The State Of Solid State: Computex 2015

[2Be_or_Not2Be]

I have a test rig with two Samsung Pro SSDs, two 7200 rpm SSDs and one 7200 rpm HD. Here's the Windows Boot Time Results.

SSD - 15.6 seconds
SSD RAID 0 - 15.9 seconds
SSHD - 16.5 seconds
HD - 21.2 seconds

I don't know why your SSD-equipped system boots so slowly. Maybe you're timing it from when the system is still running its POST. From the moment that my Win7 system POSTs and the screen displays the "Starting Windows" splash screen, I have a Windows desktop in less than 3 seconds. In fact, the splash screen doesn't even finishing "building" the logo before it goes to the Windows logon screen.

That's on a Samsung 830 SSD - not even one of the newer ones.

 

[synphul]

Yes they do, according to benchmarks. In real world every day use for most tasks there is little noticeable difference is all I'm saying.

Well yes, if all you do is browse the web then you probably wouldn't notice any difference other than boot time and application load times. However, for power users there is a huge difference between mechanical and solid state, and that isn't even accounting for the significantly higher failure rate of a hard drive.

Playing digital media off a hard drive that can't be loaded completely into memory? You can forget about doing anything else on that drive. Regular torrent user? Again, the disk reads and writes makes doing anything else on the drive at the same time virtually impossible.

For anyone who does multitasking at even the minimal level a SSD is a significant time saver over a mechanical drive. But yes, you are correct, I wouldn't install a SSD for my parents who browse the web, use Word, and open the occasional PDF.

Which is why I said it would benefit those who are doing heavy i/o intensive operations - which torrent + playing digital media etc simultaneously is. There are more things being done on the pc than torrents or simple word docs, that leaves out a whole lot in between. You might call content creation a power user scenario, but photoshop/illustrator and cad machines aren't overly i/o intensive. Just like gaming will eat all the cpu and gpu you can throw at it, but won't max out an hdd. Definitions of power users vary. If you're serious about multitasking, you're likely using a multidrive setup anyway to avoid bottlenecking the bandwidth of a single drive/channel. There's more to i/o optimization than a single ssd drive. Talking about power users, that's why data centers are designed the way they are. They don't pop in a single 850 evo and call it a day.

 

[2Be_or_Not2Be]

Yes they do, according to benchmarks. In real world every day use for most tasks there is little noticeable difference is all I'm saying.

Well yes, if all you do is browse the web then you probably wouldn't notice any difference other than boot time and application load times. However, for power users there is a huge difference between mechanical and solid state, and that isn't even accounting for the significantly higher failure rate of a hard drive.

Playing digital media off a hard drive that can't be loaded completely into memory? You can forget about doing anything else on that drive. Regular torrent user? Again, the disk reads and writes makes doing anything else on the drive at the same time virtually impossible.

For anyone who does multitasking at even the minimal level a SSD is a significant time saver over a mechanical drive. But yes, you are correct, I wouldn't install a SSD for my parents who browse the web, use Word, and open the occasional PDF.

Which is why I said it would benefit those who are doing heavy i/o intensive operations - which torrent + playing digital media etc simultaneously is. There are more things being done on the pc than torrents or simple word docs, that leaves out a whole lot in between. You might call content creation a power user scenario, but photoshop/illustrator and cad machines aren't overly i/o intensive. Just like gaming will eat all the cpu and gpu you can throw at it, but won't max out an hdd. Definitions of power users vary. If you're serious about multitasking, you're likely using a multidrive setup anyway to avoid bottlenecking the bandwidth of a single drive/channel. There's more to i/o optimization than a single ssd drive. Talking about power users, that's why data centers are designed the way they are. They don't pop in a single 850 evo and call it a day.

I'm not sure if you're arguing for SSDs or against SSDs, or you actually trying to define what a "power-user" might or might not be. So let me comment on a few of your statements.

It's a bit simplistic to state that a SSD would "benefit those heavy i/o intensive operations" and then say that "photoshop/illustrator and cad machines aren't overly i/o intensive". You seem to imply that category of users wouldn't really benefit from a SSD. That is definitely wrong - a Photoshop/Illustrator user can have multiple 100MB+ items to place in an image. Opening & saving can all benefit from a SSD, even in a machine with a lot of RAM. CAD/CAM users (AutoCAD, SolidWorks, etc.) can also benefit from SSDs when they have to open/save/render drawings with sometimes thousands of parts (and files).

Basically, there are benefits from SSDs that virtually all professional/commercial users can enjoy. Home users also can see benefits, although somtimes the upgrade cost may not seem worth it to them.

Now if you are arguing that there is more to I/O optimization than just plugging in a single SSD, then I can definitely agree with that.

 

[synphul]

I just meant there is more than one thing defining a 'power user' and that not all are i/o intensive. I use photoshop and illustrator extensively, and while ps loads a tiny bit faster with an ssd it's not mind blowing. The majority of the work isn't opening/saving files, rather content creation - aka graphic design, not purely batch retouching 1000 photos. So 95% of the time is spent drawing, coloring, applying presets, filters and so on and working with well over 100 layers in images around 4200x4200. During the creation stage, very little i/o activity is going on. However for applying effects and filters, a much more powerful machine is required than one simply good enough to edit word docs. In that situation, more ram is far more beneficial to the multitasking of having photoshop, bridge and illustrator open and working in all three, transferring working files between one program to the other.

That's not to say that's how all users use their machines, but it's pointing out 'power user' has different meanings. A content creation rig is still basically a workstation. It's still more effective to have programs and os on one drive, source files on another physical drive and output to yet another physical drive, aka a mulitdrive setup than it is to rely on a single ssd. Same can be said for gamers where a $200-300 cpu and a $350-$650 gpu (or multi gpu) setup also defines them as power users but where ssd's advantages just aren't really seen.

