Miscellaneous Aspects and Concluding Remarks

The 'Dynamic Thermal Guard' technology adopted by Samsung appeared interesting from the marketing description (automatic throttling to keep the case temperature below 45C). In order to figure out the impacts, I deleted the volume on the X5, and set up a fio workload to write sequential data to the raw drive with a block size of 128K and iodepth of 32 to cover 90% of the drive capacity. The internal temperature, instantaneous write data rate, and total amount of data written in total till that point of time were recorded.

Sequential Write to 90% of Disk Capacity - Performance Consistency

We see that the thermal management scheme adopted by Samsung is very detrimental to long-running sequential write jobs (typical of first-time backups), despite managing to limit the internal SSD temperature to less than 56C. Write speeds often got down as low as 60 MBps. On the other hand, our DIY configuration with the SanDisk Extreme Pro 1TB SSD in the TEKQ Rapide aluminum Thunderbolt 3 enclosure managed to pass the same test without any issues in less than 1/10th of the time taken by the X5 for the same amount of data. The internal SSD temperature did go above the X5's (reached 57C at the end of our test), but, the passive-cooling friendly ridged aluminum enclosure managed to do an effective job.

Compared to the TEKQ Rapide, it does appear that the X5 chassis is unable to keep temperatures down without extensive throttling. The choice of magnesium compared to aluminum is puzzling (given that the thermal conductivity of aluminum and magnesium are 205 w/mK and 156 W/mK). In addition, the absence of additional surface area in the chassis to dissipate the heat (in the form of ridges or fins) is also a negative. Combined together, we see that the Samsung Portable SSD X5 throttles more under heavy loading compared to some of the other Thunderbolt 3 SSDs in the market.

SLC Cache Size Determination (using Sequential Write to 90% of Disk Capacity)

The experiment also allowed us to determine the approximate SLC cache size of the internal SSD. On plotting the instantaneous write transfer rate along with the total data written up to that point of time, we see that the X5's SSD has around 40 - 45GB of SLC cache, while the SanDisk SSD in our DIY configuration has around 16 - 20GB of SLC cache.

The Thunderbolt 3 specifications indicate that the host port must be able to supply up to 15W for bus-powered devices connected to it. Since the Samsung Portable SSD X5 is a bus-powered device, it is given that its power consumption can't exceed 15W in order to be a certified Thunderbolt 3 device. It is still relevant to take a fine-grained look at the power consumption profile.. Using the Plugable USBC-TKEY, the bus power consumption for both SSDs was tracked while the CrystalDiskMark workloads were processed. The workloads were set up with an interval time of 30s.

Drive Power Consumption - CrystalDiskMark Workloads

We find that the power consumption stays mostly south of 7W, though we did notice spikes of up to 9W in our high-queue depth sequential write tests.

The Samsung Portable SSD X5 has no trouble supporting TRIM commands from the OS. The performance for most consumers is good enough - in fact, in our DAS suite, the X5 managed to lead the charts in almost all the tests. The issues start cropping up only under sustained loading in cases such as transferring 50+GB into the X5 without any break inbetween. I am still waiting to hear from Samsung on why they opted for a magnesium chassis compared to the aluminum used by most other competitive offerings. [Update (Samsung's response, paraphrased): Since the X5 is a portable product, it needs to be durable against external impact, light-weight, and dissipate heat well. Magnesium is much lighter than aluminum/steel and also durable, enabling use in products that require rigidity and lightness such as cell phones, cameras etc. Although aluminum has better thermal conductivity, we have found that magnesium is the optimal metal solution to meet all the three conditions - weight, durability, and heat dissipation.]

Moving on to the pricing aspect, we find that the X5 is not particularly cheap - we would have no trouble with that had the X5 not throttled significantly in our new extended sequential writes test.

Price per GB

Consumers would be prudent to treat the X5 as a premium product - it performs admirably for the vast majority. However, for power users who frequently transfer 100s of gigabytes in one go, a solution like our DIY Thunderbolt 3 SSD is a better choice. Our DIY device does not look as sleek as the X5, but, it is cheaper and has more consistent performance.

AnandTech DAS Suite and Performance Consistency
Comments Locked

23 Comments

View All Comments

  • HStewart - Tuesday, August 28, 2018 - link

    I search around this drive is $700 for 1T and I hope since Intel has open up the spec for TB 3 - that the prices would drop. I have also heard that new controller will be backwards compatible with USB 3.x - of course at much slower speed.

    One thing that would be nice in review - is compare a similar drive using USB 3.x - actually both Gen 1 and Gen 2.
  • Reflex - Wednesday, August 29, 2018 - link

    Someone mentioned recently that despite Intel's statement that they would open up TB3, it hasn't yet happened and communication on the topic has ceased. I haven't looked into it myself yet, but if true that implies that it really won't ever go mass market or get much cheaper.
  • The_Assimilator - Thursday, August 30, 2018 - link

    In that case, TB has cemented its legacy as Firewire's successor, and in a few years will vanish entirely unless Intel decides to keep it on extended life support.
  • shelbystripes - Tuesday, August 28, 2018 - link

    The throttling scheme is by far the most concerning part of this. I doubt there are firmware changes that could enable greater performance at the same temperature... but I wonder if Samsung could release firmware to increase the maximum internal temperature. 56C seems a bit low as a threshold, given that the thermal protection works as planned. Perhaps raising the throttling threshold to 60-65C could increase performance? The older 850 EVO series was rated up to 70C, and AnandTech preciously reported that the 970 EVO (which this is supposedly based on) gives thermal warnings at 85C.

    Running hot might have a small effect on shelf life, but I doubt most people will be writing 100GB+ to it constantly.

    I would be pissed if I used one of these to make an initial backup and it took several hours due to throttling, but if that one-time load was successful, incremental changes shouldn’t hit the higher thermal threshold at all.
  • shelbystripes - Tuesday, August 28, 2018 - link

    Previously, not previously. It’s 2018, why don’t comments have an edit button?
  • shelbystripes - Tuesday, August 28, 2018 - link

    Sigh.
  • jordanclock - Tuesday, August 28, 2018 - link

    If it means the NAND lasts longer, I think it's a fair trade off.
  • shelbystripes - Tuesday, August 28, 2018 - link

    It depends on how balanced the trade off is. 56C seems rather low given the higher thermal limits on other SSDs, and seems to be heavily impacting performance after just a couple minutes of sustained transfer. Raising it to 60-65C shouldn’t heavily impact long term lifespan, since their desktop version can handle higher temperatures just fine.

    I have trouble believing the durability impact, just from raising the throttle temp into the 60-65C range, is so massive that it justifies this much throttling. Especially since this is a mobile device, which isn’t going to serve as a boot drive or anything like that.
  • notashill - Tuesday, August 28, 2018 - link

    It's 1C cooler than the TEKQ model and 1/10 the performance, in spite of a larger and heavier package.
  • Dug - Wednesday, August 29, 2018 - link

    Where do you get 1/10 the performance from?

Log in

Don't have an account? Sign up now