• 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.

Anand Article: Seagate's Momentus XT Reviewed, Finally a Good Hybrid HDD

Idontcare

Elite Member
http://www.anandtech.com/show/3734/seagates-momentus-xt-review-finally-a-good-hybrid-hdd

On the topic of hybrid drives...in my naivety I just assumed that when these things finally came to market the arbitrating logic would be to intentionally house the small files on the flash and the large files on the platters.

Since the lower latency in access time provided by flash is beneficial in boosting the IOPs of small files, and for all the obvious math reasons they don't do so much to boost large file IOPs, it just seemed like the obvious implementation.

Something along the lines of "if filesize <= 64KB then store in Flash, else store file on platter".

But it doesn't look like that is the approach Seagate has taken with their arbiter on their hybrid drive.

So what is it that I'm missing in my naive view of how a hybrid drive ought to function?

Personally I'd rather have 4GB worth of 512B-64KB files stored on the SLC chip while my 1GB ripped VOB files stayed on the platters where I could care less if the access time to the 1GB is 100us or 10ms.
 
I think how it works is by scoring files for "cacheability" and size, e.g.

If file size < 64KB
AND accesses > 100
Then cache

And "accesses" is updated every N hours or something.
 
Forgive me, I just quickly browsed the story but didn't see anything about this, is this drive OS specific for any reason? Will it work as well for OS X?

-KeithP
 
But it doesn't look like that is the approach Seagate has taken with their arbiter on their hybrid drive.
The NAND is so small, there is no spare area, there are almost limitless small files in a system (OS plus apps), and some of them change often enough. Given this scenario, perhaps it would not be beneficial in the long run (small capacity and no spare area might make it hard to wear-level, for one), especially if the small NAND is always almost full.

They probably figured that out in their test labs, and had to design for a way to only let in the files that are accessed often enough.

Or perhaps the more performance-oriented solution would have been to actually increase NAND capacity, but you have to start somewhere and this is a good enough first step.

I failed to catch the price as I skimmed the article and took note of just the charts. As long as it is cheaper than the velociraptor, then I believe it is good.
 
Forgive me, I just quickly browsed the story but didn't see anything about this, is this drive OS specific for any reason? Will it work as well for OS X?

-KeithP

To my understanding there isn't anything about this in the article which is why I brought it up for discussion here in the forum.

They have some smart people working on these things and I am compelled to believe they know better than I do as to why they configured the caching scheme as they did, but what is wrong with a caching scheme that applies more bias to the small files (and keeps them persistent)?

There must be a fatal flaw with the logic I laid out given that it was not the logic employed by the arbiter...but what is the flaw?
 
just picked up two after reading this article.

almost as fast as an ssd and cheaper to boot. no brainer for me.

already own two ssd, one intel and one ocz. but if this thing can compete for less $$$. I'm sold.
 
I think there is another thing that would have been interesting to go in the review: total energy consumption - which is important for laptop drives.

The cache has a benefit, as once a system has booted, only a few hundred MB of files typically need accessing during use (unless you are doing high-end work - but office, surfing, etc. aren't). With a large flash cache, with time, the drive could bring all those files into flash - therefore allowing the platters to spin down for longer periods.

They have some smart people working on these things and I am compelled to believe they know better than I do as to why they configured the caching scheme as they did, but what is wrong with a caching scheme that applies more bias to the small files (and keeps them persistent)?

The drive has no way of knowing the size of a file, and there is no sensible way for it to guess either (especially with modern file handling techniques such as memory mapping of large files, etc.)

As Jimhsu said, this sort of thing is best handled at the file system level - the FS will know how big files are, how frequently they are accessed, how frequently they are changed - and (although there aren't at present) in the future, there may be ways for apps to tell the FS how a file should be cached (e.g. normally an OS will cache files in RAM for the next time they are accessed - but media files are rarely accessed again, they are usually streamed in order. Ideally, a media player app could tell the OS that it won't need the data again, and would prefer a reliable stream - so the OS would discard data once it had gone to the app, but would fetch data from HD ahead of the app, so that the app doesn't lag out).