It's not simplistic at all, the advantage of an ssd is i/o intensive scenarios. Photoshop can be i/o intensive, as can cad. Are you trying to say that cad users are spending 5% of the time drawing/plotting and the vast majority is frantically opening/saving files? It's quite the other way around. Not all who use those programs to create content are hitting the storage disk (ssd or hdd) that hard during the creative process. My photoshop usage didn't change pre ssd to after installing an ssd. There's very little difference in loading images and I'm quite aware of how large they get. It didn't take but a split second to load those images off an hdd either so again the gains of the ssd are very slight. Aside from the cpu, the next largest upgrade was ram since system memory was constantly maxed with only 8gb ram. If I would have left the 8gb of ram and only upgraded to the ssd I'm sure the improvement would have seemed more significant since the swap file use would've been faster on an ssd. However that's one way to look at it. Was it really a problem of the hdd being too slow? Or more a side effect of not having enough system memory? With the proper amount of system ram, the system shouldn't be hitting the swap file and need the ssd as a bandaid for a different bottleneck in the system.

 

[2Be_or_Not2Be]

I just meant there is more than one thing defining a 'power user' and that not all are i/o intensive. I use photoshop and illustrator extensively, and while ps loads a tiny bit faster with an ssd it's not mind blowing. The majority of the work isn't opening/saving files, rather content creation - aka graphic design, not purely batch retouching 1000 photos. So 95% of the time is spent drawing, coloring, applying presets, filters and so on and working with well over 100 layers in images around 4200x4200. During the creation stage, very little i/o activity is going on. However for applying effects and filters, a much more powerful machine is required than one simply good enough to edit word docs. In that situation, more ram is far more beneficial to the multitasking of having photoshop, bridge and illustrator open and working in all three, transferring working files between one program to the other.

That's not to say that's how all users use their machines, but it's pointing out 'power user' has different meanings. A content creation rig is still basically a workstation. It's still more effective to have programs and os on one drive, source files on another physical drive and output to yet another physical drive, aka a mulitdrive setup than it is to rely on a single ssd. Same can be said for gamers where a $200-300 cpu and a $350-$650 gpu (or multi gpu) setup also defines them as power users but where ssd's advantages just aren't really seen.

It's not simplistic at all, the advantage of an ssd is i/o intensive scenarios. Photoshop can be i/o intensive, as can cad. Are you trying to say that cad users are spending 5% of the time drawing/plotting and the vast majority is frantically opening/saving files? It's quite the other way around. Not all who use those programs to create content are hitting the storage disk (ssd or hdd) that hard during the creative process. My photoshop usage didn't change pre ssd to after installing an ssd. There's very little difference in loading images and I'm quite aware of how large they get. It didn't take but a split second to load those images off an hdd either so again the gains of the ssd are very slight. Aside from the cpu, the next largest upgrade was ram since system memory was constantly maxed with only 8gb ram. If I would have left the 8gb of ram and only upgraded to the ssd I'm sure the improvement would have seemed more significant since the swap file use would've been faster on an ssd. However that's one way to look at it. Was it really a problem of the hdd being too slow? Or more a side effect of not having enough system memory? With the proper amount of system ram, the system shouldn't be hitting the swap file and need the ssd as a bandaid for a different bottleneck in the system.

I'm still not sure what points you're trying to establish. But if I take your first sentence - "I just meant there is more than one thing defining a 'power user' and that not all are i/o intensive." - then I agree.

If your main point is saying a SSD is only beneficial in I/O-intensive operations, I would disagree because small seconds/minutes saved here and there with a SSD over a HDD definitely add up. Time is money, so any amount of time saved is benefiting you and/or your employer (disregarding what you actually do with it). Even gamers benefit from small (or significant) savings here and there; you can't just say "ssd's advantage just aren't really seen."

All users have times when there are bursts of I/O and times when there are sustained I/O - all those times when you save minutes to possibly even hours with a SSD simply add to the balance in favor of SSDs.

Right now the only reason why you might want a HDD over a SSD is price and/or capacity. In time, those differences will be overcome as well.

Edit: shortened for clarity

 

[CRamseyer]

I don't see Intel bringing a 256GB class SSD 750 to market. The controller and DRAM is too expensive to justify that small of a product. It would cost quite a bit more than competing 256GB class products.

For smaller capacity NVMe drives we will have to wait for the Phison E7, Silicon Motion's controller in 2016 and Marvell's Eldora / Eldora Lite. Seagate should have the new SandForce SF3000 ready before the end of 2015 but we've all heard that line before.

 

[synphul]

I'll just agree to disagree. Everyone has their own opinion as to value of performance. A corvette on the racetrack beats a cobalt any day. On a hwy where both are limited to 65mph, that extra performance is worthless. A gamer will see no fps increase and that's already been well established. My programs already load in milliseconds, I don't need them to load any faster. Going beyond the physical limits of the human body is only worthy for epeen status. Hopefully you don't stop for a bathroom break at any point during the day or the precious nanoseconds saved on non i/o intensive uses will be lost.

They're a real boon to mobile devices that might see more sporadic infrequent use, being powered up repeatedly for the almight boot time savings. Power efficiency and shock resistance help as well. I could probably look on the calendar to see the last time I booted my pc. It may have been 5 or 6 days ago to perform an update. Those 3 seconds shaved off the boot time are really paying in dividends lol. It just depends on the individual, it's not a guaranteed performance upgrade. The only thing I have to wait for on my pc is my internet connection and an ssd won't improve that. That's the slowest part of loading any game anymore, since every time it requires a connection to dink around checking for updates to either uplay or steam.

