In The Lab With Seagate's Momentus XT 750 GB Hybrid HDD

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Anything that spins is inherently less reliable than anything that doesn't. HDD performance is more than adequate for nearly all computer tasks. I see this product as a solution in search of a problem.
 
[citation][nom]Farrwalker[/nom]With the relatively low price of 4GB 240-Pin SDRAM DDR3 1600 four sticks can be had for under $100. With a system that has 16GB of RAM, is it possible for software to accomplice what a hybrid hard drive does?Assuming that a system has 16GB of RAM and a 1TB hard drive, is there software to use the RAM in the motherboard as a hybrid uses the flash memory contained with in the hybrid drive to increase read and writes?[/citation]
Memory is volatile. Computer loses power, the cache is gone. Makes it a really poor solution for caching.

You can, however, create RAM disks if you really want to. In practice they are rarely used for anything you want to keep around for more than a couple hours, though, because of the volatility. You can set them up to save on shutdown and restore on load, but this relies on always having a clean shutdown, and you've got to wait for the RAM disk to write out its entire contents to your slow HDD on shutdown, and wait again to read from your slow HDD on startup.

RAM works very well for things you don't need to be persistent, and very poorly for things that need persistence.
 
[citation][nom]willard[/nom]No they aren't, you can only add probabilities when they are mutually exclusive.[/citation]
Incorrect. If they are not mutually exclusive, you just have to use a different formula:
P (A or B) = P (A) + P (B) - P (A and B)
this is equivalent to
P (A or B) + P (A and B) = P (A) + P (B)
So the combined probability of either one of the drives, or both of them, failing, is equal to the sum of the probability of each of them failing individually.

You can find the formula at http://en.wikipedia.org/wiki/Probability#Not_mutually_exclusive
 
I've been dual-booting with a 120GB Intel 320 and an old Raptor 3 drive RAID0 for quite a while. I am finally tired of listening to those noisy drives, so I dropped in a 1 TB HDD, and can definitely feel (subjectively) the slowdown. Something has happened to my WIN7 install o the SSD, so I think I'm going to set it up as a caching drive and see how it feels.
 
[citation][nom]Sakkura[/nom]Incorrect. If they are not mutually exclusive, you just have to use a different formula😛 (A or B) = P (A) + P (B) - P (A and B)this is equivalent toP (A or B) + P (A and B) = P (A) + P (B)So the combined probability of either one of the drives, or both of them, failing, is equal to the sum of the probability of each of them failing individually.[/citation]
I'm aware of the basics of statistics, but that's still not how this works. It works in large sample sizes due to the law of averages (for example, by adding up hours to compare to MTBF and get number of drives failed in a time period), but not on individual drives.

Refer to my example of a drive with five components. 20% failure rate for five components does not equal a 100% failure rate as a whole.

[Edit] Clarifying the math to put this to rest.

Or, think of it this way. If there's a 50% chance of the HDD component failing after a week, and a 50% chance of the SSD component failing after a week, do you really think 100% of drives are going to have failed after a week? Because if you simply add the probabilities together, that's what you get.

Here's how the math actually works.

We've got the same hypothetical drive with two components, 50% failure rate of either component after a week. We want to know the number of drives which have not failed after one week to get the overall probability of failure.

A drive can fail because of either A or B failing. After one week, half of the drives should have failed because of A. Now, we need to know the percentage of drives that fail because of B given that A has not failed. This is known as conditional probability, and is a totally different formula than the one you listed.

The equation is P(B|A) == P(A and B) / P(A).

The probability of A and B is easy since the events are independent, you just multiply them together. The probability of A is also easy, since we know it. That gives us the probability of B given A == 0.25 / 0.5, or 50%. So, we expect 50% of the drives which did not fail as a result of A to fail as a result of B.

Since we know that 50% of the drives did not fail because of A, we can multiply that fraction again by 50% to get the number of drives that did not fail as a result of A but failed as a result of B, or 25% of the whole

Now you can add the probabilities together, for 75% of the total drives failed.

Like I said, you can't just add the probabilities of failure together to get an overall probability of failure.
 
I have been considering upgrading a PC running 4 Seagate 500GB in a RAID 5 to this hybrid drive - would the benefits of the hybrid still translate to the RAID, or would the drive usage in the RAID configuration negate the benefits of the SSD cache?
 

People have claimed that they work very well in RAID.
 
After reading this article, I am wondering if there is a way to exclude certain file types or even directories from being cached? Let's say I have 20GB of legal (yes, I went there) music & video files. It doesn't seem worth caching these as they are low bitrate for what an HDD can handle, but could potentially remove files worth caching. I didn't know if there was a way to manually do this or if the HHDD is smart enough to not do this, or maybe it just doesn't matter?
 
I don't know how we got into this stats discussion. The formula "reduces" to adding the probabilities for small probabilities.

