Is A SATA 3Gb/s Platform Still Worth Upgrading With An SSD?

[ericjohn004]

[citation][nom]godfather666[/nom]That's not at all what the article is saying. It shows that there are sizable differences in performance between Sata II and Sata III, but that even with Sata II, and SSD provides enough of a boost in performance to justify an upgrade from a hard disk.[/citation]

SSD's definitely benefit from SATAIII. If you read and studied the whole article you would know this, but I guess you just skipped right through it to the conclusion because it's pretty obvious that SATAIII is sometimes twice as fast in synthetics, and faster in real world testing.

The only reason they have similar real world performance is because what we normally do on our computers doesn't push our SSD's hard enough to show a sizable difference between SATAII and SATAIII. But I bet you can find some real world tests that'll push these SSD's to their limits and really show a big difference.

SSD's are already saturating the SATAIII interface, thats why they are making a SATAIV. Normally when you reach a ceiling, you need to make the ceiling higher to continue to improve. So that's what they are doing. Would you rather they stop at SATAIII for the next few years and settle for 500mb/s? Or would you rather have a blazing fast 1,000mb/s SSD with 300,000 random I/O's?

 

[Marcus52]

One thing I would say, is that it is probably better to go ahead and buy a premium drive (or at least a modern one) than try to save buying a cheap 3Gb/s rated drive. Some bottom end SSDs won't give you the performance you are looking for, especially over time.

Besides, when you do upgrade to a 6Gb/s SATA capable rig, you won't have to buy a new SSD for it.

 

[godfather666]

[citation][nom]ericjohn004[/nom]SSD's definitely benefit from SATAIII. If you read and studied the whole article you would know this, but I guess you just skipped right through it to the conclusion because it's pretty obvious that SATAIII is sometimes twice as fast in synthetics, and faster in real world testing.The only reason they have similar real world performance is because what we normally do on our computers doesn't push our SSD's hard enough to show a sizable difference between SATAII and SATAIII. But I bet you can find some real world tests that'll push these SSD's to their limits and really show a big difference. SSD's are already saturating the SATAIII interface, thats why they are making a SATAIV. Normally when you reach a ceiling, you need to make the ceiling higher to continue to improve. So that's what they are doing. Would you rather they stop at SATAIII for the next few years and settle for 500mb/s? Or would you rather have a blazing fast 1,000mb/s SSD with 300,000 random I/O's?[/citation]

Were you replying to me, or to the person I was replying to?
Because that's exactly what I was saying.

 

[logainofhades]

SSD's are still too high of price for me to consider one. I still see SSD's as a luxury item. They aren't going to give me more FPS, so money would be better spent upgrading my graphics. That is the only thing I really care about. SSD really won't improve file transfers to my 4x2tb Raid 5 array as they are all 5400 rpm drives. Boot time means nothing to me. I start up my system then go get something to drink or whatever most of the time anyway. My 2x 500gb Seagates in Raid 0 have been sufficient for my needs.

 

[balister]

[citation]Modern SSDs slam right up against the SATA 6Gb/s interface's throughput ceiling...[/citation]

Really guys? I know of not a single SATA III based SSD that hits 750MB/s sequential read (which is the limit of SATA III -- 6 Gb/s). As noted in the article sequential reads are not reaching 600MB/s (which means there's still head room, about 1.2Gb/s or more -- 550 MB/s translates to 4.4 Gb/s, still 1.6 Gb/s below the saturation point for SATA III). Less sensationalism would be nice.

Given that saturation for SATA II is 375 MB/s (3 Gb/s), it makes sense that putting a new SSD on a SATA II port wouldn't be an issue unless you are doing a lot of sequential reads (and even then, if you have an SSD that is around 400MB/s sequential read, you wouldn't lose that much performance).

