SSD general ?, and SSD in a 2008 2.8 8-core?

[300D]

So are we on the same page yet?

Yes. Its still painfully obvious you have no experience with an SSD.

Thing is, SSDs are glorified CF cards. Even if there are no moving parts, I would never do a back-up on a CF card

I'd feel far safer keeping my important data on a flash drive than 3-6TB of data across 3 mechanical drives were ALL the data will be lost if 1 fails. Good luck finding a 6TB drive to backup onto.

 

[Tesselator]

Yes. Its still painfully obvious you have no experience with an SSD.

Why do you keep saying that? It's just like you're only out to cause trouble or something. You offer no numbers. I have. What's the deal?

I'd feel far safer keeping my important data on a flash drive than 3-6TB of data across 3 mechanical drives were ALL the data will be lost if 1 fails. Good luck finding a 6TB drive to backup onto.

It would ONLY be your feeling. The stats and numbers clearly show rotational data to have a MUCH higher integrity and fault tolerance! Also backups only need to be as large as the sizeof the data you're backing up - not the sizeof the media! Nice strawman tho...


.

 

[300D]

Why do you keep saying that? It's just like you're only out to cause trouble or something. You offer no numbers. I have. What's the deal?

Because you keep talking as if you actually know what you are talking about!

The stats and numbers clearly show rotational data to have a MUCH higher integrity and fault tolerance!

Not in a RAID, the failure potential is exponential to the number of drives.

 

[nuckinfutz]

I don't think there's enough relevant data out there on SSD regarding durability to make an informed decision.

 

[Tesselator]

Because you keep talking as if you actually know what you are talking about!

Yeah, see. Just trouble-making. Why? It's not conducive to the discussion at all.

Not in a RAID, the failure potential is exponential to the number of drives.

As has been shown again and again here at MR this is 100% false.

I don't think there's enough relevant data out there on SSD regarding durability to make an informed decision.

I think in the case of SSD user data isn't needed. The manufacturers have spoken. They know exactly how many times they can write to a cell before it fails. From there the rest is just math. This is dissimilar to rotational magnetic media where bit failure rates vary greatly depending on any number of conditions.

 

[300D]

Yeah, see. Just trouble-making.

I don't know, why are you here causing trouble? You have added nothing to the discussion about SSDs, only spreading your biased opinion against them in an effort to persuade people into wasting money on a low-end RAID instead.

As has been shown again and again here at MR this is 100% false.

That is a false statement. All mechanical devices have a failure rate and having your data depend on the integrity of multiple mechanical drives increases that risk exponentially with every drive you add.

This is dissimilar to rotational magnetic media where bit failure rates vary greatly depending on any number of conditions.

That is because there is physical electron movement in an SSD cell where a HD only changes polarity on the disk surface. The average SSD will long outlive the average computer's lifespan, even a mac. In a professional environment the computer will be replaced with a newer machine well before it starts to wear out. The bigger the SSD and the more free space available, the longer its lifespan. This is why Intel's SSDs are such odd capacities, they leave an amount of capacity unaddressable so the wear-leveling programming will always have free space to move data around.

 

[Loa]

All mechanical devices have a failure rate and having your data depend on the integrity of multiple mechanical drives increases that risk exponentially with every drive you add.

Remove mechanical from your sentence and you'll be even more right. Every device has a failure rate.

And when you're using SSDs, since they have very low capacities, you'll be forced to add more then by using HDDs... 🙂


Also, you use a RAID for performance and reliability, not storage or back-up... Most people that have RAIDs for performance only use the fastest part of their drives, and thus plan to have 2 or 3 times the capacity they actually need. As Tesselator said: you don't need to have a back-up as big as your RAID, simply as big as you need it to be.

You "spec" your RAID for performance, and "spec" your back-up for dependability.

Loa

 

[300D]

And when you're using SSDs, since they have very low capacities, you'll be forced to add more then by using HDDs.

Thats why people use an SSD (in desktops) for applications, frequently accessed data and booting, not general storage. There is no lifespan limit to reading data in an SSD, only writing.

