be worthwhile to pay a little extra for the peace of mind.

 

[escrow4]

The speeds suck badly. Especially with 4K/IO speeds that gut any SSD even top shelf ones. Sandisk, Kingston and other cheap rubbish should be avoided. You will tell if you hammer the drive daily even with semi serious multitasking and writing at the same time.

 

[deustroop]

The speeds suck badly. Especially with 4K/IO speeds that gut any SSD even top shelf ones. Sandisk, Kingston and other cheap rubbish should be avoided. You will tell if you hammer the drive daily even with semi serious multitasking and writing at the same time.

If its faster, its worth extra money. PERFORMANCE is EVERYTHING. If you go from 850 EVO to 960 Pro you can notice a quickness which makes it worth the cost.

 

[Valantar]

I would say this topic is similar to buying a known-brand GPU vs buying a no-name one. The hardware is the same because many of these small brands use contract manufacturers anyway, making them practically marketing and logistics companies. The real difference is in support (and possibly add-on software) because the smaller companies don't usually have the necessary infrastructure to provide proper customer support and RMA in case of any problems. For most drives sold this won't make any difference, but on the other hand if the drive is going to a system that is used on a daily basis it might be worthwhile to pay a little extra for the peace of mind.

That's probably true if you go for some cheapo no-name chinese SSD brand, but not common ones like OCZ/Toshiba, Crucial or SanDisk.

The speeds suck badly. Especially with 4K/IO speeds that gut any SSD even top shelf ones. Sandisk, Kingston and other cheap rubbish should be avoided. You will tell if you hammer the drive daily even with semi serious multitasking and writing at the same time.

The thing is, 99% of PC users dont "hammer the drive daily even with semi serious multitasking and writing at the same time." 95% of users don't even come close to that.

If its faster, its worth extra money. PERFORMANCE is EVERYTHING. If you go from 850 EVO to 960 Pro you can notice a quickness which makes it worth the cost.

Do you drive a Formula 1 car when you go shopping? Do you shower with a power washer? Shave with a belt sander? Performance is relative to the task to be performed. I'm not saying that too much performance is a bad thing in a PC (unlike my rather silly examples), but mostly you'll be throwing money out the window, essentially paying for a placebo effect.

 

[deustroop]

[QUOTE

I'm not saying that too much performance is a bad thing in a PC (unlike my rather silly examples), but mostly you'll be throwing money out the window, essentially paying for a placebo effect.[/QUOTE]

No.