 

[bambiboom]

As usual, TH has produced a well-documented and useful description and measurement of hardware. Of particular value is the summation including the experiential context that in many everyday uses, SSD's, whether connected at 3 or 6GB/s, are not astounding time savers overall. But, there are probably many of us that are still on the mechanical /solid state fence.

For we Great Undecided, this drive comparison would have been even more useful had 7200 RPM 6GB/s and a 3GB/s mechanical drives been included. The usefulness would derive from the probability of the proportional presence of 7200 RPM drives in typical systems. The Western Digital Velociraptor at $230 is only about $10 less than the Samsung 840 used in the tests. I can't support it with figures, but I can imagine that many more systems will have something like 1TB, 6GB/s WD Caviar Blacks at $95, Seagate SV-35 at $110 than any 10K RPM drive, and of course, many, many of us use antediluvian 3GB/s drives. My ancient banger Precision T5400 uses a Western Digital RE4, 500GB that makes average (sequential and random) reads and writes at around 130MB/s . I have purchased an LSI SAS/SATA RAID controller that works in the old PCI-X 133MHz (not PCI Express) server slot to see if I can really get anywhere near the kind of amazing transfer numbers LSI used in the adverts- 2.4GB/s for very large sequential files!

I for one, don't mind that Windows 7 takes an extra 30 seconds to load- I get in two more sips of coffee before settling down to work. I do however, move/copy large blocks of files- typically 15 to 20GB and do get impatient then. My major storage frustration is a Patriot 8GB flash drive that never ran over 1.5MB/s. Yes, probably formatted FAT32, but I never think about that drive until I need it.

The future of mechanical drives is not over, mainly in that, as the real-world use it is still competitive and sudden fail mech'l HD's have hope of file recovery as compared to SSD's, but more so that a 1TB ,7200RPM, 6GB/s mechanical drive is 1/3 the cost of a 256GB SSD. When four times the storage costs one third, there have to be incredible performance benefits and I don't yet see this, and for me, this comparison does not significantly illuminate the entire equation.

BB

 

[swifty_morgan]

be nice to see you do this test with an 8 year old machine running XP.................. and yes, xp actually runs on an ssd...............

 

[sanilmahambre]

Such a heroic demonstration of samsung 840 SSD. I am sure this is just the beginning of a blazing speed ERA.
But best technology is something that comes in handy! isn't it?

 

[jcamel24]

I can definitely vouch for this. Have an old emachines with a single core Celeron and 5400rpm HDD. First I put in a T8300 dual core proc (used for $23 online) and it didn't make much difference. But when I put a 90gb Corsair Force 3 in, feels like new again!

 

[Combat Wombat]

[citation][nom]jimmysmitty[/nom]Any SSD on 3Gb/s kill Raptors outright. Even my older X25-M does due to the sheer IOPS compared to a Raptor or any mechanical HDD. I do need to upgrade but not for the speed, mostly for size. 80GB is not enough even for OS and a few apps. I have messed with everything from a SATA II SSD to a PCIe SSD (Revo 3) and as long as you are on SATA II or better its going to be more than fast enough.But that said, I might just wait for Broadwell and SATA Express.[/citation]

Yeah, true dat, but if I were to get an SSD, i wouldnt muck around, i would just get a 500GB+ and be done with it haha

 

[king smp]

Great article
wouldve been nice to see the Seagate Momentus XT Hybrid SSD/Mechanical drives thrown in there for comparison.
I have two of the 500gb sata II hybrids in raid 0 with a sata II Gksill Phoenix Pro as an OS drive
good setup for what I do and relatively cheap
total cost was $200 for all the drives
I do plan on going to a Sata III mobo soon but this article makes me think it might not be a priority

 

[mapesdhs]

jemm, indeed do not use the Marvell ports, they're pretty bad. The only
time I've ever had weird issues with SSDs (eg. failed secure erase) is
when a unit has been connected to a Marvell port, and as you have
probably been told already, performance is often woeful.