P(HDD fail) = 0.1%. P(SSD fail) = 0.1%.
P(HDD fail) || P(SSD fail) given P(HDD not fail) = 0.1% + 0.999*0.1% = 0.1999%.

willard, are you happy that it's not exactly 0.2%?
 
After reading article, it makes me think that comparing this drive to a fairly old raptor drive really isn't a particularly realistic comparison especially given the capacity. I think a comparison to similar sized drives would have been better with the inclusion of various drives ranging in size from 600GB to 2TB. Western Digital now has 250GB, 500GB, and 1TB VelociRaptor drives which I think would have been a better comparison. I also think comparing them to various high performance SAS drives such as the 15,000 rpm ones from Hitachi or Seagate would give a detailed comparison to the different technologies and the speeds you can expect or observe from those drives. I realize that people are going to say that most don't have those enterprise/workstation drives in mind when buying hard drives, but I think they should at least be reported for comparison.
 
[citation][nom]jsowoc[/nom]I don't know how we got into this stats discussion. The formula "reduces" to adding the probabilities for small probabilities.P(HDD fail) = 0.1%. P(SSD fail) = 0.1%.P(HDD fail) || P(SSD fail) given P(HDD not fail) = 0.1% + 0.999*0.1% = 0.1999%.willard, are you happy that it's not exactly 0.2%?[/citation]
*Disclaimer* I just learned that I'm receiving a promotion, so I'm celebrating and a little drunk. */Disclaimer*

Sorry if I offended you, but I'm a stickler for correctness, and somebody telling me I'm wrong and citing math that not only do they not understand but is patently false to anyone who's had even the most basic statistics course bothers me.

As for how we got into a statistics discussion, this is why.

When you're dealing with an either/or scenario, probabilities are actually additive; P (drive 1 or 2 failing) + P (drive 1 and 2 failing) = P (drive 1 failing) + P (drive 2 failing).
Click to expand...

The fact that the end result is negligibly close with small probabilities doesn't make this statement any more correct. Furthermore, he backtracked later and posted another formula that was also incorrect, at the same time flatly ignoring a disproof by contradiction in the post he responded to.

I can be a bit abrasive at times, but my rigid adherence to facts and correctness makes me invaluable in my field (software engineering). I've stepped on more than a few toes in the workplace by calling people out for poor design practices, but ultimately the product improves as a result. Professionals don't hold grudges, and anybody whose ego I wound quickly gets over it in the interest of simply getting on with things.

I do damn good work, and despite my occasional interpersonal issues, I am a highly valued member of the team and have never failed to beat the average annual raise and bonus. Despite what you may expect from such behavior, I'm actually good friends with a lot of my coworkers. We're all adults, and understand that matters of correctness aren't personal. When I'm wrong (like I was in this thread about Sakkura's use of orders of magnitude) I simply politely admit my wrongness and get on with things. To me, it's not a big deal.

On the internet I might be an asshole, but in college I was the guy who could fix your project the night before it was due. In the workplace I'm the guy they turn to when the hard requirements come in. My mindset allows me to pick up new programming languages in a matter of hours for use in professional development environments, and I've done precisely that on more than one occasion. I spend the vast majority of my time on Tom's Hardware helping people with their computer problems (feel free to look through my thousands of posts). In general I'm a very nice guy, but I'm not one to shy away from confrontation either.

Not everyone likes me, but I'm not so vain that I treat life as a popularity contest. My friends like me, even if not everyone at work does, but I'll be damned if my coworkers don't respect me. To me, the respect of my peers is more important than their friendship.

It's easy to be everyone's friend. It's much harder to earn their respect.
 
[citation][nom]hunshiki[/nom]The idea is great in my opinion, but they could include a 16gb SSD inside the drive. Or 32.[/citation]
But then the price for the drive would go up and less would sell. Would be nice for that option though. It's like electric cars, like the Plug-in Prius. Right now it only goes for 12 miles or so on battery power, if it came with a larger battery pack it would have longer range but would cost a lot more. So they take it easy and get people to accept it one step at a time.
 
Thanks for all the feedback. What I tried to raise for discussion in the article was, “how fast does your storage solution need to be” and “which storage solution will best meet those requirements, taking into account the wider spectrum of storage metrics, such as capacity/ power consumption etc.”.

I’ve been using the Momentus XT for around a month now as a boot drive. For the vast majority of tasks I rarely notice a perceivable difference in performance between a SSD and the Momentus XT. (I’ve been using SSD’s exclusively for over three years now, so I’m quite sensitive to differences in real world performance when using a different storage solution). Boot up is consistently fast and the desktop is responsive as soon as it loads. There are of course occasions, such as installing applications, when the Momentus XT is noticeably slower than a SSD.