I wish there were better resources, bench's what have you for real world performance in a variety of tasks to compare ssd and hdd. They seem to be lacking and the most common ones I come across make the ssd look like the holy grail - against a 2.5" 5400rpm mobile hdd. I'd also be curious to find out what hdd's are compared. For instance, someone posted on a forum where linustech pitted an ssd against an hdd. Calculating a folder it took the ssd 15s for 311gb of data. Assuming this was right click and choose properties to calculate file/folder count and usage. Their hdd test took 158s to complete. I did the same test albeit not the same exact file size/count just whatever was on my storage partition - 346gb and it took my hdd 59s to complete. So basically the ssd did the job 4x faster on a task I personally seldom have the need for - one that again is obviously i/o intensive - and the results were 4x as fast not 10-20x as many state. Other benchmarks have shown boot times to be around 5-6min or longer from a cold boot. My system never takes that long on a cold boot. So I think it has to do with the hardware being tested. Maybe there are some janky hdd's out there, idk.

It would be nice to see prices continue to fall. If they became competitive with hdd's for price/gb I wouldn't be against them. Looking at my own systems performance and the difference in performance between my general use/gaming and everything under the sun pc to my workstation (ssd vs hdd) I'm just not that wowed. So far my own personal real world experience is that ssd's look much better on paper. Just like passmark scores, dual channel vs single channel ram bench's and other things that somehow get lost in translation to the real world realistic scenarios.

Maybe it's me, I'm also not one of these people with 3 dvd drives burning 3 movies at once, watching two others, live streaming my 2 simultaneous games I'm playing while downloading 10 tor files. 😛

 

[xyriin]

A gamer will see no fps increase and that's already been well established. My programs already load in milliseconds, I don't need them to load any faster.

Calculating a folder it took the ssd 15s for 311gb of data. Assuming this was right click and choose properties to calculate file/folder count and usage. Their hdd test took 158s to complete. I did the same test albeit not the same exact file size/count just whatever was on my storage partition - 346gb and it took my hdd 59s to complete. So basically the ssd did the job 4x faster on a task I personally seldom have the need for - one that again is obviously i/o intensive - and the results were 4x as fast not 10-20x as many state.

It's true that a SSD won't increase FPS performance, however it makes a HUGE difference everywhere else. Just think about the common install size of a top tier title these days (20GB+). First if you're using digital you have to download all that data. During the download and install afterwards you're not going to be doing much of anything on your computer when it comes to disk access. Even after that long all stop period you have to load this data from your drive to memory at some point, and even while you don't do it all at once a SSD is going to read a 400-500MB chuck of data and load it immensely faster than a hard drive.

As for your properties calculation the huge variable is the number of files and file size. It would make sense that your storage partition is fewer files and/or larger files, not a few hundred thousand tiny files. If that same file sampling had been on your hard drive the results would have been much closer to a 10x value, you're literally comparing an apples test to the oranges on your drive. However the principles remain. Let's say you aren't doing properties but you have to regularly sift through thousands of photos, a SSD would kill in this kind of rapid IOPS function. As a sample I went to my system drive on my work computer and did a properties check on everything in C:\ ... 44 seconds for a mere 54.2 GB (188,210 files). By that timeline if I had the same type of files but 311 GB it would have taken 252 seconds or over FOUR minutes! Obviously in that scenario I'm at a 17x increase compared to the SSD test data. I believe the example you cited because they did the test on the SAME data, you have to compare apples to apples.

For me the biggest personal issue is media watching. I don't have enough SSD storage space to be entirely off hard drives, so a lot of the movie and TV series on my Plex server are still on mechanical disks. If I'm just watching a single thing it's no big deal. However, the second someone else pulls up a movie on the drive I run into issues. Need to add media to the drive while watching something? Forget about it. These issues simply don't exist with a SSD as I've been able to slowly move some of the media to 1TB SSDs. I mostly mitigated the disk load torrent issue a few years ago by using a dedicated hard drive for torrenting. However, it wasn't solved until I swapped my torrent drive to an older small SSD...I don't even remember what a disk overload issue is anymore on my torrent drive. It's also pretty cool to be able to play a game without issue while one is downloading and/or installing in the background.

For me it all comes down to multitasking...
- Stop watching a movie to add a movie to my media server - time wasted
- Stop playing a game to download, install, or update a game - time wasted
- System slowdown while Dropbox, Drive, or OneDrive syncs - time wasted
In fact any time you're reading/writing to hard drive by more than one source you're waiting and I just find that unacceptable in today's world of NAND storage.

Obviously my use case isn't everyone's use case and while a SSD will ALWAYS be faster some people won't justify the cost increase because they're fine waiting. The classic YMMV tag applies well here. However my main point is that the technology is vastly superior. Hard drives have quite literally peaked. The only appeal these days is capacity and they're already pulling stunts like using Helium (8TB) and shingled tech (10TB). Keep in mind that shingle tech (SMR) makes hard drives even slower which is a serious drawback.

 

[akula2]

The time is ripe to change your old spinning catastrophe and embrace the now of SSD.

Old spinning catastrophe? Seriously?
I use over 100 SSDs and also hundreds of HDDs in a variety of systems. Performance alone can't be a deciding factor, say how about the cost factor like in this simple ZFS SAS setup?

2 x {6 x 6TB} = 72TB theoretical.

Now, how would you propose to achieve the same using SSDs?

 

[xyriin]

The time is ripe to change your old spinning catastrophe and embrace the now of SSD.

Old spinning catastrophe? Seriously?
I use over 100 SSDs and also hundreds of HDDs in a variety of systems. Performance alone can't be a deciding factor, say how about the cost factor like in this simple ZFS SAS setup?