I recently tested a number of different SSDs on SATA2 ports (P55 and
nForce 790i to begin with, will test AMD, X58, etc. later), the upgrade
is absolutely worthwhile, and (slomo4sho!) I included a Samsung 840
250GB in the tests, which performed very nicely indeed (so much so that
I bought one for my 3930K setup). See:

http://www.sgidepot.co.uk/misc/ssd_tests.txt

As a more real-world example, my brother's P55 system (an overclocked
i7 870) used to have a WD VR 150GB 10K system disk. The main game he
plays atm took more than 3 minutes to load from the WD VR, but after
replacing it with a Samsung 830 256GB (my xmas gift), the game now
loads in less than 10 seconds. As others have said, the difference
really can be night & day, but it depends on the task. Boot to Windows
is much faster, but the speedup is less significant than what he saw
with the game loading time.

I'll be testing other SSDs as & when I can obtain them (expecting a
Vector 256GB this week). Hope the data is useful to those still using
SATA2! And I will try to get round to testing an AMD board soon, and an
X58. Not yet tested the 830 or Vector/128 on the 790i either, will do
that next week if I can.

Ian.

 

[mapesdhs]

[citation][nom]Marcus52[/nom]One thing I would say, is that it is probably better to go ahead and buy a premium drive (or at least a modern one) than try to save buying a cheap 3Gb/s rated drive. ...[/citation]

Nope, not so, look at my results (see link above).

As toms has said many times before, the real benefit is in having an
SSD in the first place. When used with SATA2, compared to a mechanical
HDD, there isn't that much difference between a premium Vector/128 or a
newer Samsung 850, vs. an old Vertex2E. This is why I snap up used
V2Es, etc, when I can, they're ideal for use with the older mbds I
have, eg. ASUS M3N-HT Deluxe.

Ian.

 

[-Fran-]

I really don't feel my cheap-dirt 500GB 7200 RPM x2 in RAID 0 are slow.

SSDs need to go down more in price for me (at least) to consider recommending them over cheap RAID-0 setups.

Cheers!

 

[mapesdhs]

bambiboom writes:
> For we Great Undecided, this drive comparison would have been even more
> useful had 7200 RPM 6GB/s and a 3GB/s mechanical drives been included.

See my earlier link; I tested a typical Samsung F3 1TB 7200rpm aswell.


> in the tests. I can't support it with figures, but I can imagine
> that many more systems will have something like 1TB, 6GB/s WD Caviar
> Blacks at $95, Seagate SV-35 at $110 than any 10K RPM drive, ...

That's why I also tested the Seagate 750GB. Though an Enterprise drive,
its speed is typical of older SATA2. The difference vs. an SSD is
just enormous for some tasks (look at the AE loading speed).


> purchased an LSI SAS/SATA RAID controller that works in the old
> PCI-X 133MHz (not PCI Express) server slot to see if I can really

I have the same card and have run numerous tests on a wide range
of drives (actually I have more than a dozen different cards, from
3041s to 3800Xs, both PCIX and PCIe versions). See:

http://www.sgidepot.co.uk/diskdata.html


> anywhere near the kind of amazing transfer numbers LSI used in the
> adverts- 2.4GB/s for very large sequential files!

Ah, no, you won't. It's quite likely your mbd implements the PCIX
link by using a PCIe/PCIX conversion chip, as mine does I believe
(an ASUS M2N32 WS Professional), so the max I/O is somewhat less
than LSI's PR. Until recently the old ASUS M2N had a 300GB 15K
drive, leaves standard SATA in the dust, but it's no match for
the 120GB Vertex2E with which it was replaced back in Jan.


> I get in two more sips of coffee before settling down to work. ...

Sips?? I use lots of milk, gulp my Earl Grey in seconds.


> however, move/copy large blocks of files- typically 15 to 20GB
> and do get impatient then. ...

