Seagate Momentus XT

linjy2

Limp Gawd
Joined
Jun 30, 2005
Messages
423
So who is getting one?

I just bought two. Might even sell my ssd if I can't tell a difference between hybrid and the ssd, Probably throw one in the laptop and another in the HTPC.

Own a intel ssd and a ocz ssd.
 

leh18621

[H]ard|Gawd
Joined
Jun 18, 2008
Messages
1,082
Might be a stupid question but......

In the reviews I have read on this drive they only mention using it in a laptop. Can you use it in a full size desktop pc?
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,925
It's most likely slower than a desktop drive well unless your whole OS can fit in 4GB. But yes it will fit in a desktop with a 2.5inch to 3.5inch adapter or other suitable mounting rails.
 

linjy2

Limp Gawd
Joined
Jun 30, 2005
Messages
423
Might be a stupid question but......

In the reviews I have read on this drive they only mention using it in a laptop. Can you use it in a full size desktop pc?

yes, like all other ssd drives. with a 2.5 -> 3.5 mounting bracket.

It's most likely slower than a desktop drive well unless your whole OS can fit in 4GB. But yes it will fit in a desktop with a 2.5inch to 3.5inch adapter or other suitable mounting rails.

its 500GB hybrid drive with 4GB flash onboard. real world test shows it to be just a tad slower then a ssd from reviews i read. synthetic benchmarks shows it to be faster then a regular desktop drive.
 

white_devil

Limp Gawd
Joined
Apr 21, 2007
Messages
225
The reviews all seem pretty positive, think I'll be picking one up. Was gonna get an FDE drive for the netbook but now think I'll get this and run PGP.
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,925
its 500GB hybrid drive with 4GB flash onboard. real world test shows it to be just a tad slower then a ssd from reviews i read. synthetic benchmarks shows it to be faster then a regular desktop drive.


I know. I still expect it to be slower than a WDC 500GB blue in real world applications once the 4GB flash cache has been used up. I assume it auto balances what stays in the flash based on usage. So the 4GB of most used stuff will be in the flash cache.
 

dccmadams

[H]ard|Gawd
Joined
Nov 25, 2007
Messages
2,010
I saw it tested against a 600 raptor, and outperformed it most of the time, see anandtech site for review.
 

Bahamut

n00b
Joined
Apr 27, 2010
Messages
0
Yah, but, we here at the [H] prefer to see real testing done by our own members on their own hardware more often than not so, as soon as someone gets their drive(s), HDTune/HDTach/CrystalDiskMark/etc 'em and let's see those benches! :p
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,925
I would like to see performance over time. I mean how well does it perform 2 months after installing.
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,925
Maybe I am overestimating the size of most users most used files. I mean if most users only use 4GB of their data often this would be at near SSD speed.
 

Bladestorm

[H]ard|Gawd
Joined
Dec 10, 2006
Messages
1,232
None of the reviews I've seen mention anything about using this as a game drive. Would this drive perform better in gaming vs. a conventional HDD?
 

dccmadams

[H]ard|Gawd
Joined
Nov 25, 2007
Messages
2,010
I saw a review that had 6 apps in the startup folder. It booted up, and had apps ready to go faster than a wd 600 raptor.
 

Trepidati0n

[H]F Junkie
Joined
Oct 26, 2004
Messages
9,193
None of the reviews I've seen mention anything about using this as a game drive. Would this drive perform better in gaming vs. a conventional HDD?

I don't think it could do any worse. If you played the game a lot, it should help in general. Imagine things like FPS with player maps. If that is all you did, eventually those maps should be in the flash and you would load in quicker.

The main reason I want this drive is pretty much for boot time. The faster app load time is nice (a second or two), but not usually the thing that annoys me. Watching windows load is like watching paint dry. At least when loading an app I can alt-tab to something else.

In something like a laptop, this would be pretty useful.
 

sfsilicon

Limp Gawd
Joined
Nov 1, 2005
Messages
505
I've been monitoring the release of this product and from my accessment the key will be the caching algorythm which decides what to keep in the NAND SSD.

Seagate is doing a good job at positioning this product (getting a couple of friendly sites to do benchmarks and pre-release testing) and will do well if they keep pushing the technology as a mobile disk drive replacement. I'm very sceptical of it being a SSD replacement due to the small size of the cache (4GB) and that they are not doing any write caching at all.

The performance benefits you will see really depend on your usage. If you have a few files that you use often and need to load fast and the majority of your other files are used seldom then this will work well for you. You just have to run the application 2-3 times to see the benefit.