Regarding reliability I have asked Seagate if they can provide some more insights into this and I will update the article if I can provide more information. I hope I can also update the article on issues such as the impact of defragging the Momentus XT and raid 0 performance.

Cheers,
Richard Hart
TomsHardware.com

PS. I made a (rather bad quality) video of the boot up processes using the Momentus XT. Once the desktop appeared I manually loaded a few apps to give an idea of responsiveness.

https://skydrive.live.com/redir?resid=541C442F789BF84E!757
 
When I bought my Momentus XT 750 in late 2011, it was with the understanding that Seagate would be issuing a firmware upgrade in "Spring 2012" that would enable write caching for these drives. So far - nothing. It's interesting that this whole article got written without any mention of write caching. FWIW, my experience upgrading a two generation old MacBook Pro with this drive has been extremely positive. I just wish Seagate would let us actual owners know what is going on with that promised firmware update.
 
Willard,

You seem to be overlooking the fact taht we are talking about MASSIVE sample sizes and that as you correctly said it works like that with the law of averages.

We are talking MILLIONS of drives. whether they be SSD, HD or Hybrid. I stand by the notion that it is misleading to say that a hybrid drive does not have a higher probability of failure when compared to the other 2.

So it is YOU who is mistaken mister "knows statistics and probability". The fact that I used a single drive as an example was purely illustrative. They probability is additive over very large samples.

You neglect to point out that in your week of 1 device with 5 components you will get some with 2 failures and 3 and 4 and 5.

 
[citation][nom]americanbrian[/nom]Willard,You seem to be overlooking the fact taht we are talking about MASSIVE sample sizes and that as you correctly said it works like that with the law of averages.We are talking MILLIONS of drives. whether they be SSD, HD or Hybrid. I stand by the notion that it is misleading to say that a hybrid drive does not have a higher probability of failure when compared to the other 2. So it is YOU who is mistaken mister "knows statistics and probability". The fact that I used a single drive as an example was purely illustrative. They probability is additive over very large samples. You neglect to point out that in your week of 1 device with 5 components you will get some with 2 failures and 3 and 4 and 5.[/citation]

An enterprise hard drive is probably far more reliable than a consumer SSD. You're making over-bearing statements by assuming that just because there are more points for failure, maybe each one is still of a higher quality than some solid state drives and some hard disk drives. If you were arguing that there are more points for failure, then yes, that is correct. However, arguing that it is inherently less reliable purely because of this is arguing an unknown and you don't have any way of proving otherwise with the information that has been presented so far. You also exclude the fact that should the solid state cache fail, these drives are designed to then operate as regular hard drives instead of failing completely, so the flash is not a reliability problem in anything but keeping full performance on cached data.

So, these hybrid drives could fail more often than other drives, but they might not. They might fail less often than other drives. Not all SSDs or HDDs are equal in reliability. I can guarantee that many SSDs and HDDs over the years have been far less reliable than others. These hybrid drives are probably fairly reliable, although like I said, I can't prove that without more information. If these are well-built, then they could be more reliable than any HDD or SSD. Heck, a well-built HDD can be more reliable than any SSD. Being an HDD does not mean that it must be less reliable than an SSD. Being hybrid between the two does not necessarily mean that these hybrid drives are less reliable than drive models from both the hard disk and the solid state technologies.
 
Best thing I ever did for both my and my dad's laptop... I needed the space the spinning disk affords, and his aging laptop needed the speed...

We now use those (per my recommendation) in our factory built turn key systems. The factory doesn't trust SSDs by themselves yet (older generation thinking), but 2 of these in a RAID1, speeds up the normal launch of our software beyond belief....

and if the client adds anything more... it acts like any other normal 7200rpm HDD.

Just RAID them... every mobo for the last 5 years has it on board! Then the concern of SSD fail or HDD fail is kinda moot.
 
[citation][nom]blazorthon[/nom]An enterprise hard drive is probably far more reliable than a consumer SSD. You're making over-bearing statements by assuming that just because there are more points for failure, maybe each one is still of a higher quality than some solid state drives and some hard disk drives. If you were arguing that there are more points for failure, then yes, that is correct. However, arguing that it is inherently less reliable purely because of this is arguing an unknown and you don't have any way of proving otherwise with the information that has been presented so far. You also exclude the fact that should the solid state cache fail, these drives are designed to then operate as regular hard drives instead of failing completely, so the flash is not a reliability problem in anything but keeping full performance on cached data.So, these hybrid drives could fail more often than other drives, but they might not. They might fail less often than other drives. Not all SSDs or HDDs are equal in reliability. I can guarantee that many SSDs and HDDs over the years have been far less reliable than others. These hybrid drives are probably fairly reliable, although like I said, I can't prove that without more information. If these are well-built, then they could be more reliable than any HDD or SSD. Heck, a well-built HDD can be more reliable than any SSD. Being an HDD does not mean that it must be less reliable than an SSD. Being hybrid between the two does not necessarily mean that these hybrid drives are less reliable than drive models from both the hard disk and the solid state technologies.[/citation]

Blaze, you are dead right... without meaningful metrics... who knows? I've had VERY expensive enterprise SSDs die in within 1 year in my SAN solutions, while other consumer grade SSDs are still chugging along. (it took a bit and a couple client willing trials to get SSDs into my boxes at all, and consumer came first cause they were cheap)

But per my experience, consumer, prosumer, enterprise... it all boils down to turn around... how fast can \ will you get my replacement out to me? I think that's most of what you are paying for.