If you're working a lot with large sequential files, then using
mechanical disks in RAID0 (or RAID10 for protection) will still
offer good performance characteristics.

The key as always is to use the right type of storage for the
appropriate task. The more sequential the nature of the task,
the more a mechanical drive may offer adequate performance; I
say adequate because an SSD will still be quicker, but of course
some types of sequential I/O also mean the manipulation of very
large amounts of data, eg. 2K editing in Flame, in which case
mechanical RAID0 is more sensible.

Btw, you can of course connect SSDs to SAS RAID cards, but older
cards will only configure the ports at 3Gbit/sec, though this
can in some cases be changed by using particular fw releases.
And remember that in RAID0 the fast random I/O of SSDs will be
underused due to the lesser max IOPS rate of the card's controller.
Only much more recent or more expensive RAID cards have chips
that can really exploit the random performance of SSDs.

Also, you can mitigate the issues surrounding random I/O with
mechanical drives in RAID to some extent by using a RAID card
which has some onboard cache RAM, eg. cards such as the HP P410.
Some of these can be obtained quite cheaply 2nd-hand, including
variants that are PCIe-based, allowing for more flexible use,
and the backup batteries for the cache RAM can still be bought new.
I ran a lot of SAS RAID tests with a P67 board, got some pretty
good numbers (better than with PCIX).


> real-world use it is still competitive and sudden fail mech'l
> HD's have hope of file recovery as compared to SSD's, ...

You'll hear endless conflicting stories on this, but in my case
out of several Samsung 1TB SATAs, two have had to be replaced
under warranty so far, whereas all of the SSDs I've obtained
(more than 40 in all) are still working A-ok. The key is to
update the fw as soon as you obtain an SSD.


> ... a 1TB ,7200RPM, 6GB/s mechanical drive is 1/3 the cost of
> a 256GB SSD. When four times the storage costs one third, there
> have to be incredible performance benefits and I don't yet see
> this, and for me, ...

There _are_ signifcant performance gains, huge in some cases,
but as always it depends on what you're doing. What is your main
task or tasks?

An SSD really does improve general responsiveness when used as a
system drive (background tasks like virus scans become far less
intrusive), and/or main apps can be installed on a separate SSD
if preferred (eg. my gaming PC has all the game data on a 120GB
Vertex3 MAX IOPS, while the C-drive is a normal 120GB Vertex3;
the same system has a 1TB SATA for video archive and DVD images).

The right tool for the right job. 8)


> ... this comparison does not significantly
> illuminate the entire equation.

I hope my data can help, and feel free to describe in more detail
for what purpose you use your system, I'd be interested to know.
Perhaps I could run a relevant test or two?

Btw, I also have a Dell Precision T7500 (two XEON X5570s, 24GB
RAM, 600GB 15K SAS, Quadro 4K, etc.), so I can use that to run
tests with a 'newer' pro system if you prefer.

Ian.

 

[mapesdhs]

Manuel/Achim wrote:
> ... the fastest desktop hard drive we've ever benchmarked, Western Digital's ValociRaptor.

Have you guys ever tried a Seagate 600GB 15K SAS? (the 15K.7 edition) You really should.
And it's more representative of the top-end drives used in OEM pro systems, etc.

Btw, there's a typo in your sentence; after so many JP movies, spelling that word should
be 2nd-nature by now.

Ian.

 

[dimar]

How could you leave SATA I?

 

[bambiboom]

Ian,

I very much appreciate your very detailed reply and additional information regarding the performance of mainstream mechanical drives relative to SSD's. The last several months I've become a Passmark Performance Test junkie, and perhaps there's a weighting to the various test parameters at work, but in looking at a lot of benchmark system ratings results, the SSD systems are improved in almost every parameter and there is clear divide in the overll rating where the mechnical drives become SSD's. Having a fast disk system seems to be coincidental with higher CPU and memory scores. The fastest systems seem to have what I think of as "direct injection" PCIe drives- I see a lot of RevoDrive 3 X2's in top rated systems- as well as LSI SAS/SATA 92XX 8 port controllers. The key to this being useful is making an experiential connection between a fast disk and real world applications' use.