All the benchmarks that show they are getting close to SSD performance are a bit suspect without further details on the testing parameters. Even Seagate says run the game 2-3 times and you'll start seeing benefits (e.g. run Game A 2-3 time for it to get accelerated). This works if you run the same game over and over again. What happens if you play 2-3 games and run 2-3 different applications over the week as well as shutdown your PC everyday. Does this push your OS files out of cache after playing 2-3 games? For other applications that require fast writes like video editing or DB applications no write caching will reduce performance to an equivalent 7200 drive.

If the NAND cache were bigger (some where betweeen 32 and 64GB) then this would be a better solution but a much more expensive one as well. I do think this has potential to eventually replace all hard drives. The question is if the HDD mfgs can reduce the incremental cost vs. a HDD to a parity. Right now the 500GB model commands almost a 50% premium ($156 vs $85) All that for 1 4GB SLC chip? I think Seagate has quite a bit of margin in there to make this cheaper and close the gap.

I see folks planning to get 2+ drives for RAID. I wonder if that will improve or worsten the read cache hits since you have no control over what gets stored where so you might have one drive cache well in a RAID 0 scenario and the other doing terribly. It really depends on how much statistical analysis went into the caching algorythms and what kinds of usage scenarios they played with.

Personally I prefer the control and faster interfaces of dedicated SSDs. You can pick which applications need acceleration vs letting an arbitrary algorythm do the work for you. The key benefit of this HDD hybrid will be to the mobile folks that need more storage than a SSD provides (both from a $/GB as well as total capacity). It misses out on one key benefit that a SSD provides which is lower power consumption.

Oh on a side note watch out for when Intel releases their 25nm SSDs this will not only increase capacities by 2x (320-1024GB) but potentially lower the price of a 80GB SSD to <$100. That will make an OS/Programs dedicated drive affordable for a lot of people and dedicated business SSDs for notebooks as well.

Looking forward to seeing your benchmarks guys!

Check out my blog for additional analysis on hybrid HDDs and SSDs.
 

Darakian

Supreme [H]ardness
Joined
Apr 12, 2004
Messages
4,698
I've got one of these on my test box. Anyone want any specific tests done?
 

Darakian

Supreme [H]ardness
Joined
Apr 12, 2004
Messages
4,698
Cool :) How about crystaldiskmark 3.0

Any particular settings?

Edit: Default settings
11072239.png
 
Last edited:

Michaelius

Supreme [H]ardness
Joined
Sep 8, 2003
Messages
4,684
I think that doing syntetic random read benchmarks on those hybrids is pointless.

The drive is intended to learn your usage patters and load most often used files into SLC cells. Therefore randome read test is artificially recreating worst case scenario of randomly accessing data that wasn't preloaded into flash and you can see this on Anand where in random4k drive shows numbers like 5400rpm.

But Seagate you are too late i already got Intel G2 for my Asus UL :D
 

Tau

Limp Gawd
Joined
Mar 3, 2010
Messages
196
I've got one of these on my test box. Anyone want any specific tests done?

Atto please, default settings will be fine,

Also what controller are you running it on?

and an HDTune read/write would be awsome as well. We just got a crate of them in at work here thinking i might liberate one for my laptop :)
 

Michaelius

Supreme [H]ardness
Joined
Sep 8, 2003
Messages
4,684
It would be interesting if you could repeat all the tests after two weeks to see how it behaves after learning.
 

sfsilicon

Limp Gawd
Joined
Nov 1, 2005
Messages
505
When running the tests you need to run them 3x times. From the benchmarks I've seen the algorithm takes that long to recognize if something is used a lot and then it transfers it to the cache.

Are you using it as your boot disk? That would be one of the main advantages of any Solid State booster. If you are try rebooting 3-4 times then take note of the boot time. I'd be interested to see what happens when the boot files are in the cache then you run a couple of big game benchmarks a few times and play a bit. Will the cache loose the OS files and thus you loose the benefit of having solid state? I know with enough stress you can force this, but the question is how close this scenario is to normal usage.
 

Darakian

Supreme [H]ardness
Joined
Apr 12, 2004
Messages
4,698
No I can't compare to an SSD (I don't have an SSD to play with :(), but I am trying to get some "real world" data. I'll post it when I have it on hand.
The controller is an ICH8 in IDE mode so NCQ is not being used.
As for running the benchmark 3 times, note that crystal mark was run 5 times.

Edit:
I ran HDTune on the drive the other day. Max throughput was a little over 100MB/s and the min was around 80MB/s. The latency was curious as it read ~.04ms for all sectors. My guess is that it was hitting the flash and the drive was taking over from there.
 