Yes. americanbrian, the market speak talks about MTBF, and IOPS, and latency... But in my experience, they are all pretty close to the same. I know you talk sample size... but that number is only a comfort if you are on the side of "your drive lasting beyond its MTBF"... if you aren't on that side of the fence... and for me, on multiple occasions, it boils down to other things... FAST!

Market speak doesn't mean jack if it takes a week or more to get a replacement drive... ESPECIALLY in any sort of business environment. You get what you pay for... and I just see the extra cash we pay for enterprise dives going to things like 24-7 REAL person call center, advance replacement, them willing to give me the tracking number of my drive... things like that.

 
[citation][nom]egowhip69[/nom]Blaze, you are dead right... without meaningful metrics... who knows? I've had VERY expensive enterprise SSDs die in within 1 year in my SAN solutions, while other consumer grade SSDs are still chugging along. (it took a bit and a couple client willing trials to get SSDs into my boxes at all, and consumer came first cause they were cheap)But per my experience, consumer, prosumer, enterprise... it all boils down to turn around... how fast can \ will you get my replacement out to me? I think that's most of what you are paying for.Yes. americanbrian, the market speak talks about MTBF, and IOPS, and latency... But in my experience, they are all pretty close to the same. I know you talk sample size... but that number is only a comfort if you are on the side of "your drive lasting beyond its MTBF"... if you aren't on that side of the fence... and for me, on multiple occasions, it boils down to other things... FAST! Market speak doesn't mean jack if it takes a week or more to get a replacement drive... ESPECIALLY in any sort of business environment. You get what you pay for... and I just see the extra cash we pay for enterprise dives going to things like 24-7 REAL person call center, advance replacement, them willing to give me the tracking number of my drive... things like that.[/citation]

To be fair, the Vertex 3 has a higher MTBF on newegg than the Samsung 830 and the Crucial M4, yet those two SSD series tend to be far more reliable than Vertex 3 drives. MTBF seems to be just whatever the company wants you to believe. Also, I've tried several different SSDs and performance can be similar, but it can be radically different. Try comparing a Vertex 4 with the latest firmware to a Vertex 3. Sometimes, it will lose noticeably. Sometimes, it will win by huge margins. Enabling NTFS compression will probably more or less eliminate its losses while increasing its wins a little.

My Vertex 4 is noticeably faster than my previous Samsung 830 in a lot of workloads, but most obviously in writes. However, my Chronous Delux was not noticeably faster than my Vertex 3 in most situations, so sure, they can be similar. it depends on the drive. If they have very different controllers and flash, then they can have very different performance. If they have the same controller and somewhat similar flash, then they can have very similar performance, especially in real-world usage.

If you don't believe me, then you could try comparing a high-performance non-SandForce SSD to a high-performance SandForce SSD in different workloads. If you do anything where low latency or not very compressible data are large factors in performance, then there should be an obvious difference. Each drive can behave differently and not only that, but can even change radically with different firmware.

With SandForce drives that have the same controller, the main performance differences come from the flash interface and the capacity. For example, any second gen SandForce drives with the same capacity and flash memory will perform very similarly, if not identically. I'd be surprised if anyone can tell the difference between them by performance alone. Perhaps you've only had experience with SandForce and that's why you think that there is little difference between different SSDs.
 
Any chance of a review of the OCZ Synapse Cache products? They come in 128GB and 64GB versions (64GB and 32GB usable as cache due to over-provisioning) and work together with any hard drive and motherboard.

I'm using the 64GB version and it's working great. You barely notice you're running a mechanical drive and you don't have to worry about what to install on the HDD and on the SSD since it dynamically caches whatever you access most frequently.

There are obvious drawbacks - it takes up an extra 2.5" slot and SATA connector and requires drivers to function, but for some users, it's the a great solution.
 



I made 2 statements and I stand by them.

No.1 It is LIKELY that these drives will be less reliable...Due to the laws of probability, I believe the spider graph is misleading. Unless they make both the HD and SSD components 2x as reliable as normal. I don't think this is realistic, they will be made the same as current tech allows.

No.2 Probabilty does work like that... (when applied to large sample sizes with which were being discussed)

I do not see what is overbearing about either of these statements.

That is all.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom