Welcome Guest ( Log In | Register )

Bump Topic Topic Closed RSS Feed
10 Pages « < 7 8 9 10 >Bottom

Outline · [ Standard ] · Linear+

 The Solid State Storage Thread

views
     
everling
post Aug 29 2011, 06:41 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(iaknesnah @ Aug 29 2011, 05:47 PM)
finally i ended up using intel 320 series 80gb in raid 0. pretty happy with it no hickups no bsod no problem as for now... hahahaha
*
You bought two Intel 320 Series 80GB and then RAID-0 them? It would have been cheaper to just buy a single 160GB SSD. And then consider buying another 160GB for a RAID-0 at a later time.
everling
post Aug 29 2011, 08:20 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
That seems to be a good deal. Also, TRIM isn't overly critical to performance. Most SSDs claim to be able to do internal garbage collection, although less efficiently.

I hope the firmware is up-to-date or you've updated the Intel firmware. happy.gif
everling
post Aug 30 2011, 04:13 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(AlamakLor @ Aug 29 2011, 08:24 PM)
Aih, regardless I currently have 2 x 60gb vertex 2. One in my laptop and one in desktop now. For some reasons I only have 13gb free space on the desktop and I have nothing installed besides windows update.
*
If you have a lot of RAM, you should disable the pagefile. Also, try disabling the hibernate feature.
everling
post Sep 1 2011, 12:37 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
While manufacturing costs do go down as their yields improve, I think they're trying to make some extra profit while SSDs are still considered luxury and not commodity goods. It has been stated a long time ago that Intel's SSDs were sold up to 60% profit. No idea how true is that today.

On the other hand, if you were tracking certain models, their prices have went down significantly. I have seen an old SSD model reduced in price by 40% before it finally dropped out of the retail market.
everling
post Sep 1 2011, 06:06 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
stringfellow, other than some nice benchmarks, I'm not sure what else you're trying to inform us. You're comparing high performance, low capacity SSDs with low performance, high capacity HDDs. And it appears that you were very unhappy with your HDD's performance and have concluded to your satisfaction that your expensive high capacity (8TB) external storage was not what you should have bought. Is that it? unsure.gif

This post has been edited by everling: Sep 1 2011, 06:10 PM
everling
post Sep 1 2011, 06:36 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
I see, but I think your analysis is slightly flawed.

People who get blinded by SSD's high sequential read/write performance are not evaluating SSDs properly. It is the extremely high random read/write performance that are the SSD's primary strength and value. A 4x RAID-0 HDD simply cannot compete SSDs on operations that demand extremely high random I/O, and it is what makes SSDs feel snappy. It is for this reason that Facebook was willing to replace tons of expensive HDD Velociraptor RAID arrays with SSDs.

If you only require high sequential performance, HDD RAID arrays can indeed perform adequately for a cheaper price. That is what your benchmark program was testing, since it was testing sequential writes and reads.

This post has been edited by everling: Sep 1 2011, 06:36 PM
everling
post Sep 1 2011, 07:14 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
I had made a mistake in interpreting your earlier post with the graphs. I had read the third graph as another SSD benchmark and not the Pegasus benchmark. Sorry about that. sad.gif

I was never against the idea of using HDD RAID arrays to get sequential performance on par with single SSDs. But they serve different purposes and aren't suited to replacing SSDs for the niche that SSDs serve due to the random read/write performance. However I get what you mean now.
everling
post Sep 1 2011, 09:27 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
If your concern is the write limit, please do use SSDs as scratch disk. Its lifespan actually isn't very limited as most might think, unless the amount of scratch data written is a damn lot.

Assuming a perfect situation, the 3,000 write limit spanning over 5 years.

3,000 writes / 5 years / 365.25 days = 1.648 write/day

Thus to kill a 120GB SSD via write exhaustion on the last day of the fifth year, you need to write more than 192.50GiB per day. For a 160GB SSD, it is 256.67GiB per day.
everling
post Sep 11 2011, 02:07 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
People who already have 8GB of RAM really should consider disabling their pagefile. It is unlikely for most users to need more than 8GB. And if you don't use memory intensive applications (Firefox doesn't count as it isn't demanding enough to qualify) and your computer somehow managed to use more than 8GB, chances are that something is terribly wrong and that you need to look into it. Lately, the only reason I ever used more than 4GB of RAM is because of my virtual machines.
everling
post Sep 14 2011, 12:05 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(turion64 @ Sep 13 2011, 10:14 PM)
May i ask is the lifespan of SSD longer than normal HDD?
i mean if i want to use SSD for 24/7 read/write operations, is it a good idea?
*
QUOTE(AlamakLor @ Sep 13 2011, 10:20 PM)
Well to be honest, you won't feel the difference. My 60gb is almost outdated now, and the wear level is not even 1%......This SSD lifespan thing is overrated.
*
It clearly is overhyped if you do the maths.

With 3,000 writes over 5 years, you have to write over the entire SSD 1.643 times per day to kill it via write exhaustion by the end of the fifth year. If you want to kill it in your first year, you need to overwrite it 8.21 times per day. It is obvious that the average person are not going to be able to use their SSDs to such an extent.
everling
post Sep 14 2011, 03:36 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(turion64 @ Sep 14 2011, 02:20 PM)
can you explain it in layman terms?
i have used normal hdd for 24/7 and on average it died after couple of years.
how many years the ssd will be able to withstand on-going 24/7 read write operations?
*
A 128GB SSD should be able to write on average ~192GB per day for 5 years.
A 160GB SSD should be able to write on average ~240GB per day for 5 years.
A 240GB SSD should be able to write on average ~360GB per day for 5 years.
A 300GB SSD should be able to write on average ~450GB per day for 5 years.

Nobody would normally write that much per day. Even torrenting or downloading will have a difficult time. You need at least a constant 19MB/s download speed (Unifi VIP20 would be a start) to wear out a 128GB SSD at the end of the fifth year AND you will need to either delete or move your downloaded files out of your SSD everyday (because you can't write 192GB of data to a 128GB SSD without deleting something in the process).

tl;dr: SSDs are more likely to die for reasons other than its limited write lifespan. I have no idea what the average real world lifespan of an SSD is.

QUOTE(AlamakLor @ Sep 14 2011, 02:23 PM)
I'd say it's going to be more or less the same. The QUICKEST way to kill a hdd/ssd is running torrent on it. Always always keep download drive separated.
*
I'm not a heavy torrent user, but when I find the need, I always point it to my SSD. icon_rolleyes.gif

This post has been edited by everling: Sep 14 2011, 03:43 PM
everling
post Sep 14 2011, 11:39 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
That benchmark was run on your netbook? You could try running it on your desktop first. It may then look similar to most other users, due to different hardware.
everling
post Sep 15 2011, 12:05 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
If you're new to SSDs, how come you have both a Corsair Force 3 and a Kingston V100? The benchmarks according to ATTO looks fine.
everling
post Sep 17 2011, 01:18 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
Bad news if SandForce doesn't fix the BSOD problem. But Intel might buy it out and then fix the problem via a more thorough and exhaustive testing process.

I wonder if it is because of a patent issue, if SandForce managed to acquire the key patents to on-the-fly data compression, that would be a hell of an iron grip on the SSD market. As the feature size continues to decrease, as well as the corresponding write lifespan, on-the-fly compression would be the best way to reduce the number of writes while improving performance.
everling
post Sep 28 2011, 01:33 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(AlamakLor @ Sep 28 2011, 01:25 AM)
Regaining trust is not easy but OTOH people seem to forgive Toyota Camry's pedal issue rather quickly thumbup.gif
*
Because there was no real problem with the Camry's pedal issue. It was mass hysteria, and Toyota gave up on trying to reason with the hysteric mobs and 'fixed' the problem by issuing 'fixes' for the 'foot pedal problem'.
everling
post Oct 2 2011, 11:05 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(Anand Lal Shimpi)
When Intel entered the SSD market one of its declared goals was to bring the technology into the mainstream. The goal was so important to Intel that its consumer drive was branded X25-M, with the M standing for mainstream. Intel's desire for SSD ubiquity wasn't entirely altruistic however. Mechanical storage acted as a potential gate to increasing CPU performance. Eventually, without significant improvements in IO performance, CPU improvements would be less visible to most users. SSDs would help alleviate this bottleneck.