 

[Tesselator]

I don't know, why are you here causing trouble? You have added nothing to the discussion about SSDs, only spreading your biased opinion against them in an effort to persuade people into wasting money on a low-end RAID instead.

I just kicked your ass in every test. I have numbers on my side. All you have is a rather childish attitude.

All mechanical devices have a failure rate and having your data depend on the integrity of multiple mechanical drives increases that risk exponentially with every drive you add.

Again not true at all. Haven't you been reading along in the RAID discussions here? Or how about manufacturer data. It also claims you're wrong.

There is no lifespan limit to reading data in an SSD, only writing.

Bzzzt, wrong again. There is a lifespan even to it's physical structure (shelf life). There is for reads, and there is for power-on (start/stop) cycles too.

 

[300D]

I just kicked your ass in every test. I have numbers on my side.

HA! Delusion and lack of experience is all you have.

Again not true at all.

I agree, what you said was not true at all.

Bzzzt, wrong again.

Yep, you are wrong again. What was that you were mumbling on about earlier? "Haven't you been reading along in the RAID discussions here? Or how about manufacturer data."

There is a lifespan to everything, but that is well beyond when it becomes impractical for normal use. I still have 20MB and 40MB hard drives from Mac Classics and other computers that still work very well. Remember how dismal their expected lifespan was back then? By your logic those drives should have died long ago!

 

[Tesselator]

I'm not sure your rubber & glue tactics are having the desired affects. 😀

I'll stick with the numbers and the manufacturer claims. 😉

 

[nanofrog]

There is no lifespan limit to reading data in an SSD...

Unfortunately, this is incorrect. It should be listed in the specs somewhere. It is for Intel's X-25 models. So far, they all use either MLC or SLC flash. Which translates to ALL SSD's have both read and write limits. Read is just quite a bit higher with current models.

For now, SSD's are a bit immature for write usage due to the limitations of the technology it's using. Manufacturer's are well aware of the limitations, as well as their component suppliers. Alternatives are already being developed, but nothing's hit the market yet. 🙁

Like it or not, the current mechanical drives actually have a higher reliablilty for writes, and are on par for reads. Despite the fact it's done with mechanical components. It's been around for a lot longer, and has had multiple refinements/improvements through the years. Three decades worth. 😱 😉

 

[300D]

I'm not sure your rubber & glue tactics are having the desired affects.

Are you sure? You are embarrassing yourself pretty badly. You still have ZERO experience with an SSD yet you still claim a RAID is better and you still know nothing.

I'll stick with the numbers and the manufacturer claims.

I have yet to see you do anything of the sort.

Unfortunately, this is incorrect.

Fortunately, its not.

Like it or not, the current mechanical drives actually have a higher reliablilty for writes, and are on par for reads. Despite the fact it's done with mechanical components. It's been around for a lot longer, and has had multiple refinements/improvements through the years. Three decades worth.

Yep, remember that two years ago SSDs were little more than a $1000+ fantasy for most people? If manufacturers would stop using JMicron's junk controller, things would be far better sooner.

 

[thepawn]

Some numbers:

http://www.storagesearch.com/ssdmyths-endurance.html
http://www.tomshardware.com/reviews/Intel-x25-m-SSD,2012-5.html
http://www.embeddedstar.com/weblog/2007/05/21/ssd-longevity/
http://www.esacademy.com/faq/docs/flash/lifetime.htm (published in 2000!)

The new ram buffers in drives help significantly reduce wear. My OCZ Vertex drive has 64mb of buffer.

Guy who actually worked to GET a flash (thumbdrive) to write fail:
http://www.bress.net/blog/archives/114-How-Long-Does-a-Flash-Drive-Last.html

With software implementations like SteadyState, it helps preserve the flash even longer. Longevity is still very much a use process - so pick your SSD smartly and you are in for an enjoyable ride. Intel, Samsung or Indilink controllers, get one with a large buffer, implementing some effective software caching and using quality chips and you are much safer...SSD aware OS wouldn't hurt. Pay $20 for your 120gb SSD drive and get $20 results.