2 x {6 x 6TB} = 72TB theoretical.

Now, how would you propose to achieve the same using SSDs?

You are right in cost being a factor. While it's certainly cheaper to purchase hard drives there are other cost factors. If you ignore capacity and cherry pick the best value per GB you can get hard drives around $0.03 per GB (using low end model pricing, not enterprise level drives). SSDs have a much higher entry price around $0.30 per GB.

So for initial setup the simple choice looks like hard drives. However, when you start factoring in that hard drives use significantly more power you run into the first SSD cost saver. A hard drive will typically use about 6.5W under load and 4.5W idle. By comparison a SSD is normally around 1.5W load and 0.5W idle. Rough calculations say a hard drive uses 4 times more power under load and 9 times more power idle. However the load calculation is deceptive. Since the hard drive is slower it's going to be under load longer than a SSD. In fact, even for 100% sequential transfers a SSD is going to be at least twice as fast a even the fastest hard drives. This means you're using at least 8 times more power under load and 9 times more power idle. If you account for the real world and random access it's pretty safe to say your operational power costs for hard drives are about 10 times that of SSDs.

Funny how that works out, SSDs cost around 10x more to purchase but cost about 10x less to operate. Further skewing the initial purchase price numbers is the fact that hard drives fail far more often. So let's say the SSD only lasts 6-10 years, and the hard drive 3-5 years under regular use. You effectively just doubled or tripled your initial purchase cost of hard drives because you have to buy them more often. So even if you only double the purchase cost, SSDs now only cost 5 times more than a hard drive, but still operate at 10x less than hard drives.

Use case obviously is going to play a large factor in operation costs but even huge data centers that need massive amounts of storage use SSDs and they try to prioritize their mediums. They try to use hard drives for cold storage applications and SSDs for anything that is frequently accessed. This allows them to take advantage of the operational cost savings of SSDs, place the majority of wear and tear on the more reliable SSD instead of failure prone hard drive, and of course place infrequently accessed data on large slow volumes.

 

[akula2]

My storage support is broken down into:

a) Enterprise (Linux Servers, NAS etc)
b) High-end workstations (R&D, production etc)
c) Desktops (executive/office)
d) Mobility: Laptops and Tablets

Currently, I've 500+ machines in total as mentioned above. In this type of heterogeneous setup, only viable solution is mixing SSDs and HDDs. It means hundreds of storage devices. In some solutions, SSDs almost eliminated. E.g., NAS server based on ZFS (baring ZIL etc). Nevertheless, I've adopted SSDs whenever it's efficient and highly production say in workstations or office machines. However when it comes to Network storage, HDDs rule for now (volume cost factor).

 

[akula2]

I forgot to add:

Power isn't a factor at all. My facilities are 100% solar in three nations.

 

[JackNaylorPE]

People used to say that back when I was really into RAID for my OS. Then they would wonder how I was already loaded in game, destroyed the enemy jets and reloaded back on the carrier before anyone else was even fully loaded in game.

The thing about RAID is you have to work to get the most out of it. There are several ways to tune the array but one wrong setting can negate any performance increase.

I started running RAID for the boot drives back in the SCSI days with the early 10K and eventually 15K RPM Seagate drives attached to Adaptec cards. Just like back then, some combinations work better than others. A proper review of an add-in-RAID card takes several weeks just to test. There is a lot of trial and error. Once the settings are right you can get a nice performance increase in the real world but it needs to be tuned for the workload you want to accelerate.

I always bought 15,000 rpm SCSI drives for my CAD workstations and with the business paying for them, enjoyed the speed benefit when gaming on those same boxes. However, in testing, I never say any improvements in RAID. **No** test / review has ever showed an improvement on the desktop and it's hard to accept the fact that "all these guys" got it all wrong and no one has ever been able to document this performance increase outside very specialized applications.

"However, many have tried to justify/overlook those shortcomings by simply saying "It's faster." Anyone who does this is wrong, wasting their money, and buying into hype. Nothing more."

The most cost effective solution is to have an SSD boot drive with Windows, Office, and some other stuff, and an HDD for storing your pictures, music, videos, documents, and large programs. I imagine it will remain that way for some time.

A SSD with apps / OS and an SSHD for data games gives you the best of both worlds. As users tend to play games and use files in streaks (starting editing projects in office / starting finishing games over a matter of a week.... with all the 'frequently used files" sitting on the SSD portion of the drive. That performance increase costs ya $20, a very worthwhile investment

I don't know why your SSD-equipped system boots so slowly. Maybe you're timing it from when the system is still running its POST. From the moment that my Win7 system POSTs and the screen displays the "Starting Windows" splash screen, I have a Windows desktop in less than 3 seconds. In fact, the splash screen doesn't even finishing "building" the logo before it goes to the Windows logon screen.

That's on a Samsung 830 SSD - not even one of the newer ones.

In order to eliminate reaction times, from the time you push the power button till the time you are on the desktop.

If your main point is saying a SSD is only beneficial in I/O-intensive operations, I would disagree because small seconds/minutes saved here and there with a SSD over a HDD definitely add up. Time is money, so any amount of time saved is benefiting you and/or your employer (disregarding what you actually do with it). Even gamers benefit from small (or significant) savings here and there; you can't just say "ssd's advantage just aren't really seen."

The problem is you are removing "the human factor". I had an employee who argued for putting an SSD in his box back when a 120 GB was $300. I suggested that he "make a financial case" for time saved versus cost.

I got back an analysis of published boot time differences x 5 days a week x 3 years and it didn't cut the mustard (< 1/3 return on investment) . In addition, in the days it took for him to prepare it, I observed his morning activity ....

