Upgrade Advice: Does Your Fast SSD Really Need SATA 6Gb/s?

[cmcghee358]

[citation][nom]SpadeM[/nom]And you obviously didn't bother understanding my post. This isn't something that only Tom's does, most if not all sites when it comes to SSD testing, go for higher capacity drives while the low end kind of gets the short end of the stick. And since performance for SSD is even capacity related, then the majority of ssd sold (those under 128GB) should get more lime light. I do hope now my point gets across, if not, I'm sorry but there's nothing I can do to fix it.[/citation]

Yes it makes more sense now thank you. You are obviously too lazy to actually find what I was referring to so I will provide you with the link.

http://www.tomshardware.com/reviews/tests-ssd-review-solid-state,3103.html

 

[fausto412]

one of these days i will comprehend all this ssd info. so many caveats

 

[Draven35]

[citation][nom]Andrew Ku[/nom]If the trace was 100% random, none of the accesses would be zero sectors away.[/citation]

This is an incorrect assumption. If the trace was 100% random, there would be an equal chance of it being zero sectors away and of it being the maximum number of sectors away. The 'percentage of total data transferred' for every position in the chart would be equal (around 5.88% in this chart).

 

[JohnnyLucky]

Great article. I thoroughly enjoyed reading it. For the past 18 months there have been a few bits and pieces of information in technical reviews hinting the difference between ssd's is not as great as synthetic benchmarks would lead people to believe. This article confirms it for the typical user as well as acknowledges that for some specialized applications the differences might be important.

 

[Guest]

Such pragmatic advice! Certainly maps to my experience in upgrading a MacBook Pro / 15-inch Early 2008 with an SSD. The improvement in overall responsiveness on this dated machine is remarkable; new life to still useful machine!

 

[superflykicks03]

This was a great article...I've been wondering what the answer was to this question for a long time. Clearly using a newer SSD with SATA II does not result in much of a handicap for the average user and even most high end users. That being said, I have a theoretical question.

I've been wondering, theoretically, if using 2 64Gb SSDs in RAID 0 on a SATA II controller would "make up" for the differences between SATA II & SATA III. So, would x2 64 GB SSDs perform better than a single 128 GB SSD on SATA III. Kind of a noob question, but curious what you guys think...

If the question (assuming money is not a problem) is should I upgrade to a SATA III mobo and get a 128Gb SSD, would the best answer be, no, keep the SATA II board and RAID0 2 64GB SSDs?

 

[capt_taco]

This is perhaps the most informative and thorough article I've read here in quite some time, not to mention a much-needed one. Not only answered several lingering questions I'd had about SSD vs. HDD performance, but answered them completely. Well done.

 

[GullLars]

I wonder how they managed to get those 4K random read speeds at QD 1 on page 4 of this article.
The highest i've ever seen for SATA interface NAND SSDs is 40MB/s, and that was for an old read latency optimized SLC (Mtron). Most MLC SSDs are in the 20-35MB/s range, and a few dip all the way down to 15MB/s.

I've benchmarked a lot of different SSDs, and analyzed benchmark data others have generated, and i frankly can't see how these numbers can be correct when a Fusion-IO IoDrive 160GB SLC only pushes 65MB/s of 4KB random reads, and the MLC versions push around 45MB/s.

If all the other numbers except the M4 (which seems to agree with all benchmarks i've seen of it at 24MB/s) were taken at a queue depth of 3 or 4, the numbers could be correct.

Other than that, great article.

 

[soistheman]

Now I wonder, ignoring trim issue with raid 0, would it be overall better to have 4 ssd disks in sata2 vs 2 ssds on sata3?

 

[Draven35]

No, because the bus would get saturated with the 4.

 

[Guest]

Great article, I have this exact situation SATA 3 SSD on SATA 2 motherboard and I did wonder what I was missing out on.

 

[cgorange]

HDD manufacturers are starting to reduce their warranties from 5 years to 1 yearhttp://www.openpr.com/news/208868/M...y-on-HDDs-reduced-from-5-years-to-1-year.html

 

[peevee]

Andrew, what is wrong with random data reads on your m4? Usually this was the fastest drive in that category. Some firmware update broke it?
Also, you write that most data we deal with is sequentially is uncompressed. This is not true. We access sequentially only large files, and almost all large file formats (video, music, pictures) are compressed, specifically because they would be even larger otherwise. Even MS Office compresses its files now. And there are no 0-filled files in reality at all, even DLLs and such are only somewhat-compressible.

 

[cybersans]

i will buy after they reduce the price at least USD300 for 1TB. thank you.

 

[Draven35]

ok, so... 7 years or so?

 

[blazorthon]

Are people just thumbing down every comment regardless of content?

 

[deanjo]

[citation][nom]blazorthon[/nom]Are people just thumbing down every comment regardless of content?[/citation]
Yup, same old morons.