 

[iamcheerful]

thepawn, nice links. thanks for the easy info. saves us googling 😉

 

[nanofrog]

Yep, remember that two years ago SSDs were little more than a $1000+ fantasy for most people? If manufacturers would stop using JMicron's junk controller, things would be far better sooner.

Still too expensive for the capacity yet, so most people will want to wait. I'd like to get them into a RAID, but is way out of my budget. 😉

It would also be nice if the flash would be changed over for another form that's capable of higher write cycles. Hopefully, both will happen simultaneously.

Some numbers:

http://www.storagesearch.com/ssdmyths-endurance.html
http://www.tomshardware.com/reviews/Intel-x25-m-SSD,2012-5.html
http://www.embeddedstar.com/weblog/2007/05/21/ssd-longevity/
http://www.esacademy.com/faq/docs/flash/lifetime.htm (published in 2000!)

The new ram buffers in drives help significantly reduce wear. My OCZ Vertex drive has 64mb of buffer.

Guy who actually worked to GET a flash (thumbdrive) to write fail:
http://www.bress.net/blog/archives/114-How-Long-Does-a-Flash-Drive-Last.html

With software implementations like SteadyState, it helps preserve the flash even longer. Longevity is still very much a use process - so pick your SSD smartly and you are in for an enjoyable ride. Intel, Samsung or Indilink controllers, get one with a large buffer, implementing some effective software caching and using quality chips and you are much safer...SSD aware OS wouldn't hurt. Pay $20 for your 120gb SSD drive and get $20 results.

Nice links. 🙂

For me, I think in terms of worst case. Manufacturers don't list the chips they're using, how they tested their hardware, and certainly not the grade (if they actually do this). So for me to be able to use such disks with current flash, they'd need to be larger.

This IMO, is the real issue with SSD in enterprise environments as it exists now. For consumer use, it's acceptable, depending on usage patterns. Though I don't exactly like the methodology. (Consumer is unaware of this from the published data for most manufacturers). Simply put, I want to know how they derive a 1-5 million write cycle on the chips they use (i.e. indication of percentile based write endurance), and haven't seen this.

Then there's the cost... 😱 😛

 

[thepawn]

Yeah, I had all the links bookmarked because from researching my OCZ Vertex for my laptop, was easy to share... 🙂

At the bank I do systems architecture for we rarely use and SSD in enterprise systems. We have the occasional Texas Memory drive for DB caching enhancement or our systems themselves are design to suck up memory to cache themselves. Everything is dual-port 15k SAS for DASD or straight up SAN (EMC or Hitachi usually).

The place I found chip types and such was from Toms or Anand, etc, where they were willing to tear everything apart and read the print on the silicon. 🙂

It is still a rather "raw" market, you do really need to research what you buy or you can get pretty burned...they are definitely not all created equal.

 

[Tesselator]

Are you sure?

Yes! I'm also pretty sure that you're the one with egg on your face here. Jumping up and down like an overbred terrier on speed with no evidence or test results of course would make anyone look foolish.

I showed you the numbers very clearly here: https://forums.macrumors.com/posts/7829393/ and here: https://forums.macrumors.com/posts/7829635/ . Additionally I count 3 people in this thread with a reputation of having a high technical aptitude telling you that you're wrong and saying to paraphrase one user that your subjective, unmeasured observation is BS compared to an objective, spec-for-spec comparison.

Allow me to clarify one last time. Yes SSD is fast and makes a huge difference in overall system performance. And 3 or 4 drive RAID0 with 500GB platter drives offers the same relative speed advantage. Faster at some things (most actually) and a little slower at others (very few actually). The difference becomes shockingly one-sided when you consider that in SSD $350 gets you 80GB but in a RAID0 that same $350 will get you 4.5 terabytes. To acquire even 1TB worth of SSD media one would need to spend nearly $4,000 USD. Even with the slower cheaper 250GB SSDs one would need to spend nearly $3,000 USD. And as my benchmarks clearly show you would still be slower at most things than 4.5 terabytes costing one tenth as much. I spent $291 USD on mine. A fourth drive would bring it up to 6 terabytes and would offer even greater speeds for a mere $400 total.