Arrive, take off jacket, start PC, go to coffee machine, make pot of coffee, chat w/ the girls..... pour his coffee, sit at desk, reviews his phone messages, read what was dropped in his in box, log into PC. The SSD wasn't going to shorten any of those human activities.

Similarly, when handed a letter / report to edit, most load the program and the file while reviewing the markups....the file is open long before fingers hit the typewrite as the user looks to see what he / she has got to do .... same on the CAD workstations. Now with video editing, mass file conversions the extra speed does have a productivity impact but in most instances SSD speed will have 0 impact on office productivity.

It's true that a SSD won't increase FPS performance, however it makes a HUGE difference everywhere else. Just think about the common install size of a top tier title these days (20GB+). First if you're using digital you have to download all that data. During the download and install afterwards you're not going to be doing much of anything on your computer when it comes to disk access.

Multitasking perhaps ? When I am downloading / installing a game... I am reading tips and hints about potential problems, recommended settings, strategies, reading e-mails, ... The disk write is certainly limited by the download speed and I often also run disk backups during those times which, with this setup, has no observable impact as games and data are on different SSHDs.

Starting the machine a week after Patch Tuesday, if there's been no outcry of horrors, WU gets run and that's an ideal time to return phone calls, answer e-mails. If your "Things to Do" list for the day includes only "play games" perhaps the lost minute is infuriating but most peeps will be productive in those short breaks..... I may drive a Porsche to work but it's 185 mph speed doesn't mean much when stuck in 45 mph traffic. So I get productive by taking care of phone calls .... at the end of the day, I get the same amount of work done when I drive the Porsche or I drive the SUV. At the office, loading up that big CAD file I am working on, I can see a performance difference w/ a benchmark. But subjectively, the system is as responsive whether it's on the SSHD, SSHD or the HD..... the performance difference is smaller than my reaction time.

Same with gaming..... on my box where my oldest son (pilot) does his flight sims and lays the big AAA games of the day, (other sons do to when they have freinds over playing on their box) I have moved the games from the SSD, to the SSHD to the HD w/o telling anyone and no one has noticed. You can measure differences w/ a benchmark ..... but like booting Windows, you can't subjectively tell the difference between 16.5 and 15.6 seconds.

1 second a day x 220 work days @ $50 an hour / 3600 seconds a day = $3 productivity savings .... how many years it going to take to earn back that SSD cost ?

 

[2Be_or_Not2Be]

My storage support is broken down into:

a) Enterprise (Linux Servers, NAS etc)
b) High-end workstations (R&D, production etc)
c) Desktops (executive/office)
d) Mobility: Laptops and Tablets

Currently, I've 500+ machines in total as mentioned above. In this type of heterogeneous setup, only viable solution is mixing SSDs and HDDs. It means hundreds of storage devices. In some solutions, SSDs almost eliminated. E.g., NAS server based on ZFS (baring ZIL etc). Nevertheless, I've adopted SSDs whenever it's efficient and highly production say in workstations or office machines. However when it comes to Network storage, HDDs rule for now (volume cost factor).

I have found SSDs to be very effective in server storage for virtualization hosts when you want to use DAS rather than a SAN/NAS across the wire. They are also a good candidate for primary storage when high utilization/high throughput is a big consideration. True, cost is a factor, so everyone's situation will be different.

 

[2Be_or_Not2Be]

I always bought 15,000 rpm SCSI drives for my CAD workstations and with the business paying for them, enjoyed the speed benefit when gaming on those same boxes. However, in testing, I never say any improvements. **No** test / review has ever showed an improvement on the desktop and it's hard to accept the fact that "all these guys" got it all wrong and no one has ever been able to document this performance increase outside very specialized applications.

"However, many have tried to justify/overlook those shortcomings by simply saying "It's faster." Anyone who does this is wrong, wasting their money, and buying into hype. Nothing more."

So you're saying that all the reviews & tests on sites like Anandtech that have clearly shown differences in performance between a SSD and a HDD are just "wrong, wasting their money, and buying into hype"?

 

[WyomingKnott]

So you're saying that all the reviews & tests on sites like Anandtech that have clearly shown differences in performance between a SSD and a HDD are just "wrong, wasting their money, and buying into hype"?

That's not how I read it. I read it as saying that many people want it just because it's faster, although they will get no benefit from it. Since it wasn't directed at any particular user, but at a class of people making a mistake, it seems reasonable. Heck, I'm one of those people - I'm waiting for an m.2 NVME SSD to come out, and then I'll try to build a new system. Even though the one I have now fills my computing needs.

This has been a long thread full of different and sometimes opposing opinions. I have been pleasantly surprised that even the toe-to-toe arguments have not devolved into personal attacks; this is forum discussion at its best.

My personal take: In general, SSDs are faster than HDDs, especially for random IO. My SSD, which is only on an SATA II port (all I've got), is more "responsive" than my velociraptor, and makes a lot less noise (which, to me personally, is important). However, SSDs are smaller in capacity and more expensive than HDDs. Whether the SSD is a necessary upgrade depends on the situation, and whether or not it is cost effective depends on the situation. Discussion of the effectiveness or lack thereof in an SSD in various situations is constructive and illuminating.

JackNaylorPE wrote '**No** test / review has ever showed an improvement on the desktop and it's hard to accept the fact that "all these guys" got it all wrong and no one has ever been able to document this performance increase outside very specialized applications'" I worked on a project where we did a lot of clean-and-build-and-unit-tests on a large Java application. About 16 minutes on my HDD; about 9 minutes on my SSD. Jack, can you make what you said clearer, because my personal experience seems to contradict it.