Advanced file systems like ZFS already have an SSD file cache function. If you've got a file server, with 10 TB in a RAID array, you can attach a 80 or 160 GB SSD, and use it as a read cache - the most frequently used files will get cached, dramatically increasing performance. Similarly ZFS, can also use SSDs as a write cache - data waiting in RAM to be written to HD, will be backed up on SSD - so if the power goes off, or there is a system crash, all that data is safe (if the SSD goes out, the data is still in RAM, and once it gets saved to HD the RAM cache will be disabled until the SSD is replaced)
 
To my understanding there isn't anything about this in the article which is why I brought it up for discussion here in the forum.

They have some smart people working on these things and I am compelled to believe they know better than I do as to why they configured the caching scheme as they did, but what is wrong with a caching scheme that applies more bias to the small files (and keeps them persistent)?

There must be a fatal flaw with the logic I laid out given that it was not the logic employed by the arbiter...but what is the flaw?

It is my understanding that basic hard drives (such as Seagate drive mentioned) have no inherent knowledge of "files". They basically and simply handle the data transfer of sectors/blocks as addressed by means of LBAs (logical block addresses). And they have no knowledge that a particular sector/block (or some group of sectors) represents a particular "file".

Of course, they can try to infer that the LBAs presented to them represent a "file", but this likely entails a lot of assumptions.
 
So what is it that I'm missing in my naive view of how a hybrid drive ought to function?

It seems that "hybrid" drives are in their infancy, so IMO there is no "ought to." Maybe months or years in the future all the manufacturers will settle on a particular "tried and true" method. Right now all we can go on is... does it blend? Er, I mean, does it work?

Apparently it does work, and works well enough to be a viable alternative to "slow" hard drives and "small/expensive" SSDs.

Think about this... the quick Windows booting and "snappy" feeling of SSDs at less than 1/10 the cost/GB. Heck, in many cases this is better than a VelociRaptor for less than 1/2 the cost/GB. I can't see this as anything but win.

Wait, it's made by Seagate. :sneaky:

That's what backups are for! ^_^

Now, knowing how well a notebook HDD does with just 4GB cache, imagine what WD can do if it added say 16GB/4ch SLC to a VelociRaptor 600GB without drastically increasing price? 😱
 
I might be getting this drive. It's getting good reviews. The 320gb size is a good buy because of it's reliability (maybe only 1 platter?), even though the dual 250gb platter 500gb size driver are slightly faster. Only thing that is a small concern is the reliability behind the seagate brand after all the reported failures recently.

I think i will still get this. My 3 options being considered were:
$70 Western Digital Caviar Black 320GB 7200BEKT
$110 Intel X25-V Value series SSD 40GB
$120 Seagate Momentus XT 320GB Hybrid SSD/HDD
 
Ultimately, this is not a replacement for a desktop SSD, but is an obvious improvement for notebooks in particular many of which come with only one drive slot. That is all.
 
The drive has no way of knowing the size of a file, and there is no sensible way for it to guess either (especially with modern file handling techniques such as memory mapping of large files, etc.)

It is my understanding that basic hard drives (such as Seagate drive mentioned) have no inherent knowledge of "files". They basically and simply handle the data transfer of sectors/blocks as addressed by means of LBAs (logical block addresses). And they have no knowledge that a particular sector/block (or some group of sectors) represents a particular "file".

See that just goes to show you there is no limit to the number of things I am clueless about. I had zero notion of this now very obvious factoid!

OK, I think I've reached my "doh!" quota for the day D:...thank you gentlemen for not smacking me with a wet fish too hard in regards to my ignorance of existence of file-system management...yeah don't ask me how it had completely escaped my scrutiny that different file systems exist for a reason and the drive is file-system agnostic!