So in conclusion I re-stress my original point that SSD is certainly not by any stretch of the imagination (no matter how active YOURS may be) the best upgrade one can perform on their MacPro.

For a laptop I think it is actually. But there's no way it is for a MacPro. For MP or any desktop with open drive bays it's actually one of the worst.

.

 

[VirtualRain]

So in conclusion I re-stress my original point that SSD is certainly not by any stretch of the imagination (no matter how active YOURS may be) the best upgrade one can perform on their MacPro.

For a laptop I think it is actually. But there's no way it is for a MacPro. For MP or any desktop with open drive bays it's actually one of the worst.

.

Tessellator, I've a high level of respect for your knowledge and contributions here but I'm having difficulty with your comments in this thread. (BTW, you have a PM)

What do you mean by best and worst here?

How, in your eyes, is an SSD the worst upgrade you can make for a Mac pro?

 

[Tesselator]

Tessellator, I've a high level of respect for your knowledge and contributions here but I'm having difficulty with your comments in this thread. (BTW, you have a PM)

What do you mean by best and worst here?

How, in your eyes, is an SSD the worst upgrade you can make for a Mac pro?

No PM is showing up. Try again maybe?

Well "best" would be a $5 part that increases performance 100x. 😀
Worst or "one of the worst" would be $3,000 worth of storage that has the same relative speed advantage as a $300 solution and is only one sixth the size. Making it in total 600 times worse - or 600 times more silly. 🙂

If you require JBOD then I guess there's no other option but otherwise there's no rationale for SSD that I can see in a desktop machine. When 250 GB SSDs are $70 instead of $700 as they are now then I'll of course be forced to reconsider my position.

 

[VirtualRain]

No PM is showing up. Try again maybe?

Well "best" would be a $5 part that increases performance 100x. 😀
Worst or "one of the worst" would be $3,000 worth of storage that has the same relative speed advantage as a $300 solution and is only one sixth the size. Making it in total 600 times worse - or 600 times more silly. 🙂

If you require JBOD then I guess there's no other option but otherwise there's no rationale for SSD that I can see in a desktop machine. When 250 GB SSDs are $70 instead of $700 as they are now then I'll of course be forced to reconsider my position.

I see, well, I define "Best" as the upgrade which provides the most noticeable performance improvement compared to similarly priced alternative options.

There are no other upgrades I'm aware of for around $300 that can equal the impact on performance I've witnessed with nearly every computing task I do. You don't need to spend $3K to enjoy the benefits. Even moving just your OS/Apps to SSD will have a dramatic impact.

 

[Tesselator]

I see, well, I define "Best" as the upgrade which provides the most noticeable performance improvement compared to similarly priced alternative options.

Yes that's the same thing I'm saying.

There are no other upgrades I'm aware of for around $300 that can equal the impact on performance I've witnessed with nearly every computing task I do. You don't need to spend $3K to enjoy the benefits. Even moving just your OS/Apps to SSD will have a dramatic impact.

So you're saying RAID0 isn't an option for you then? Because it out performs SSD and is the same $300. It offers the same "impact" as you put it on apps and nearly as much on the OS.

And if you're like me then 80GB is not enough for OS and apps. You would need at least two bringing it up to $600 (and aren't they really $350 each? So that would make it $700 for 2). Actually I would need 300GB and even that would be running at near 100% capacity not leaving room for any sizable project data. For me to be able to work with only two 80Gig drives I would have to keep way too much data segregated on slow or slower-external space. Then what's the point of the fast drive? Just boot and app loading? So $700 to save 30 seconds to 1 minute a day? Hmmm...

Nah, not for me. I want fast, and reasonably priced, and large enough to put everything on so that the total system really is speedier. For that I need a minimum of 1TB. And that's if I keep my photo, video, and music libraries off.

 

[chowmein]

Yes that's the same thing I'm saying.



So you're saying RAID0 isn't an option for you then? Because it out performs SSD and is the same $300. It offers the same "impact" as you put it on apps and nearly as much on the OS.

