Slashdot Mirror
Why Size Matters For Your SSD Purchase
Vigile writes "Performance analysis on solid state drives is still coming into clarity as more manufacturers enter the fold and more of the drives find their way into users' hands. While Intel's dominance in the SSD market was once undoubted, newer garbage collection methods from Indilinx and Samsung are now balancing performance across the the major players. What hasn't been discussed in great detail yet is the effect that drive capacity can have on overall performance. Some smaller drives (64GB versus 128GB) will actually use fewer data channels from the controller chip and thus will have lower transfer speeds. The article compares drives using controllers from Indilinx, Samsung and Intel." Note that PCPer greedily spans this review over 12 pages. Next time maybe they can keep it down to something more reasonable.
6.40 inches ought to be enough for anybody.
Intel X25-M 160GB totally dominates in IOPS and doesn't suffer in the other categories. A clean win.
Help stamp out iliturcy.
The odds of me reading page 2 of any article not paginated sensibly (reading a single page should take several minutes) are probably around 10%. Page 5? never.
I'll just be uninformed until information is published with a sensible pagination system. I'm okay with that.
paul reinheimer
They're not going to be more reasonable until we take a stand. Vote down the story, and make sure not to click the links.
I have never bothered with firmware updates and additional configuration steps with standard IDE, SATA, SCSI, and SAS drives. While looking around at various SSD, I found that you need to go though all of this additional crap to get things working right. OCZ, for example, has a whole forum dedicated to help tweak out their drives. http://www.ocztechnologyforum.com/forum/forumdisplay.php?f=186
From the article:
I was going to include a price comparison, but a few of the units tested (like the Corsair P64) don't seem to be carried anywhere as of yet. That said, prices generally do not sway far from the cost/GB of ~$2.75 set by Intel when they released their G2 drives at record low prices. The exception here is the SLC-based PhotoFast V4S, which will retail for a whopping $499 (that's $15/GB in case you ran out of fingers and toes).
Help stamp out iliturcy.
Um ... why would they do that if their 12-page version gets slashdotted anyway? The whole point of the splitting it up is to get page views.
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
Each page was was broken at logical point too. Sure, they could have done the same in 6 pages that didn't have logical breaks, but you know, some people think that everything should be on one long page. Yay for them.
Is it too hard to click "next" ???? REALLY???
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
What's with the 'signoff' tag?
(Off topic, I know... if only slashdot had a 'General Discussion' thread.)
Actually, fewer pages with more text content delivered per http request would reduce the load on the server. The bigger impact on the server is repeated visits to the hard drive and trips to the database. When one article requires 12 separate page requests, that cranks up the number of http requests coming in that have to be responded to with hard drive file reads and database queries.
Not knowing their specific server architecture, the above is a generalization. Caching, virtual memory mapped file systems, etc. can alleviate these bottlenecks.
Seth
$5 / month hosted VPS on linux = awesome!
I use adblock primarily for these sites.
Then you're doing it wrong! The plugin you want for 12-page reviews is AutoPager. It works like the /. home page, loading 'next' pages as you get near the bottom. It's even smart enough to strip off headers and footers.
That's why I boycott books. I mean the scroll format was obviously much better. Turning pages wastes time and saliva, plus there is the mortal danger of paper cuts. Until books and magazines come in a sensible scroll format I won't be buying either.
I know what you're talking about. However, for a report formatted the way this one is, likely not a good idea.
Chas - The one, the only.
THANK GOD!!!
Saliva???
Do you chew on your books?
Have you driven a fnord... lately?
You must wait a little bit before using this resource; please try again later.
How is this news?
The number of channels increases the theoretical and actual read and write speeds.
Did anyone NOT know this?
No one needs to look at capacity to guess the number of channels, and no one needs to dig around for a review site that cracked the bitch open / contacted the Chinese manufacturers to get the number of channels used.
All you need to look at is the specs, and in case they're lying, benchmarks. No guess work. No hunting for obscure information that might not apply to your particular hardware revision of the same SKU. No bullshit.
Is your biggest problem with my post really the fact that some people lick their thumb to help them turn pages.
You also have twice as many points of failure. Three times as many, if you count the RAID controller. Four, if you count the firmware in the RAID controller. Five, if you consider the increased load on your PSU, having to power the additional drive and RAID controller.
That's ignoring the additional traffic on the bus. That RAID controller doesn't work on FM[1], you know. Then again, it probably doesn't matter; it's not like you'd want faster disk access in a machine being used for video or audio capture, where bus congestion may be an issue.
Oh... That's precisely where you'd want it. Well, there, and servers; again, bus-congestion could be a bitch.
[1] F*cking Magic
APK quotes people (including myself) without context and should not be trusted. Just thought you should know.
You obviously never had the pleasure of owning an IBM Deskstar 75GXP. Alas, updating firmware, incantations, and anything/everything else did not keep me from suffering a failed drive and RMA hell. Twice.
Slightly off topic, but it's often forgotten that the filesystem also plays an important role in drive performance. Newer filesystems like NILFS (http://www.linux-mag.com/cache/7345/1.html) are created to suit SSD's instead of the legacy rotating media. It claims to hold the same performance, no matter how large the filesystem is.
Back on topic: We're seeing the same evolution with SSD's now like we saw it with spinning media several years back, when they started to increase the drive size ever more. Eventually these performance differences between larger and smaller drives will disappear: they will simply not be an issue anymore at all when you won't be able to get SSD's smaller than 200GB, like the similar trend with spinning media.
