Adaptive Memory Technology
Seagate had to develop new software/management algorithms for the hard drive that can take advantage of the additional 4GB flash memory. The datasheet mentions SLC NAND flash memory, which tells us that the flash portion should be very suitable for heavy I/O operation, as well as strong on both read and write operations. Unfortunately, the datasheet does not tell us how the flash memory is organized. Most flash SSDs reach 200+ MB/s throughput by combining multiple flash memory channels. This is certainly not the case on the Momentus XT, but we don’t have specific information.
The technology working inside the Momentus XT is called Adaptive Memory. According to Seagate, it monitors activity and gathers so-called hot data within the fast 4GB flash memory. You could say that it works like a cache, but the contents won’t be lost in the case of a power failure. Frequently used files, user data, and operating system files should technically all be located within the flash memory section to be instantly available. However, we found that a technology that operates so dynamically to be difficult to measure and benchmark. We had a similar situation with Adaptec’s MaxIQ solution that utilizes an SSD to cache an entire RAID array. Therefore let’s look at what Adaptive Memory does.
Seagate mentions ‘constant monitoring’ performed by the drive controller to ‘customize system performance for the user.' So far, our performance numbers only reflect low-level performance, but it’s hard to come up with robust testing that really exposes the impact of this feature. Yet, the Momentus XT beats the competition in most low-level benchmarks without getting anywhere near the performance we came to appreciate on flash SSDs.
There are many benchmarks where the drive is faster than its 7,200 RPM competition in the 2.5” hard drive space, but typically the drive doesn’t really outperform them by a wide margin. And there are a few benchmarks where it is even slower than the competitors. At this point, I have to admit that I expected a few performance spikes here and there, providing visible proof of concept for the Momentus XT and Adaptive Memory. Apparently, these either do not exist or our benchmarks are simply incapable of showing the performance improvements that we were expecting to materialize over time.
Be that as it may, expectations should clearly be based on the fact that the drive is not a flash SSD. Rather, it is a Momentus hard drive with the XT suffix. In this context, its performance is just about right--and on top of the competition. We’ll keep testing with a second Momentus XT and will analyze the inner workings of Adaptive Memory Technology with any more information we can glean from Seagate. I’m sure there is more to come, as only a firmware update could introduce additional performance in the blink of an eye.
Related Stories :
- Momentus XT 750 GB Review: A Second-Gen Hybrid Hard Drive
- In The Lab With Seagate's Momentus XT 750 GB Hybrid HDD
- Deals January 16: 750GB Seagate Momentus XT $119
- Deals Dec 27: 750GB Seagate Momentus XT $99.99
- Seagate to offer hybrid hard drives for Vista notebooks
- Momentus XT Review: Seagate's Marriage Of The HDD And Flash Memory
- Speedy Notebook Storage with Seagate's latest Momentus Drive
- Seagate Says Future is Hybrid HDD, Not Pure SSD
- Seagate Hybrid-HDD-SSD Unveiled for Notebooks
- Seagate, Toshiba to Make SSD + HDD Hybrid?
I think these hybrid drives need an all-new testing methodology that is neither like a hard drive (as you have classified it) nor a pure SSD, nor even simply the sum of the two tests.
You mentioned that performance *might* improve if the benchmarks were rerun. Could you re-run PCMark 3 times and see if the 3rd time is any faster than the first? The idea being that the first time is simply the hard drive, and hopefully by the 3rd (or 4th time), the data ends up in the flash.
Am I the only one that's not impressed in the least?
Im not. I would easily shell out the money for an 80GB ssd if it also included 500G of mechanical storage. 4G is nothing. Absolutely worthless for this. Give us something real and this will be nice.
I absolutely agree with the conclusion. You look at this like a hard drive. In some things, it appears to outperform the competition, and the price is not unreasonable. For too many other things, it is no better, or even worse. I think it is promising, but as essentially "1.0" tech, it is a little underwhelming. Hopefully all it needs is a few minor tweaks.
Some what unimpressive but then again it does well enough for most uses. What they need to do is make drives with much larger caches instead of the meager sizes that are common place. 64mb cache on the high end drives is a joke ware 128mb or 256mb would make a difference ware unlike today ware much of the bandwidth is wasted before even hitting the cable. Internally typical hard drives have up to 500mbs before that is wasted by the controller due to cache misses and seek ques.