And especially relevant to the Mrch'l / SDD dicussion opened by the article this site id the information on your site regarding a larger variety of drives as well as obsolete LSI SAS/SATA RAID controllers. I would encourage readers of the article under discussion to visit the links included in "mapeshs: comments. As soon as I can find an alternative to my Linksys 600 Wifi card- thereby freeing a PCI-X slot, I will give the LSI SAS3080X a spin- literally. I didn't really expect the 3080X in the T5400 to meet the LSI chart- I'd be very happy if it performed at 50% those claimed rates. When I make transfers from partition to partition or from RE4 to Barracuda, Win Explr reports never more than 30MB/s !

I've toyed with the idea of RAID 0 and 1- and their combination- 10- but haven't taken the plunge. I'm very much behind in the disk realm. Actually, I'm going to be setting up my first RAID (1) on Thursday. As a gift, I'm upgrading a Dell Precision 390 in my brother's architectural office- new 2.93GHz Core 2 Duo to replace the 1.86, a Quadro FX 1700 (512MB) to replace a Quadro 550 (128MB), 64-Bit XP Pro, improve RAM from 2 to 4, set up a RAID 1. Total cost- $100.

I see mention on your site of Irix and that brought back a flood of memories of the early 90's when I first started with CAD. In the summer of 93, I bought an IBM 486 with a 50MHz DX2 ("math co-processor"), 2MB RAM, and 85MB HD, running DOS 6, Win 3.1, AutoCad 10 DOS, Corel 3, WP 6- the first GI version, and various others. Adding 2MB RAM to have the maximum 4MB cost $180 -that's $9,000/GB and the replacement 540MB HD was $570- which at that rate would make a 1TB cost over $1,000,000.

An office in Santa Monica where I knew some people had just installed a number of Silicon Graphic Indys and I was jealous of their 100MHz processors and 16MB RAM, though they were more than twice as expensive as the IBM- about $5,000 without software. I thought Irix and the CAD applications much more elegant than DOS/Windows and SGI had an office on Ocean Blvd where you go for training and they had a vast server- it must have been a 5' cube in a glassed in in room. It may have been 1TB, I'm not sure. An architectural client in Del Mar, CA had a neighbor who had an SGI Indigo and sat in a custom reclining chair with a keyboard and monitor on a stand, designing Fender guitars in 3D. You could click on a drawing area and a popup viewport would display a detail block with annotations- astounding to me then. I thought I'd time-traveled to the future! I think that Indigo setup was in the $20,000+ range. The good old days are now!

Speaking of HD's and cost/ storage/performance, my father was involved in the system for a shipyard building nuclear submarines and came home one day (1971?) quite excited about an impending upgrade to the system- I think it was an IBM 360 and he described the upgrade as having 50- 2' diameters platters, was the size of a small oil drum, and cost $50,000. Capacity of this huge and $50,000 device- 5MB! Many forget the days before HD's of programming on cards and programs and storage was on tapes.

A bit more about my system in case you have disk system or test suggestions > Dell Precision T5400, 2X Xeon quad core X5460 @ 3.16GHz, 16GB DDR2-667, Quadro FX 4800 (1.5GB), WD RE4 /Segt Barcda 500GB, M-Audio 2496 (soundcard) > Windows 7 Ultimate 64 > AutoCad, Revit, Solidworks, Sketchup Pro, CS4 MC, Kerkythea (soon Vray) Corel Technical Designer, WdPft Office, MS Office. Tasks include 3D models, industrial design assemblies (up to 6,000 parts), renderings of same, Patent applications and drawings, architecture, and descriptive industrial design proposal/descriptions. I believe the T5400 disk system (along with the slow memory) is a serious weak link in the chain.