Obligatory link: http://www.youtube.com/watch?v=pQHX-SjgQvQ
Back when I was a kid, books used to come in a nice, easily scrollable format. Of course, that was before Gutenburg came along and messed things up with his new-fangled page-at-a-time printing... now get off my lawn!
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Yes it is when each spam filled page takes at least 5 seconds to load (at least in some cases).
Why OpalCalc is the best Windows calc
Quoting from TFS:
Note that PCPer greedily spans this review over 12 pages.
Methinks kdawson meant to say
Note that to allow reading on multiple web browser windows at once in parallel, PCPer spans this review over 12 channels, er, pages.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
What "additional traffic on the bus"? Between that outlandish claim and the claim that you have FIVE TIMES AS MANY POINTS OF FAILURE you sound like a FI[1].
[1] I bet even you can guess this one.
Brian Fundakowski Feldman
"Note that PCPer greedily spans this review over 12 pages." Guru3D is also guilty of this. They review at least 500 GTX 295 cards by different makers over the course of a month and the shortest review is always in the neighborhood of 14 pages. Usually they span 16-18 pages, though. I wouldn't mind that if it was mostly new information, but they always throw in the same filler over and over like how to calibrate your monitor. Seriously... if you need to throw that into every review just take up 18 pages for something that is no different from BFG than it is from XFX, then you need to rethink how you compose your reviews.
I think that in order to be classified as a Super Star Destroyer, it has to be at least 15km long. Although the salesman will tell you it's worth at least 20 regular Star Destroyers, the price you pay should be no more than the cost of 15 Imperial-class SD's. Also, be on the lookout for used SSDs. They may be infested with Conduit worms, affecting the ability of the SSD to fire its cannons.
Battlemaster--Game with friends in medival realms
Additional hardware on a bus leads to... oh fuck it, here's a car analogy.
I have one road. On that road, I have one car; this car can do whetever it pleases, whenever it pleases. When I add another car, both cars now have to watch out for each other, or they'll eventually crash.
This is traffic.
Also, yes, the more parts, the more points of failure.
I hope you enjoyed your meal; now, please, go troll elsewhere. I hear 4chan is nice this time of year.
APK quotes people (including myself) without context and should not be trusted. Just thought you should know.
So two 64GB drives in RAID0, or even on just on different SATA channels are faster than one 128GB drive.
Spread the word! This will change database disk design for decades to come.
---
I've got several OCZ 32GB SSD with the partitions aligned. Wildly faster for running my VM's off of. Windows boots in about 10 seconds onto a LAN. Don't really see a notable difference with my *small* database servers. That's likely because they load huge pages of the database into RAM and serve it from there. I have yet to try this on a large database server.
Yes, with you there: I provide all my content free but it does cost me, out of my pocket. I typically only recover 30% of my costs from ad revenue, and that fraction continues to fall.
Remember that 'sticking it to the man' on some kind of principle is no kind of principle at all. In some cases that 'man' is a fellow geek being kind to you; would you piss in his/my beer too on some kind of principle (CmdrTaco excepted, my liege)?
And *no*, I absolutely seriously do not want advertisers' money stolen by clicking on ads that visitors are not interested in as some kind of misguided 'tip jar': I'm an advertiser to bring new traffic in too.
Just allow a few more shades of grey in!
(I run NoScript, BTW, but primarily for safety; it's a jungle out there!)
Rgds
Damon
http://m.earth.org.uk/
In a RAID situation, with a real raid card, there is no additional bus data at all. The data is sent along the same bus to the RAID card. The raid card then splits it into whatever is requireed and writes it to the disks that are connected to that same RAID card. The RAID card creates more data yes (with parity etc) but this is all onboard and the card is designed for this. Data over the PCI-E bus is exactly the same
Normal people worry me!
Size DOES matter.
Data still travels the bus to control the card. Unless you're removing the onboard SATA controller... which is onboard... presumably you're not ignoring its existence...
APK quotes people (including myself) without context and should not be trusted. Just thought you should know.
You're saving 100GB. Whether it ends up on one drive or a RAID of a billion drives, it's still 100GB of data going through the bus.
I'll concede on that point. On modern systems, it's negligible. More hardware DOES still mean more bus traffic, however; that RAID card does use bus bandwidth to report status, as well. Unless, of course, you're not running any sort of daemon to monitor the status of the disks in your array; a highly unrecommended configuration, at any rate.
And still nobody can legitimately attack my point about increasing the number of possible points of failure.
APK quotes people (including myself) without context and should not be trusted. Just thought you should know.
I just ordered a Thinkpad T400S with the 128GB SSD option. It was impossible to determine whose SSD it is - some reports said Samsung, some said Toshiba, reps don't know. Small random writes (eg generating class files, but very commonly used in apps ranging from Pidgin to Firefox) can be quite slow with SSDs. I intend to use the device with Ubuntu installed for typical desktop use, and developing server ware that uses Java, PHP, and MySQL.
Anyone have insight with this device combination and know of any issues?
My "drive" is in a solid state, and it's size is massive.
Property is theft.
The codex was invented long before the Gutenberg printing press.
"[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz
Not if you put it out publicly. Want rules, go private such as a subscription model.
My computer, My time, my choice.
You can say anything you want (obvious 'fire' in a crowded theater, liable, slander etc. exceptions).
and people have a right to ignore any portion thereof.
Free speech != guaranteed audience or acceptance.
Haven't read the article yet, but if you want it read and the adds viewed, give folks a reason past a third grade 'not fair' argument.
Mcyroft
https://signup.leagueoflegends.com/?ref=4c3ed6600b6ea