Last edited:

Tau

Limp Gawd
Joined
Mar 3, 2010
Messages
196
Just had a 2 hour sit down with my Seagate rep.... she was unable to ansear alot of my questions though im sending the engineering department an email here tonight and should have a responce early next week... Any questions anyone wants me to ask? Ill take a list for the next few hours and i can include the results.

Biggest one i have is "How will defragmenting effect the 'frequently' used algorithem" and how big is the "frequently used algorithem"'s memory.... 3 years down the road on a heavily used drive will it crash when the memory fills up?

those are just a couple that I have... and im sure you guys must have a load, so lemme have em.
 

sfsilicon

Limp Gawd
Joined
Nov 1, 2005
Messages
505
Just had a 2 hour sit down with my Seagate rep.... she was unable to ansear alot of my questions though im sending the engineering department an email here tonight and should have a responce early next week... Any questions anyone wants me to ask? Ill take a list for the next few hours and i can include the results.

Biggest one i have is "How will defragmenting effect the 'frequently' used algorithem" and how big is the "frequently used algorithem"'s memory.... 3 years down the road on a heavily used drive will it crash when the memory fills up?

those are just a couple that I have... and im sure you guys must have a load, so lemme have em.

See if you can get the to explain to you how the cache works. When is it updated. Since there is no write caching it must be doing it in the background. What is the trigger? I'm surprised they are being so secretive with this. If you had some details (like with the Silverstone HDD boost) you can figure out its weaknesses and strengths so that you can determine if you can live with the trade-offs.


Some of these questions were answered during the Seagate webcast. If the flash dies then the drive will act like a 7200 rpm drive. Defragmentating will work, since there is no write caching the it will just by-pass the cache. The only thing that could happen is that if a block is written frequently as part of the defrag then it might get transferred to the cache. From what I've heard so far, this is my guess of how the cache works.

1) The HDD controller tracks which blocks are being loaded from the computer. Since you only have 32MB DRAM they can't be tracking the data itself, most likely the blocks in a look-up table.
2) If the block is read 2-3 times the algorithm recognizes this as a frequently accessed block and starts the process of transferring the block into the NAND cache.
3) Now as new files are loaded from the drive a compare happens to see if the data is in the cache. This means that if there is no cache hit the read access will be slower than a regular drive. You can see this in some of the benchmarks where the initial runs are slower than a 7200rpm drive.
4) The table must be kept in RAM for speed purposes and save to NAND at shutdown and reloaded on power-on. So a portion of the 32MB ram must be reserved for this. The question is how much and does this make the DRAM cache more comparable to standard drives.
 

Tau

Limp Gawd
Joined
Mar 3, 2010
Messages
196
See if you can get the to explain to you how the cache works. When is it updated. Since there is no write caching it must be doing it in the background. What is the trigger? I'm surprised they are being so secretive with this. If you had some details (like with the Silverstone HDD boost) you can figure out its weaknesses and strengths so that you can determine if you can live with the trade-offs.


Some of these questions were answered during the Seagate webcast. If the flash dies then the drive will act like a 7200 rpm drive. Defragmentating will work, since there is no write caching the it will just by-pass the cache. The only thing that could happen is that if a block is written frequently as part of the defrag then it might get transferred to the cache. From what I've heard so far, this is my guess of how the cache works.

1) The HDD controller tracks which blocks are being loaded from the computer. Since you only have 32MB DRAM they can't be tracking the data itself, most likely the blocks in a look-up table.
2) If the block is read 2-3 times the algorithm recognizes this as a frequently accessed block and starts the process of transferring the block into the NAND cache.
3) Now as new files are loaded from the drive a compare happens to see if the data is in the cache. This means that if there is no cache hit the read access will be slower than a regular drive. You can see this in some of the benchmarks where the initial runs are slower than a 7200rpm drive.
4) The table must be kept in RAM for speed purposes and save to NAND at shutdown and reloaded on power-on. So a portion of the 32MB ram must be reserved for this. The question is how much and does this make the DRAM cache more comparable to standard drives.


How she explained the cache to me was that it records an activity log for each file (im assuming by file she meant block) and the ones that are accessed most often get put into the NAND storage... now with defragmenting... my question was what happens if someone defrags every night... and the cache fills up with those files being defragmented since they would in essence be "accessed" every time the defrag goes... seems to me the NAND would fill up pretty quickly.


Im going to shoot off the email tonight i have a huge list of questions about how the drive operates, i will include yours as well and post the results here when i hear back from them (should be early next week)
 
Top