 

[logainofhades]

I used a 250gb MX200, in a build, for a friend. The difference was quite noticeable, vs my 2x 500gb seagates, in raid 0. Boot up time was amazingly faster, and programs opened quicker as well. I would love to have an SSD, but the prices are still a bit high, for what I would want.

 

[xyriin]

I always bought 15,000 rpm SCSI drives for my CAD workstations and with the business paying for them, enjoyed the speed benefit when gaming on those same boxes. However, in testing, I never say any improvements. **No** test / review has ever showed an improvement on the desktop and it's hard to accept the fact that "all these guys" got it all wrong and no one has ever been able to document this performance increase outside very specialized applications.

"However, many have tried to justify/overlook those shortcomings by simply saying "It's faster." Anyone who does this is wrong, wasting their money, and buying into hype. Nothing more."

As for the actual performance benefits, they've been thoroughly discussed. Anyone who needs rapid IOPS will benefit from a SSD period. The only question is how much your individual use case benefits. I'll also toss it out that high RPM enterprise hard drives do still exist today, and they're MORE expensive than SSDs. They're priced for suckers who swear off NAND storage.

Some very common examples of cases where a user will easily notice the difference...
Torrents
Playing a movie while performing any kind of disk copy operation
Boot time
File operations (like copying)
Loading or saving any data that takes more than a couple seconds
Loading applications located on disk
Video game save, load, and install
Video production
Massive photo processing

And that is just performing any one of those items in a vacuum. You rapidly make it worse when you start multitasking any of those items.

Are you going to notice a difference after your web browser is already open? No
While working in a spreadsheet or word document? No

 

[jimmysmitty]

I used a 250gb MX200, in a build, for a friend. The difference was quite noticeable, vs my 2x 500gb seagates, in raid 0. Boot up time was amazingly faster, and programs opened quicker as well. I would love to have an SSD, but the prices are still a bit high, for what I would want.

I had the same RAID 0. Now I have 2x 2TB Seagates in RAID 0 for games/data and a 240GB Intel 520. I think you need to wait till the mass majority have SSDs. I hate working on normal PCs now, especially ones with the cheapest slowest HDDs possible. I get impatient working on them.

That said, I am waiting for multi TB SSDs to become much cheaper before I even think of upgrading my current one. That and I am hoping for the newer NAND tech that has much better P/E cycles.

SSD do give a bit of performance improvement, but not in everything. Games benefit in load time while not seeing any advantage in FPS. I would say now they are worth it but still a HDD has a much better $/GB so a small SSD for OS and a few apps with HDD for everything else.

 

[logainofhades]

I would need a 500gb SSD, at minimum, most likely. Just not important enough for me to spend the money on it. I have other projects that need my attention.

 

[JackNaylorPE]

So you're saying that all the reviews & tests on sites like Anandtech that have clearly shown differences in performance between a SSD and a HDD are just "wrong, wasting their money, and buying into hype"?

No I am saying that a 165 mph car doesn't get you to work any faster than one that tops out at 65. I am saying that benchmarks does not equate to productivity. I am saying we don't build computers to transfer 500 GB of files all day long, we build them to run applications and, except in rare applications there is no perceptible improvement. I am saying that no one has ever shown a positive ROI in adding an SSD to a system.

And yes, I am saying EXACTLY what anadtech said about RAID 0

http://www.anandtech.com/printarticle.aspx?i=2101

"We were hoping to see some sort of performance increase in the game loading tests, but the RAID array didn't give us that. While the scores put the RAID-0 array slightly slower than the single drive Raptor II, you should also remember that these scores are timed by hand and thus, we're dealing within normal variations in the "benchmark".

Our Unreal Tournament 2004 test uses the full version of the game and leaves all settings on defaults. After launching the game, we select Instant Action from the menu, choose Assault mode and select the Robot Factory level. The stop watch timer is started right after the Play button is clicked, and stopped when the loading screen disappears. The test is repeated three times with the final score reported being an average of the three. In order to avoid the effects of caching, we reboot between runs. All times are reported in seconds; lower scores, obviously, being better. In Unreal Tournament, we're left with exactly no performance improvement, thanks to RAID-0

If you haven't gotten the hint by now, we'll spell it out for you: there is no place, and no need for a RAID-0 array on a desktop computer. The real world performance increases are negligible at best and the reduction in reliability, thanks to a halving of the mean time between failure, makes RAID-0 far from worth it on the desktop.

Bottom line: RAID-0 arrays will win you just about any benchmark, but they'll deliver virtually nothing more than that for real world desktop performance. That's just the cold hard truth."

If you bought your SSD to get your name on web site leader boards, then yes by all means get an SSD.

If you bought an SSD with the mindset that the purchase is going to pay back financial dividends in worker productivity, it ain't gonna happen. The cost of the time saved will never equal what you paid for it.

 

[jimmysmitty]

So you're saying that all the reviews & tests on sites like Anandtech that have clearly shown differences in performance between a SSD and a HDD are just "wrong, wasting their money, and buying into hype"?

No I am saying that a 165 mph car doesn't get you to work any faster than one that tops out at 65. I am saying that benchmarks does not equate to productivity. I am saying we don't build computers to transfer 500 GB of files all day long, we build them to run applications and, except in rare applications there is no perceptible improvement. I am saying that no one has ever shown a positive ROI in adding an SSD to a system.

I will add to and agree with you in this way:

I think SSDs do have an advantage over HDDs but not in GB/$. But I have worked with and tested a SATA 6Gbps SSD to a OCZ Revo Drive 2 PCIe SSD that has reads at almost 2GB/s. In terms of how it felt in Windows, the SSD felt the same. The only benefit to the Revo drive was writing files from one part of the SSD to another part. But moving through the OS and running most applications felt very similar.

HDDs have been one of the biggest bottlenecks but SSDs are now changing that for the better. They just need to get cheaper and more reliable.