...and I can't even claim I was drunk-posting 🙁
 
I like how you can get them in 320 & 500 GB flavors for ~ $100-150. Newegg has the 500GB XT up now for $129 ETA 5/27. From real world benches it appears to bridge the gap between the speed of a Velociraptor/Cheetah and an average popular SSD. But in the synthetics it is always slowest/last place. Good thing I don't care about synthetic HDD benches. If it loads apps and boots up faster than the fastest HDDs, thats good enough (even though it still isn't as quick as the SSDs)

The major drawback is the "SEAGATE" brand. I've been checking some of the Newegg reviews for Seagate 2.5" HDDs and recently they are horrible. ~35&#37; failure rate reported in the 500GB drives. Compared to the WD Caviar Black which has an 83% 5-egg rating, the Seagate drives in the same category have a 55-59% 5-egg rating.

I hope Western Digital makes one of these 7200rpm Hybrid SSD/HDD drives.
 
I like how you can get them in 320 & 500 GB flavors for ~ $100-150. Newegg has the 500GB XT up now for $129 ETA 5/27. From real world benches it appears to bridge the gap between the speed of a Velociraptor/Cheetah and an average popular SSD. But in the synthetics it is always slowest/last place. Good thing I don't care about synthetic HDD benches. If it loads apps and boots up faster than the fastest HDDs, thats good enough (even though it still isn't as quick as the SSDs)

The major drawback is the "SEAGATE" brand. I've been checking some of the Newegg reviews for Seagate 2.5" HDDs and recently they are horrible. ~35% failure rate reported in the 500GB drives. Compared to the WD Caviar Black which has an 83% 5-egg rating, the Seagate drives in the same category have a 55-59% 5-egg rating.

I hope Western Digital makes one of these 7200rpm Hybrid SSD/HDD drives.


The seagate brand got hammered with their first gen 1TB drives but for the most part are as reliable as the next brand. I did stop buying them about a year ago because the caviar black and green series were just too damn good.
 
One thing I'm not quite sure after reading the article: How did Anand test the RL performance (or his whole test suite for that)? He could either run the whole testsuite several times and measure performance in the last run or he could run one benchmark several times and use the last numbers.

I'd say that the first one would be much more interesting to see how the drive works in general (with that tiny 4gb flash I don't think it can cache lots of files, even small ones).
Running the RL benchmark in that fashion would give some interesting insights (obviously no IOmeter or similar things, since they probably destroy the performance if those get cached..)
 
I have a 300GB 2.5" VelociRaptor as my storage drive. My primary drive is an Intel 160GB G2. I had the VelociRaptor first and when I got the Intel, I chose to use the VelociRaptor as a storage drive (personal documents, audio / video files, backups).

Would it be worth switching to the 500GB Momentus XT as a storage drive if I can sell the 300GB VelociRaptor for at least as much as a 500GB Momentus XT?
 
I have a 300GB 2.5" VelociRaptor as my storage drive. My primary drive is an Intel 160GB G2. I had the VelociRaptor first and when I got the Intel, I chose to use the VelociRaptor as a storage drive (personal documents, audio / video files, backups).

Would it be worth switching to the 500GB Momentus XT as a storage drive if I can sell the 300GB VelociRaptor for at least as much as a 500GB Momentus XT?

Why would you want the momentus as a storage drive? Just use the raptor.
 
Ok, that's what I thought. So the VelociRaptor's 10000rpm spindle speed is still better for a drive that's going to be used for storage.

Does it matter that I move my "My Documents" folder onto the storage drive?
 
So the VelociRaptor's 10000rpm spindle speed is still better for a drive that's going to be used for storage.

Depends on what your needs are for storage. If it is just a bunch of documents, music or videos then just get some 5400RPM "green" drive. If it is storing huge Photoshop files and you do this for a living, then a VelociRaptor would be great.
 
so if i'm using an SSD for boot, a smaller SSD for readyboost, and a 640 GB caviar black for files (passmark says there aren't many drives that are faster), is something like this going to improve my experience?
 
Back
Top