• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

What's the most reliable 512GB SSD for ~$300?

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Heres a good idea:

Stopwatch measure loading times in games vers 1 single drive vs all in raid 0.

Bet it's barely anything...
 
I created the array to get the speed of an SSD with the capacity of a large HD. The extra speed is a bonus.
 
AdamK47 said:
Really? Explain. I haven't thought of that.
Click to expand...

raid strip.

8 drive = 8 time the speed. thertically - scaling not 100% :biggrin:.
8 drive = 8 opportunity to fail. all it take on once drive to fail to ruin the array.

vs

1 drive = 1 time the speed.
1 drive = 1 opportunity to fail.


1 < 8. any questions.
 
Its more than 8 times (can't be bothered to calculate it) because the drives are all following a probability graph of failure and since any of them failing could impact the others you are impacted by the first failure which causes a complete failure on them all, the failures are not independent events.

All things being equal if took 2 drives to get 512GB and 1 drive to 512GB the 2 drives will show near doubling of theoretical performance but also a great than 2x chance of failure. The failure chance scales more rapidly than the performance with Raid 0 such that failure chances get really high and why we have Raid 5 and Raid 6 for arrays greater than 3 disks.

You should really consider using Raid 5 at least on your array, you will loose 1 disk to parity and some performance but it will enormously reduce the risk of failure.
 
zCypher said:
I think what he is talking about is a ramdisk, not using an SSD cache to speed up SSD. Look at those numbers - 5GB/s.

and that's some nice download speed on Rogers, too bad you can eat up the 250GB cap in no time!
Click to expand...

In June Teksavvy.com will start offering the Rogers 150/10 cable package. $80 for 150/10 with 300GB cap and unlimited bandwidth between 2am and 8am. Alternately, 150/10 with unlimited bandwidth for $200 a month.

Teksavvy.com is the boss. I have 50/10 VDSL2 with unlimited bandwidth from them for $80. Also have a 28/1 cable line from them as well I may just get rid of. 150/10 is cool and all but I don't need that sort of download speed, it's more important to me to have a good solid upload (good in respect for what meagre offerings there are in the GTA) and a good enough download speed, priority being on no bandwidth caps.

The only offering that even interests me from Rogers is their 250/250 FTTH or Bell's 150/150 FTTH. Both are completely useless though as they have ridiculously low bandwidth caps and cost $250 a month. Rogers FTTH is not even available in my area, I think there is a just a small pocket downtown where you can get the FTTH service.
 
Last edited:
UaVaj said:
raid strip.

8 drive = 8 time the speed. thertically - scaling not 100% :biggrin:.
8 drive = 8 opportunity to fail. all it take on once drive to fail to ruin the array.

vs

1 drive = 1 time the speed.
1 drive = 1 opportunity to fail.


1 < 8. any questions.
Click to expand...

Actually, I was being my regular facetious self in the reply. 🙂

I already know the risks and take regular backups. I also have a two year replacement plan.
 
Last edited:
BrightCandle said:
Its more than 8 times (can't be bothered to calculate it) because the drives are all following a probability graph of failure and since any of them failing could impact the others you are impacted by the first failure which causes a complete failure on them all, the failures are not independent events.

All things being equal if took 2 drives to get 512GB and 1 drive to 512GB the 2 drives will show near doubling of theoretical performance but also a great than 2x chance of failure. The failure chance scales more rapidly than the performance with Raid 0 such that failure chances get really high and why we have Raid 5 and Raid 6 for arrays greater than 3 disks.

You should really consider using Raid 5 at least on your array, you will loose 1 disk to parity and some performance but it will enormously reduce the risk of failure.
Click to expand...

Keeping it as is. Appreciate the concern though!
 
Reason I said try one ssd vs the raid in game bench was to say if that speed is worth 30% or almost chance of complete failure. Try raid 5 or a similar redundant instead of 0.
 
BrightCandle said:
Its more than 8 times (can't be bothered to calculate it) because the drives are all following a probability graph of failure and since any of them failing could impact the others you are impacted by the first failure which causes a complete failure on them all, the failures are not independent events.
Click to expand...

Wait, what? Maybe my understanding of SSDs in RAID0 is a little rusty, but without any prior, you must assume that each failure probability is independent. In fact, the moment one drive fails, the entire array becomes unusable. The rest of the drives would no longer be accessed. If anything, this would reduce their chance of failure as they're no longer being used.

As far as I know, non-moving computer component failures are modelled using exponential distributions (i.e. CPUs, RAM, SSDs - basically anything that isn't spinning). If the part hasn't failed yet, the chance of it failing within the next month is the same as it's chance of failure in the first month of service.
 
Zxian said:
Wait, what? Maybe my understanding of SSDs in RAID0 is a little rusty, but without any prior, you must assume that each failure probability is independent. In fact, the moment one drive fails, the entire array becomes unusable. The rest of the drives would no longer be accessed. If anything, this would reduce their chance of failure as they're no longer being used.

As far as I know, non-moving computer component failures are modelled using exponential distributions (i.e. CPUs, RAM, SSDs - basically anything that isn't spinning). If the part hasn't failed yet, the chance of it failing within the next month is the same as it's chance of failure in the first month of service.
Click to expand...

LSI raid controllers can move data off a failing raid-0 set based on smart thresholds. Assuming the drive starts exhibiting failure and doesn't flat out die (or reverts to read-only mode) the system can pull the data off to a spare. It's pretty slick.

After all raid does not at all stop the need for backups.
 
BrightCandle said:
The other bonus is a significantly higher chance of loosing the contents of the drive. Its more than 8 times more likely to fail.
Click to expand...

Yeah and? You can also rebuild the array and restore the partition from backup MORE than 8 times faster, in seconds instead of the hours it takes with dino drives. That's the fun of SSD RAID 0: backing up and restoring your partitions to other SSDs at 1+ GB/sec over your 10 gig Infiniband home LAN.

If my RAID 0 goes down, I pull the bad drive, rebuild the array with either 1 less drive or a spare and it's up and running faster than it takes to get the RMA tag printed.

Stop thinking about SSDs in terms of primates with spindles. RAID failure on an SSD array is about as much as an inconvenience as rebooting a spindle for a Windows update.

Likewise the read/write is so fast, you are more likely to have up to date backups because it only takes seconds and doesn't interfere with your ability to use the computer. People tend to slack on backups of HDD data because it's so damn slow.
 
Last edited:
exdeath said:
Likewise the read/write is so fast, you are more likely to have up to date backups because it only takes seconds and doesn't interfere with your ability to use the computer. People tend to slack on backups of HDD data because it's so damn slow.
Click to expand...

:thumbsup::thumbsup:
 
KentState said:
Your chances of a reliable setup with 6 drives in RAID 0 is very small IMO. The performance hit in RAID 5 would be very small. At that point, you could risk one the cheaper 500GB drives.
Click to expand...

the performance hit in raid5 will be gargantuan and unacceptable...
Unless he gets a 300+$ hardware controller.
 
The figures I know of is that OCZ saw failures around 17% within 6 months and the general SSD failure rate was around 5% in a year. More recent (but smaller studies) have found failure rates to be somewhat higher than that but I'll go with the data from a year ago or so.

That means 8 way RAID 0 with OCZ drives has a 77% (1- (0.83 ^ 8)) of having at least 1 failure within the first 6 months.

The general industry average would be more like 33% (1- (0.95 ^ 8)) chance in the first year.

You can see the impact of compounding the failure rate of multiple drives. Its a trade off going Raid 0, and if you can't get at the extra performance normally then it may not be worth it considering the additional cost involved.
 
BrightCandle said:
The figures I know of is that OCZ saw failures around 17% within 6 months and the general SSD failure rate was around 5% in a year. More recent (but smaller studies) have found failure rates to be somewhat higher than that but I'll go with the data from a year ago or so.

That means 8 way RAID 0 with OCZ drives has a 77% (1- (0.83 ^ 8)) of having at least 1 failure within the first 6 months.

The general industry average would be more like 33% (1- (0.95 ^ 8)) chance in the first year.

You can see the impact of compounding the failure rate of multiple drives. Its a trade off going Raid 0, and if you can't get at the extra performance normally then it may not be worth it considering the additional cost involved.
Click to expand...


This makes no sense and your viewpoint is biased.

Fail post is a total fail.

Why do people think past singular issues with OCZ on early released products rushed to market suggest every OCZ drive is the same ?

Even if we all concede to you on the garbage data your suggesting about OCZ failure rates what effect at all does that have on the newer released models like Adam has ???

To my knowledge the OCZ Vertex 4 has proven far better so far and it's been out on market over a year now.

I think your reaching based in your own personal bias.
 
adam i think ur bottlenecking your PCH on your SB-E...

I bet if u had a dedicated RAID controller u could probably get 40% higher scores and faster speed.

If i recall ruby one showed me a video of a guy on an ARC-1680 controller (old controller btw) encode a move which would take 2 hours normally in about 15 seconds..
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top