 

[JackNaylorPE]

Context - You can't look at the computer's performance w/o taking into consideration the human factor ... or "human lag". If my task is to edit a proposal that I marked up on my flight home, how long it takes to open the file .... 0.7 or 0.8 seconds is immaterial as I will spend 5 or more seconds flipping thru my edits before my fingers touch the keyboard. This is very similar to "cost of upgrades" ... when the world was pushed to move from W4WGs to Win95, the cost of the upgrade was put forth as the cost of the OS + some extra RAM..... The true cost as reported by InfoWorld was $2500 - $4500 per box in components, backup time, installation, downtime, training, adaptation and so on..... what was purported to bring an increase in productivity never materialized and corporate America never recovered that investment. What made it worse here is that W95 turned out to be 40% slower than W4WGs.

The SSD is no doubt faster doing it's thing but the tasks it excels at don't really change productivity. You can not separate the machine from the operator.... Opening a file 0.2 seconds faster doesn't get the operator's fingers on the KB any faster.

Some very common examples of cases where a user will easily notice the difference... or not

Torrents - usage not permitted in home or office

Playing a movie while performing any kind of disk copy operation - no problem .... unless you choose to do both on the same disk, should not be an issue... does anyone routinely do disk copies on a daily basis ? I run 4 backups with no issue daily.I often schedule streaming activities (i.e. CAD Training) to occur during such times. I stream movies, and when backups start or users accessing files, it's not like you notice.

Boot time - 15.6 seconds on SSD / 16.5 seconds on SSHD ... need a stopwatch to tell which is which

File operations (like copying) - I did bunch of those the day I built my machine.... none since that took more than 2-3 seconds tops. No impact as while file is saving, I'm multitasking on to the next task...while file is opening, I am reading what edits need to be made to it. I copied 36 GB of files to a thumb drive....would an SSD have done it faster ? Ot an read as fast as it can but it won't change the write time to the thumb dive ... in the end it doesn't matter .... I was doing something else (editing a CAD file) while it happened and the file copy had no impact.

Loading or saving any data that takes more than a couple seconds - Non issue, human factor is bottleneck

Loading applications located on disk - Non issue, human factor is bottleneck

Video game save, load, and install - So far no one has bee able to tell difference when I have put their game on SSD, SSSH or HD - That doesn't mean you can't measure a difference with a stopwatch, it means the differences were so small that no one noticed. Also in some of the tests I did, the server handshake was the bottleneck with test on SSD, SSHD and HD all taking 45.6 seconds till character could be controlled.

Video production - worthwhile investment

Massive photo processing - Worthwile investment ... if you do that sort of thing

And that is just performing any one of those items in a vacuum. You rapidly make it worse when you start multitasking any of those items. - My box serves as a file server with up to14-16 networked connections, runs backups (disk to disk) daily, runs external backups nightly, while peeps sleeping....no impact on my daily activities at all.

Are you going to notice a difference after your web browser is already open? No - agreed

While working in a spreadsheet or word document? No - possibly, if working a large spreadsheet or database open, you will see impact especially with multiple users accessing. I had one database that took about 18 seconds to load.

Back in the day, the storage subsystem was a bottleneck and AutoDesk apps force a lot of writes to disk. Being the biggest, most demanding program we have, I just opened in on my laptop SSHD and my desktop's SSD. Both took between 6 and 7 seconds..... digits scroll by too fast to take a read on the 2nd and 3rd digits but regardless, not a difference you can count on to improve the bottom line at 1 second a day.

I will add to and agree with you in this way:

I think SSDs do have an advantage over HDDs but not in GB/$. .

I'm looking at it a little different. The impact of GB/$ strictly as a purchase cost is immaterial..... the impact on worker productivity could easily show a return on investment, long before ... if there was one.

I haven't bought a HD in 4 years and I put an SSD in most of our builds simply because I am a geek and for $100 why not. But I won't try and justify the expense by saying it makes anyone work faster. Same w/ water cooling .... I like WC cause it's fun and gives way more aesthetic options.... it also makes the system waaaay quieter. But there's no way I could justify the expense from a performance standpoint as with today's CPUs and GFX cards, heat is not normally the limiting factor.

 

[xyriin]

Let's take your example of the flight home. SSDs use less power so even if you can't keep up with speed of the drive you're going to be able to get more work done before the battery in your laptop dies. That is a cost savings since productivity equals money. However, as has been stated many times over use case matters. I've already said that opening a Word document is such a low end task that it isn't affected. Let's use a different work example. You need to perform post processing on a few thousand images before sending them to a client. That is a work scenario where you WILL see a performance increase and save significantly more time.

You seem to be under the impression that torrents are illegal. Numerous companies (take Blizzard for example) utilize torrent technology to distribute their games or applications faster instead of archaic direct downloads. So usage of torrents is not only a legal practice, it's actually profitable for companies to use. You can obviously argue against torrents for other purposes, but the fact that they exist makes them a use case. If you don't prefer the term torrent we can always call it multi-P2P.

You tossed out the disk copy operation, which makes no sense. But then again you specifically state you reschedule your work flow to avoid this problem. So it IS a problem, you're simply using a work around to avoid it. Let's also be realistic, are you staying you should have a half dozen hard drives so that each unique disk task is only on one disk? If so you're losing the cost savings for hard drives. Additionally, if you're streaming it's not the same disk load of playing local media like a Plex server. Reorganizing common day to day activities to limit concurrent disk activity is not a solution, it's a band aid. And to answer your question, yes, if someone has a media server they're more than likely making daily disk copy operations. Granted I schedule my backups for early in the morning but that's no guarantee I don't have a friend trying to use my Plex server at that time. If the backup is happening on a hard drive and the person tries to pull up content from that disk then game over.

