HDD to SSD transition effects (pics/charts)

[jonlumpkin]

Due to the sales on the 64GB Samsung SLC SSD drive last week, I decided to jump on the upgrade. This is model MCCOE64G8MPP-OVA, and is the same model that originally shipped with the x300 (I needed a 1.8" to 2.5" converter to make it work in the x200).

This drive has consistently got good marks and is one of the few SLC drives available (almost all other drives, including the Intel x-25M are MLC). However, despite numerous benchmarks to its performance in synthetics, there have been surprisingly few about real world effects of the upgrade (especially battery life). Therefore, I am making this post as a comprehensive analysis of the real world differences between the MCCOE64G8MPP-OVA and the 320GB 5400 RPM Hitachi 5k320 it replaced (this is one of the quieter and more energy efficient conventional hard drives available).

I have run several benchmarks and practical tests. For ease of reading, I have entered them into Excel, and produced charts. The spreadsheet with complete details is attached in a .zip archive.

First is a set of practical tests for boot (defined as the time between pressing the power button, and the Power Manager guage appearing), hibernate, resume, shutdown, and a 4.37GB (DVD size) MKV file copy. As you can clearly see, the SSD has a substantial advantage in boot time (especially when you subtract the 15 seconds for the BIOS), entering hibernate, and copying a large file to itself (only half the time).



Next is the test that I consider the most important - Battery Life. I ran these tests under Windows 7 with Wi-Fi off, processor (SL9400) locked to slow, and the screen set at minimum brightness; the goal was to maximize the effect of the drive, and minimize other variables. I measured the difference in Watt-Hours (WHr) using Lenovo's power manager, and calculated this against time (WHr * 3600 / time [seconds]) to yield an accurate average wattage. Lower is better for all these tests.

The SSD leads in every category, but the real world difference is actually quite larger than the charts would imply. On a straight file copy, the SSD is 1.58 watts more efficient than the HDD; however, because the task completes in only half the time, only 44% of energy is required on the SSD.

The second test is playback of a 25:21 second Avi file (XviD video @ 2717 kbps, MP3 audio @128 kbps). This would simulate the playback of a video file on an airplane or similar location. The SSD uses 2.58 less watts than the HDD, or 1.09 less WHr for the entire video. This equates to 6:10 on the SSD vs. 4:57 on the HDD for playback of this kind of content (based on the 66.24 WHr design capacity of the x200T 8 cell).

The final mark is an idle power draw. The difference here is 0.30 watts; this is significant if not tremendous. Clearly there is a big improvement in energy efficiency for the SSD during any kind of activity, but at true idle both the SSD and a 5400RPM HDD can be quite efficient.


The third test is straight file transfers to the drive, from the drive, and to itself. These tests were all performed on my Desktop under Windows 7 (a 640GB 7200RPM WD6400AAKS served as the source drive for these tests). All these results are in seconds, so lower is better.

The SSD has an advantage in every single category (except an anomalous 7x AVI read). The difference is largest on writes of files >= 2MB, especially when reading and writing to itself. However, the SSD has a clear advantage over the 5k320 across the board.



Next up are a variety of synthetic benchmarks. I am not as fond of these, but they still serve a valuable purpose.

The first test is SteelBytes HD_Speed 1.5.4.72. This allows running reads and writes of any desired size across a drive. These tests were also performed in my desktop because the write test will destroy all data on the drive. The SSD leads at all block sizes, especially large reads. I have also included a 2GB SD card to illustrate the difference between high quality SLC flash and ultra-cheap ($6) MLC flash.



Crystal Disk Mark is a similar test to the above. It runs at three block sizes (sequential, 512k, 4k). Again, the SSD has a clear advantage in every category.



Next up is HD Tune. The clear differnce here is consistency and access time. The SSD has effectively no access time (0.2 ms), because there are no heads to relocate. Additionally, it's speed is remarkably consistent across the entire disk. Conversely, the HDD has a 17.7 ms access time, and its speed degrades as you move to the end of the disk. Thumbnails are attached.



This concludes the run of benchmarks, I have also attached supporting images below.


Conclusions:

An SSD (especially the MCCOE64G8MPP-OVA) is a very clear improvement over my prior drive, if for no other reason than energy usage. It is also a superb, and remarkably consistent, performer. Additionally, because it is SLC, it doesn't suffer any of the stuttering problems of the JMicron drives, nor potential internal fragmentation due to write combining of the Intel x-25M.