And if you're like me then 80GB is not enough for OS and apps. You would need at least two bringing it up to $600 (and aren't they really $350 each? So that would make it $700 for 2). Actually I would need 300GB and even that would be running at near 100% capacity not leaving room for any sizable project data. For me to be able to work with only two 80Gig drives I would have to keep way too much data segregated on slow or slower-external space. Then what's the point of the fast drive? Just boot and app loading? So $700 to save 30 seconds to 1 minute a day? Hmmm...

Nah, not for me. I want fast, and reasonably priced, and large enough to put everything on so that the total system really is speedier. For that I need a minimum of 1TB. And that's if I keep my photo, video, and music libraries off.

RAID 0 alone wouldn't be an option for me, either. I understand where you're coming from with the whole "SSDs are much more expensive" thing but making so many drives dependent on each other (for example, three, like yourself) is really a risk that some people would really not want to take. I'd rather shell out a little extra for a RAID 01 to keep all my data secure and not have to go out buying an extra drive for backup. (double win! 😀) Assuming that a hard drive has a 5% chance of failing within three years, a two-drive RAID 0 would have a 9.5% chance of failing within three years. Think of how that number would be with three or more drives! 😱 In a RAID 1 however, both drives would have to fail in order for the whole RAID to fail. Assuming the same thing, a two-drive RAID 1 would have a 0.25% chance of failing within three years. Given the modest performance boosts that you already gain with RAID 1, a RAID 01 would probably a better choice than just a RAID 0.

(btw, all calculations taken from Wikipedia article "Standard RAID levels")

 

[nanofrog]

...I define "Best" as the upgrade which provides the most noticeable performance improvement compared to similarly priced alternative options.

I would think most would. 😉 😀

There are no other upgrades I'm aware of for around $300 that can equal the impact on performance I've witnessed with nearly every computing task I do. You don't need to spend $3K to enjoy the benefits. Even moving just your OS/Apps to SSD will have a dramatic impact.

It all comes down to usage patterns, which includes needs to me.

So if you don't need a lot of capacity, and want really fast random read performance in particular, you can get a lot of performance for $300 - $350 for a single drive. Unfortunately, not everyone has the same needs. So other solutions are better alternatives.

In Tesselator's case, boot and app load time takes a back seat, and rightly so. 😉 Even less of a consideration if the same program is running for hours or even days. 😛

So you're saying RAID0 isn't an option for you then? Because it out performs SSD and is the same $300. It offers the same "impact" as you put it on apps and nearly as much on the OS.

SSD would have an advantage for random access, but mechanical drives in RAID can easily exceed the sustained throughput, and certainly capacity of a single SSD. 😀

All in an individuals requirements. So in your case, the $300 - $350 is better spent on say 4 disks in a stripe. Fast for sequential speeds, which happens to fit your specific usage quite well, and in the case of the Samsung Eco2's (1.5TB), 6TB is rather welcome. 😉 😛

Both POV's have their merits, and both are a compromise to suit an intended purpose. 🙂

 

[Tesselator]

(btw, all calculations taken from Wikipedia article "Standard RAID levels")

Yeah just so long as you're understanding that the chances assume it even happens at all. The chances only apply to one in 10,000 users or so. So it it's a 5% chance of a one in 10,000 occurrence. Not 5% as in 5 out of 100 users or 10% as in 1 in ten. See? Basically we're talking about the SAME near nil numbers as an SSD failing, or your CPU popping. And if even half of the drives that failed in their compiled stats did so from power surges, poor or failing PSU, and most of the reasons HDDs fail, etc. then this applies to SSDs as well. Not to mention that there's actually less wear and tear on a RAID set, and that RAID users would likely be of the more consciences sort. We've been over this already many times here. Adding drives to a RAID does not increase the chances of failure in any meaningful way and in fact may even reduce the chances of failure. If the MTBF of one drive is is 1,500,000 hours for example then the MTBF of 10 drives in a RAID is still 1,500,000.

Neither view whichever is more correct or common sensical, eliminates the need for a backup which solves the problem completely and really renders this entire sub-conversation moot.