I noticed you used a SSHD in your boot time example which are 2-3 times more expensive than hard drives. I'd like to point out that SSHDs exist because SSD technology is a speed boost over pure hard drives. I see SSHD is simply a cost saving band aid though. It doesn't solve the entire problem with hard drives, it just mitigates a few issues.

You blow off file operations yet they're the most significant time savings. Sure, if you're running a small app then loading with a faster SSD may only save you a second. Just as an example on my work machine I have to load up Outlook, Internet Explorer, and typically a couple office files each time I log on. My machine crawls through these tasks...memory isn't maxed, processor isn't maxed, yet the disk is thrashing around like it wants to die. Literally a minute or two later I've got everything open. Performing a similar workflow on a SSD equipped machine leaves me ready to work in seconds. The other thing that kills my machine is the background enterprise software that manages patches, downloads updates, etc. When it kicks on my machine slows to a crawl or stops anytime I have to pull something from the disk. Just like before it's a pure disk load not a memory or processor limitation.

Video games...
http://www.hardocp.com/article/2013/12/10/hdd_vs_ssd_real_world_gaming_performance/

Here is what was better with the SSD, as you might guess; load times. Loading each game was significantly faster on the SSD. Transitioning maps during gameplay was also significantly faster on the SSD. Loading times were improved, and we had a better experience overall simply because game data loaded faster.

So yes, video games are faster any time you're loading from the drive.

As an added bonus I'll toss in this YouTube video, first one I found on a quick search.

https://www.youtube.com/watch?v=FQpiZ44GyYU

Crysis 2 SSD : 7,4 sec / HDD : 12,7 sec
Need For Speed Shift 2 : SSD : 8,1 sec / HDD : 12,6 sec
Assassin's Creed Brotherhood : SSD 7,9 sec / HDD : 11,4 sec
Fallout New Vegas : SSD : 2.4 sec / HDD : 6,3 sec
Dead Rising 2 : SSD : 16,1 sec / HDD 22,8 sec
Crysis : SSD : sec 16,5 / HDD 24,1 sec

But hey, forget all that stuff. You admit that you work your schedule around hard drive limitations ("while peeps sleeping"). Sometimes rescheduling like that just doesn't work. Maybe there are different values placed on free time or global interactions that can't be constrained to a set schedule all the time. For those people they see value in having a drive without limitations.

I think SSDs do have an advantage over HDDs but not in GB/$.

This is the biggest thing. I still have hard drives because I can't justify the cost in moving all 15TB to SSDs. However, I am up to 5.5TB of SSD storage these days and I cherry pick the data I put on them to make the most optimal use of the faster medium. The bonus for me is that I fully expect the SSDs to last a very long time. I've had hard drives die over the years (even in the years since I started using SSDs) but all my SSDs are still going strong. I've even sold off my older sub-500GB SSDs and their flash life was still at 99% or 100%.

As for return on investment, the enterprise is always looking at the bottom line. They're heavily invested into SSD storage these days. Not for everything of course, but definitely for frequently accessed data because it is a cost saver for them. As for me, my devotion to SSDs came from hitting limitations on hard drives during my use. Most often it was needing to do two types of disk operations at the same time and seeing the hard drive choke. As someone who tweaks my machine for the best performance I found that unacceptable.

 

[2Be_or_Not2Be]

So you're saying that all the reviews & tests on sites like Anandtech that have clearly shown differences in performance between a SSD and a HDD are just "wrong, wasting their money, and buying into hype"?

No I am saying that a 165 mph car doesn't get you to work any faster than one that tops out at 65. I am saying that benchmarks does not equate to productivity. I am saying we don't build computers to transfer 500 GB of files all day long, we build them to run applications and, except in rare applications there is no perceptible improvement. I am saying that no one has ever shown a positive ROI in adding an SSD to a system.

<snip RAID-0 red herring stuff>

If you bought your SSD to get your name on web site leader boards, then yes by all means get an SSD.

If you bought an SSD with the mindset that the purchase is going to pay back financial dividends in worker productivity, it ain't gonna happen. The cost of the time saved will never equal what you paid for it.

For some reason, you are denying examples that are given to you that clearly show there are cases wherein a SSD clearly saved time. You may be able to argue that in your case, you don't find any benefit from SSDs for office worker PCs, but you shouldn't be so strident in generalizing that is the case for ALL office workers.

In regards to the "human factor" - that's why I said I am not counting what the worker does with time they might have saved as it doesn't apply to the time saved by the SSD - it's out of the control of the SSD. If they kicked off a process that finished 10 minutes faster but they were off at the water cooler, I'm only saying that the SSD made it faster. The same for an app - if it loads 5 seconds faster but they were reading an email in another window, I'm only saying that the SSD had a time-saving effect - not what the person themselves actually did with it.

I'll add another example of where a SSD could help an office productivity worker - I spend 30 minutes installing SSMS on a otherwise acceptably performing machine. It had a 250GB HDD and during the whole install, I basically watched at an almost solid HD activity light. I've installed SSMS on other machines with SSDs, and the time was half that (actually, more like a quarter of that). So basically it cost my own time as well as that of the office worker, along with aggravation on both sides on it taking so long (which is hard to put a price tag on).

I can acknowledge that in the use cases that you personally experienced, you don't believe a SSD helps. But you also need to acknowledge that others have use cases for regular office workers where a SSD did have a performance benefit over a normal HDD. You can't generalize that "no one has ever shown a positive ROI in adding an SSD to a system".

Everyone has usage scenarios where they do -or don't - find benefits from SSDs over HDDs. You can't generalize everyone's particular situation.