Chaintech Apogee Astro 266X 16GB USB Flash Drive Review

A few years back, portable drives (both Flash and Hard drives) used to be all the craze… for me at least. Those were the days when I was in college and needed to carry something more portable than a laptop all day long. Since most cell phones back then didn’t support high capacity memories, the most cost effective choice was to carry a USB Flash drive. And being a student of tech, the only thing that mattered to me – and most people around me – was the sheer performance of a device.

Back in those days, people would awe at anyone carrying a Sony Micro Vault with a mere 256MB capacity. I could put in all my semester assignments and still have room for my 12 favorite tracks and a couple of essential tools and utilities. That was 2004, and a mere six years later we have a whole range of different devices that can access data from Flash drives, with the drives themselves scattered along a broad spectrum with capacities ranging to 256GB in one end while speeds can go as high as 34MB/s reads and 28MB/s writes (on USB 2.0) at the other.

The Apogee Astro 16GB aims to strike the perfect balance between speed, capacity and price. The Apogee Astro range is divided into three drives of 16GB, 32GB, and 64GB capacity. Walton Chaintech also claims that there drive is fastest USB 2.0 MLC flash drive in the world right now, claiming theoretical speeds of 41MB/s for reads and 29.5MB/s for writes. Today, we put these claims to test.

Packaging and Design

The flash drive comes in a standard plastic package, the type most drives are bundled in. Surprisingly however, you only get the flash drive in the package and nothing else – no tethering cable or even a small manual or specification sheet.

The design of the flash drive itself is very sleek and business oriented, unlike most bulky or flashy designs from the likes of Kingston, and Corsair etc.  The material itself is pretty strong and it can definitely withstand a couple of falls without breaking. There is a red status indicator LED on its tail along with a small indent for a tether that wasn’t supplied with the package.

While Chaintech says the aluminum injected material is designed to dampen shocks from drops and collisions with other objects, I didn’t want to risk my sample by dropping it on a concrete floor.

Technology

NAND Flash memory, the type with most widespread use today has two basic memory cell structures – Single-Level Cell and Multi-Level Cell. Single-Level Cell or SLC Flash memory stores 1 bit of data per transistor cell. The obvious advantage of using this technique is that access and write times are significantly lower meaning faster performance. Another, less subtle but equally important advantage is that SLC only need a simple error detection and correction algorithms since one cell can only have two possible states. The obvious disadvantage here is that more cells are required to store data compared to MLC flash, hence increasing the price and size considerably. SLC Flash technology is generally used in high performance memory cards and flash drives which have low capacities.

Multi-Level Cell Flash stores more than 1 bit of information in a single cell. Often cells have more than two states which allow them to store multiple bits in one cell. This results in faster read operations as more data can be accessed in one operation, as well as denser memory organization compared to SLC, thus cutting manufacturing costs. This however comes at the expense of a more sophisticated and taxing error detection and correction algorithm and slower writes since the whole cell has to be wiped. Most consumer level flash storage is based on MLC technology.

Another important technological aspect regarding the Flash memory is wearing and erasing. While data can easily be read from flash memory byte by byte, writing data to flash involves erasing the entire block first, and then writing the data in the respective byte. So if you have a block size of 4kB and you want to write a small file of say 1kB on a block that already has 2kB worth of data, the solution isn’t really that straight forward. The controller might first read the contents of the block and write them to another empty block, and then finally write the new data into the target block. Alternatively, a flash controller might cache the target data in an onboard cache memory and read the contents of the target block into the cache as well. It would then combine the data with the target data already in the cache and then wipe the target block and write the new combined values. Sounds complicated right? That’s because it is.

Flash memory cells are designed to sustain a limited number of erase operations. While that number has grown a lot these days (from around 100,000 cycles to 1,000,000 cycles), flash memory does start to fail quite early if operations aren’t optimized by certain wear leveling algorithms.

The Chaintech Apogee Astro uses the latest generation quad channel Multi-Level cell technology which allow it to attain greater speeds compared to the traditional MLC based flash drives in the market. Chaintech is also employing a sophisticated wear leveling algorithm optimizing insertions and dynamically distributing writes. This gives the Apogo Astro considerably better speeds and life time, which Chaintech is also backing with a 2 year warranty on the device.

Test Setup

I chose to run Everest Ultimate to get to a low level data on the drive such as read and write speeds along with access times. For a more practical and real world data set, I used PCMark Vantage’s HDD Test Suite, which simulates a mix of real world data sets from the likes of Windows Media Player, Windows Live Photo Gallery, Windows Movie Maker etc.

To make sure the results were as accurate as possibly can be, all the tests were run after reformatting the flash drive while all killing all unnecessary background services that may hamper the results. For the sake of precision, I ran the tests three times.

The flash drive also came with its own speed boost utility, but we didn’t use it for two reasons. First was that you wont be able to install it on every time you plug it into a new machine, which might be a public computer at school or in a library. The second and more important reason was that the tool wasn’t compatible with 64-bit versions of Windows.