 

[clonazepam]

I had a SATA III SSD, the OCZ Agility III, originally purchased for a motherboard using nVidia's 750i sli chipset, and whatever the SB was... It has SATA II ports, but I believe the quality wasn't really up to snuff and the thing seemed to transmit pretty steadily at the peak bandwidth allowed on SATA I. Even with all of that crippling the performance, it was still worth the upgrade. Under these more extreme circumstances however, you may prefer 2 typical HDDs in RAID 0, except the seek times are still just as fast whether on a crippled SATA II, good SATA II, or a SATA III interface, as I saw it.

Another obvious reason for benchmarking the higher capacity drives is to present the readers with a best case scenario, and if they benched lower capacities, the winners would still win, and the losers would still lose at about the same margins in any given test. I'm assuming that to be true, and it probably is here and there.

As far as comparing RAID 0 performance between 2x sata II versus 1x sata III, or 4x sata II versus 2x sata III... if you are talking about the RAID options on a motherboard, its a software-based raid so there will be more overhead and you won't see a 100% increase in the various metrics. I'm no controller expert, but you'd want to conduct that sort of testing on pci-express raid controllers that bring their own processors and memory. The last time I looked a good entry into that arena costs more than any single 120GB SSD on the market, and more than some of the higher capacity drives too.

So the end conclusion is the worst SSD available on the market will outperform the best HDDs on the market in many of the tests... especially when shopping for a notebook upgrade. I hear it helps with battery life too.

 

[mutex7]

FANTASTIC article. I have searched the web for months for just this information regarding SATA II/III differences, sustainable speed (i.e. the clean erase issue) and the 'real world' comparisons between SSDs.

NOW, if I could get you to do just one more test. The laptop world is filled with the ubiquitous HM55/PM55 chipset and the much cited Intel power management issue with SSDs (the intelppm.sys issue). So, my question is, if you have one of these chipsets does it matter which SSD you buy (at least with regard to 4K reads)? I would really love to see this issue addressed with regard to 'real world' testing...not only for the results but so that this situation gets the attention I think it deserves.

Tom's Hardware really came through with this review. Thanks a lot and keep up the good work!!!

 

[insaneO]

All of these tests are synthetic and have nothing to do with real life.
I had 2 Raptors in raid 0 before and by using sleep mode my computer was very fast.
Sleep mode allow all of the loaded software to stay in memory so there hardly any delay to load software on the second or third etc....load.
I replaced that setup with SSD 120GB OCZ Vertex 6GB and I don't see any performance improvement at all.
I also tried moving this SSD between Sata 2 and 3 and there is no difference in real life performance either though ATTO reports double the speed on SATA 3.

And that is the reason synthetic tests can't be trusted.

 

[hbottjer]

Interesting, and timely article. I've got an ASUS P8Z68V-PRO MB (I believe Rev. 1) I am setting up for photo/video/maybe gaming so it has 2 Marvell 6gb, 2 Intel 6gb, and 4 Intel 3gb SATA ports. The MB manual states the Marvel doesn't support ATAPI devices, so no DVD drive there. The 3gb ports are for HDD RAID array. The 6GB Satas are for my Crucial 128gb M4SSD & DVD then. I've heard not to use the Marvel ports for SSD drives (no reason ever given). I have another SSD that only supports 3gb but would still like to use (for pagefiles/caching or SRT of the RAID array), even have an extra HDD.Maybe I should check the SSD in my other PC but not sure about Marvell controllers

 

[clonazepam]

[citation][nom]hbottjer[/nom]I've heard not to use the Marvel ports for SSD drives (no reason ever given). I have another SSD that only supports 3gb but would still like to use (for pagefiles/caching or SRT of the RAID array), even have an extra HDD.Maybe I should check the SSD in my other PC but not sure about Marvell controllers[/citation]

You just need to Google your particular Marvell controller. I use a motherboard with a marvell controller too and I wanted to find out more about it. I use Asus Maximus IV Extreme-Z. I haven't researched it in awhile, but I've read the marvell dont support atapi devices. There's also 2 different marvell controllers floating around there, the slightly older one not having as good performance as the newer. You'd have to look it up to find which one you have, and read up on it. Get the full model number. I noticed Asus' website specs for the mobo dont have the details just the generic name of "Marvell® PCIe SATA 6Gb/s controller" so you have to look it up. There's probably newer drivers available as well directly from marvell. Again, a Google search or two should answer every question about it. It's been awhile since I looked it up so dont quote me on specifics.

 

[insaneO]

I have Asus x58 Sabertooth with Marvell. I have tried SSD on both Marvel 6GB and Intel 3GB. Same results, they both work. What you probably heard is that if you want the full 6GB speed is to use Intel 6GB controller because Marvell 6Gb is not fully 6GB speed. More like 4GB. I don't have Intel 6GB controller on my board.

 

[insaneO]

To hbottjer: You are correct, there is a new Marvell controller on new motherboards but I don't know the model number. Still OCZ tech support advising not using for full performance.

 

[nissandets13]

is like everthing... a faster car will perform better on the hightway where there is no stops, while a smaller car will be more confortalble or perform better comuting in a busy city where you have to stop every 2 min.