Again thanks for contributing to the mechanical HD / SSD debate. My intuition is that very fast disk systems along with GPU computing will make Personal Supercomputers the future of the PC which will split off into the PS and tablets and pads/smartphones will handle PC applications.

Cheers,

Bambiboom

 

[dark_knight33]

I remember emailing Toms asking for this very article a few months back. Thanks!

 

[bloosteak]

i just put in a corsair m4 on my system with P35 mobo and core 2 e7400 and it was certainly worth it
I used to get choppiness in various console ports such as tomb raider but it's gone now.

 

[swifty_morgan]

I really don't feel my cheap-dirt 500GB 7200 RPM x2 in RAID 0 are slow.

SSDs need to go down more in price for me (at least) to consider recommending them over cheap RAID-0 setups.

Cheers!

seems they need to work out the 4k r/wr speeds. what else is more practical for "average" everyday use?........ and then drop the price....... really.

 

[rothingham]

Very nice comparison!

(typo on page 3. i7 3690x -> i7 3960x )

 

[Gautam Sharma]

I have had the same issue as Jemm. I think the design is just not up to par - we should be able to all 6 GB/Sec ports (instead of 2) to create a RAID array instead of mismatching. I have 4 Samsung 840 Pro's SSD's running in RAID 0 on my ASUS Maximus Extreme with Intel 3970X running just fine, 2 on a 6 Gigs/Sec channel and 2 on a 3 GB/Sec Channel. I am happy to report that luckily the difference is not much really even though 2 of my SSD's are running on a 3 GB/Sec port. I get a data throughput (Sequential Read) of about 1.7 Gigs/Sec or 1700 MB/Sec (4x RAID 0, Ports 6G and 3G combined). I tried every conceivable option before settling on this. Here are some results:

SATA 6GB/s Port: 2 x Intel SSD 520 Series 180GB: 440 MB/Sec
Mixed Mode [SATA 3GB/6GB Ports]: 4 x Intel SSD 520 Series 180GB: 880 MB/Sec
Mixed Mode [SATA 3GB/6GB Ports]: 6 x Intel SSD 520 Series 180GB: 1100 MB/Sec
SATA 3GB/s Port: 2 x Samsung 840 Pro 512GB SATA III SSDs: 700 MB/Sec
SATA 6GB/s Port: 2 x Samsung 840 Pro 512GB SATA III SSDs: 975 MB/Sec
Mixed Mode [SATA 3GB/6GB Ports]: 4 x Samsung 840 Pro 512GB SATA III SSDs: 1700 MB/Sec

And BTW, I am more inclined to also come to the conclusion that the Thunderbolt port on my Maximus isn't all that. I have a 8 Port Tower RAID SATA III Drive enclosure - in which I have 8 x SATA III 6GB/Sec 7200 RPM 2 TB drives, running in RAID 0 Mode which shows data transfer rate using USB 3 [SS] equal to or better than my external thunderbolt storage device. I regularly see data throughputs of up to 600 to 800 MB/Sec, which is almost the same using thunderbolt.

 

[mo976]

Its kind of a weird comparison! Taking one of the fastest SSD's available and one of the fastest HDD available and seeing what happens when you cripple it. A better comparison would have been your average 7200 rpm HDD vs what most people with an older system might buy say on sale. For example newegg shows the 128gb samsung 840 (not pro) and sandisk drives for 100 but 128gb can be found for even less then $100 on sale. Also most people looking to make that upgrade that lack 6gb sata would have the old core i chips or even the core 2 duo chips. Of course top of the line hardware makes for best benchmark results all else being even but in this case other then out of a "scientific" need to see what would happen it hardly represents the benefits one could expect when upgrading an older machine.

 

[Be0wulf22]

It's not just about maximum throughput. Response time doesn't get nearly enough credit. An SSD would still provide a dramatic difference even on PATA 66.