First samples are always less impressive, but i bet this hybrid tech will be the future, atleast in a small part, of the HDD market. Ssd`s are to expensive with low cap. In 2-3 months there will be great hybrid hdds for sale.
Not impressive at all but what more can you expect from Seagate afterall?
Truth is 4gb just isnt enough anymore. Although the performance numbers are ever so slightly better than current 7200 rpm hard drive the near 50 minute hit on battery life just kills the idea. Id be willing to take the battery life hit if the I/O's were higher and I could get read speeds around 140 mbs.
I think is better to wait for ssd price to go down
I agree, that was rather pathetic. I expected it to at least out-perform 3.5" hard drives (smaller platter + same speed 7200rpm should = faster access, reads and writes) instead we get a last gen 2.5" drive where they added 4GB flash as a buffer and poorly utilized at that, and call it a hybrid. The idea behind hybrids is to get the best of both contributors. In this case it should be capacity of hdd plus the speed of ssd. This thing wasn't anywhere near an ssd in performance and it was a sad misuse of the "hybrid" term.
Good review though guys. Very thorough. Its clear that Seagate needs to get back to the drawing board. Perhaps start with an ssd and work on adding hdd platter storage to it instead of trying to work in the other direction.
Am I the only one that's not impressed in the least?
Definitely not. Just looking at the benches, the "WOW!" factor just isn't there. Price considerations, I'm not even sure if I'd buy one if there were no price differences with a conventional platter-only drive. Some things were marginal, but it also took some hits. In a notebook/laptop environment, power consumption is not the department you want to see one of those hits. If there is a future to this as the alternative to SSD only drives, then the technology is far from perfected with this as an example.....
Is this written for an alternate universe where HD tach or better benchmark utilities don't exist?
How bout graph of throughput and seek times over drive geometry?
More practical is if the effectiveness of their algorithm. I'd be satisfied with some junk adapter that just spans flash memory with disk space if you could ensure the OS partition is on the flash portion of geometry.
Consider this is Seagate (famous for their rushed work); I’ll wait for WD, Hitachi, Toshiba and even Samsung.
nah we still want cheap SSDs
Anandtech.com has a different set of benchmarks his conclusion is that it has the performance level of A velociraptor.
Should have run at least 1 SSD in the tests. Because this drive reminds me of a toyota prius... mixing old tech with new tech and delivering a POS no one but a small niche market wants. Of course they will work on this and in 2 or 3 years have a 32gb or 64gb flash section and then we will have tests where their drive out performs a combination of a velociraptor and some other SSD in drive to drive transfer. Then people will start to consider buying one of these and they will make the biggest mistake of their lives. Listen, people, you want the SSD because its faster than pretty damned much any other drive type on the planet but they are expensive. Spending money on a drive that has less nand flash memory so you can get a regular hard drive in the same package is a marketing ploy to get you to buy this device when you really want an SSD. How many of us dont have at least 1 sata 3gb hard drive laying around thats around 320-500gb? Exactly... so what's the point in buying this drive if its half and half but the worst of both?
is it me or was the benchmarking ill conceived for such a device, the speed boost will only occur once the darn thing learns from your habits, how many times were each benchmark run can we get some stats to see if on running the same benchmark a second time yields a speed boost, or maybe an nth number of times, how many times would an application need to be run before it gets cached, once cached how much better does it perform, how comes there wasn't a windows boot up time test? im pretty sure this drive would have been pre-optimized to cache windows system files
you can't just throw a bunch of random benchmark at this thing and expect it to perform other then a standard drive. Come on now we talking about THG here, your not suppose to miss this kind of stuff.....
The standard set of benchmarks used here doesn't really reflect normal usage where this drive is supposed to shine. Some common tasks like windows startup, application start and such done repeatedly until there is no improvement in loading times would have been nice.
I have a suggestion on how they could turn this into a write cache as well as a read cache: separate logic and SLC flash storage specifically for writes. This could be half or same capacity as the read SLC cache. The writes get handed off to the drive logic for writing, enabling the OS to continue on, and the write cache continues on in the background.
This would also allow for higher data integrity and security. Encryption could also be done through this method.
It would drive the cost of the drive up a little, but it would still be cheaper than a SSD, and many things would perform almost as well.
Although, I doubt anyone at Seagate would read this...
I agree. I think you guys should give another chance and see how much repetition helps it out, especially in OS startup.