It wouldn't be untrue to say that Intel accomplished its mission. The client SSD market was in a state of disarray before Intel arrived on the scene. Although we still have problems today, there are a number of affordable options for end users and lots of competition. Samsung, Marvell, Indilinx, JMicron and even SanDisk are now vying for control of the market.

With healthy competition, significant performance improvements and (hopefully) improved reliability in the consumer SSD space, Intel will actually begin defocusing itself from this market over the coming years. Intel needs to keep margins as high as possible to appease shareholders, and the consumer SSD business is in a race to the bottom. Dollars per GB are all that matter here once you deliver a certain level of performance and reliability.

Intel won't abandon the consumer SSD market completely, it will still compete in the high end space but there's a good reason that the mainstream moniker has been dropped from Intel's product names. Intel will shift more of its attention to the enterprise space, bringing that technology to the high end desktop/workstation users where it can (e.g. Cherryville will be focused on both enterprise and enthusiast desktop users). But as you have already seen, I wouldn't expect Intel to actively compete in driving mainstream SSD pricing down further. That market now belongs to the players I mentioned above.

Source: http://www.anandtech.com/show/4902/intel-s...10-200gb-review

I hate it when such arguments make sense. Prices will come down, but it may take much longer. Here's to hoping that we can get 300GB SSDs for less than RM1,000 in a years time.
everling
post Oct 7 2011, 06:56 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
It's 98% used up? That sounds rather unusual.

A 40GB SSD with lifespan of 3,000 writes and a write amplification of 10x should be able to write: 40GB x 3,000 / 10 = 12,000GB = 11.71TB

Or is it saying that it has already used up 2% of its life?
everling
post Oct 8 2011, 12:55 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(iaknesnah @ Oct 8 2011, 12:26 AM)
that indicator is not reliable. according to the test after it reaches 0 the indicator will then goes up again.. lol.. funny


Added on October 8, 2011, 12:27 amand for 40gb ssd they managed to write more than 200tb on that ssd...
*
Writing 200TB on a 40GB isn't impossible if it has a SandForce-controller. With data compression, its write amplification factor can actually be less than 1, perhaps approaching less than 50% for text. 40GB x 3,000 writes / 0.5 write amplification factor = 240TB.

Also, Anand has stated that most manufacturers claim it isn't unusual for NAND memory to have much more than 3,000 p/e cycles.
everling
post Oct 8 2011, 07:23 AM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
From the first page alone, they were using the old stuff, I think with 5,000 p/e. With a write amplification of 1, that's 40 x 5,000 x 1 = 200TB. No surprise there. smile.gif
everling
post Oct 8 2011, 04:47 PM

Look at all my stars!!
*******
Senior Member
3,591 posts

Joined: Feb 2008
QUOTE(Anand Lal Shimpi)
In fact, I've often heard from manufacturers that hitting up to 30K p/e cycles on standard MLC NAND isn't unrealistic.

Source: http://www.anandtech.com/show/4902/intel-s...-200gb-review/2

So instead of only 3,000 p/e, you can get 30,000 p/e by chance!

10 Pages « < 7 8 9 10 >Top
Topic ClosedOptions
 

Change to:
| Lo-Fi Version
0.0539sec    0.88    7 queries    GZIP Disabled
Time is now: 4th December 2025 - 07:31 PM