I anticipate that many people will take issue with my choice to compare this SSD with a 5400RPM HDD. However, my top concerns are heat, noise, and battery life. For these reasons, I don't consider 7200RPM drives to be an attractive option for an ultra-portable notebook or Tablet PC.

One other point worth noting is noise and vibration. The 5k320 is one of the quitest drives available; however, the SSD is a very clear improvement (the system is effectively silent, except for the occasional Penryn whine). The SSD is also a solid state device, and immune from orientation. The Active Protection System driver knows this, and automatically disables head parking on SSDs. This is invaluable for a Tablet PC as I often hold it in one hand, and the constantly parking HDD would undermine performance.

If you value battery life, heat, noise, and/or performance an SSD is a very worthwhile investment. Newegg is currently selling the 1.8" version of this drive for a great price, and it's definitely worth it, assuming you can live with the reduced capacity.

[jonlumpkin]

Here are pictures of the drive in question, and the upgrade process.

This is a picture of the SSD in the 1.8" to 2.5" storage converter (FRU 42W8019). Note that it weighs only 50 grams.



Here is a picture of the 2.5" 5400 RPM HDD, with rubber rails, that it replaced. Note that it weighs 108 grams (more than double the SSD).



This is a picture of the SSD in the hard drive bay. Note the empty space that reduces overall weight.



Here is a final comparison shot. This shows a 3.5" 7200RPM drive, a 2.5" 5400RPM drive, the 1.8" SLC SSD, and a 2GB SD card. All are placed on top the x200 Tablet to show a overall size comparison.

[jonlumpkin]

Somehow the attachments got lost in the upload.

Attached are screenshots from Crystal Disk Mark, HD Tune, and the spreadsheet.

The 1st and 3rd image are for my Hitachi 5k320; the 2nd and 4th are for the Samsung MCCOE64G8MPP-OVA.

[zephir]

Your benchmark result is a little bit low. I attach my own result. Try to see if there's any setting you can tweak. I turned on advanced performance for harddrive in Device Manager.

[Jackboot]

We're lucky to have you here Jon Great info!

It might be worth pointing out that the 5400rpm HDD that you use as a reference is actually quite a fast notebook drive. The difference in performance would be even more drastic for someone upgrading from an older 5400rpm drive.

Also, according to the latest Anandtech article, the Samsung SSD performance can vary quite a bit depending on if the drive is "new" or "used" (which doesn't take long). I won't go into detail here (read the article if you're interested), but that might be the difference between Zephir's benchmark results and Jon's.

[jaredy]

Great write up jon. I would rep you if I could again. As jackboot mentioned that anandtech article might be an interesting read for people who want some additional info about the SSDs.

I am really happy with my SSD and for the price (now) it is really worth it for the mobile user. Battery life and heat are key as jon mentions.

For each user you will need to prioritize your needs for space vs battery life, heat, noise, and the ability of an SSD to take shocks. I prioritize all that over space because I store my files on a network.

[alphakenny1]

hey thanks jon. i got a friend working at intel and he hooks it up. i might get that x-25m haha.

[zephir]

We're lucky to have you here Jon Great info!

It might be worth pointing out that the 5400rpm HDD that you use as a reference is actually quite a fast notebook drive. The difference in performance would be even more drastic for someone upgrading from an older 5400rpm drive.

Also, according to the latest Anandtech article, the Samsung SSD performance can vary quite a bit depending on if the drive is "new" or "used" (which doesn't take long). I won't go into detail here (read the article if you're interested), but that might be the difference between Zephir's benchmark results and Jon's.

I believe that the article you're talking about mentions that a used drive is supposed to be slower than a new drive. The reason is because when you delete stuff on the drive, it marks it as useable, but doesn't delete the data right away. When you write new stuff on to that same cell, you have to delete it first, and then write new data on it. That's why it's slower.
However, this means that Jon's drive should be faster than mine, which is not the case. My benchmark is after 6 months of use.

[jaredy]

Benchmarking is weird since other factors and settings could affect it. I think the real world remarks are the most important in this case anyways.

[zephir]

Benchmarking is weird since other factors and settings could affect it. I think the real world remarks are the most important in this case anyways.

His first benchmark is the most alarming though. It is approx. half my read and write speeds.
His second benchmark is fine, in my opinion. There's not gonna be much difference in the responsiveness of the system from my benchmark and his benchmark.