Benchmark: Everest Read Speeds

Most accesses to any storage medium are read unless you are using it for backup purposes. Therefore it is important for the engineers to prioritize read performance over writes because that would be the most dominant operation performed.

Both reads and writes operations are divided into three distinct categories based on their nature. Random reads occur when you request data from storage in any random order – you don’t request the next sequential block but any other block. Sequential operations occur between two points where you start reading from point A then continue reading every block linearly till point B. The third type is buffered reads where the storage medium caches a larger block of data in memory in case you make a request to some other piece of data from that block. This usually speeds up the operation if reads hit the cache, while a miss would require to fetch a new block of data.

Random and Linear reads were pretty close to each other which isn’t surprising for a Flash memory. You can randomly request any block of data at virtually the same speed because there aren’t any mechanical parts or revolutions involved.

The maximum speeds we hit were 32.1MB/s during random reads which were about 8MB/s short from the speeds claimed by Chaintech. On average, we were getting 26.7MB/s speed which was pretty close to the minimum ensuring that Apogee Astro has a pretty decent read speed generally putting it in the high performing USB 2.0 memory category.

Buffered reads were constant at 28.1MB/s which is quite good. The single dip going down to 23.2MB/s wasn’t really consistent during our benchmarks, so it really should be a rare case. Of course getting a constant 28.1MB/s in real world usage would mean that your reads almost always hit the cache.

Update: We did ask Walton Chaintech about the results we got from our repeated benchmarking. The 32.1MB/s read maximum we got was noticeably less then their claimed 41MB/s. Their official response is that the speeds they had claimed were result of a typographical error and their own internal benchmarks put the speeds around 32MB/s as well.

Benchmark: Everest Write Speeds

Write operations might not be the common case here but they certainly have to be fast enough to keep up with read operations and not bog down the entire system.  Like I had explained a couple of pages ago that writing to a Flash memory is a lot more complicated than writing to traditional magnetic discs hence performance isn’t really very uniform when it comes to random writes.

While linear writes seem consistent enough, Random operations were all over the place and were (for the lack of a better) Random. Speeds were up as high as 19.5MB/s in some scenarios, while dipping as low as 3.93MB/s in some other. The average of 8.88MB/s was on the lower side as well.

Also the maximum speed of 19.5MB/s wasn’t really on par with Chaintech’s claimed speed of 29.5MB/s. In fact it was a whole 10MB/s shorter – which is quite a lot.

Buffered writes paint a much clear picture of how the cache comes into play. Whenever the target block in the cache had to be written back into the flash memory, the write speed would dip to around 5-6MB/s. Otherwise it was constant around 10MB/s.

Update: We did ask Walton Chaintech about the results we got from our repeated benchmarking. The 19.5MB/s read maximum we got was noticeably lower then their claimed 29.5MB/s. Their official response is that the speeds they had claimed were result of a typographical error and their own internal benchmarks put the speeds around 24MB/s, which was decently close to the ones we got.

Benchmark: Everest Access Times

Access Time is the time between the issuance of a read/write instruction to the point where the data is ready to be read/written. In simple words it is the time to find the the target block of data in the memory. Traditional hard disc drives have access times in the order of milliseconds since there is a mechanical overhead involved. Flash memory on the other hand stores the data in electric cells, so the time to access it is as fast as the system memory (RAM).

Writes on the other hand can be tricky, since the existing block may need to be wiped before data can be written, so the access time usually varies a great deal and is in the order of a couple of hundred milliseconds.

The read access time is stable at .75ms or 750 nanoseconds, which means this drive can easily handle most read operations without a noticeable delay.

Just like random writes, write access times are also all over the place. Of course this indeed is expected behavior in most cases.

Benchmark: PCMark Vantage HDD Suite

While Everest is a pretty standard synthetic benchmark, the performance what matters in the end is which you get in a real world scenario. Now it would be difficult to try to simulate every real world scenario you might put your flash drive through, but PCMark Vantage’s HDD Suite does have a nice simulation of synthetic workloads that get pretty close to real world figures.

According to PCMark Vantage, Apogee Astro is best suited for the photo related tasks, and gaming. Of course you shouldn’t really be installing any games on a flash drive though unless you are just going to connect it to one system.

Conclusion

The USB 2.0 specification states the theoretical maximum data rate of 480Mbits/s which equals 60MB/s. On average, about 20% or 12MB/s of that theoretical maximum bandwidth goes to the controller overhead. That leaves the maximum attainable speed to about 48MB/s, which is divided across all the connected USB devices on the controller.

Astro 266X flash drives are still among the fastest drives out there and if the performance advantage isn’t enough to convince you, then they also have the looks to back it up.

With USB 3.0 devices starting to roll out, its nice to see that manufacturers are still trying their best to max out the limits of USB 2.0. And that is a smart move because you won’t find USB 3.0 integrated in any of the chipsets by Intel or AMD yet, and the only way you can get on the action is by paying extra for a premium motherboard or an add-on card, which really isn’t the solution most users are looking for.

So if you need that little extra performance today without going through a lengthy upgrade performance, then going with a high performance USB Flash Drive like the Apogee Astro 266X is a pretty safe bet.