• Home
  • Blog
  • Testing budget SSD Crucial BX500 with 480 GB TLC and 500 GB QLC memory

Testing budget SSD Crucial BX500 with 480 GB TLC and 500 GB QLC memory

14.11.2023 08:46

Over time, the situation in the SSD market is constantly changing: top models become more advanced, and budget models become even more affordable. This is especially noticeable in the SATA segment, where there are practically no “top” options, with the possible exception of the Samsung 870 Evo. However, even this raises concerns among some users due to past events where increased failure rates were identified in certain batches.

On the other hand, this SSD is of interest to many users, and we will return to considering this model in the near future. Today we will not focus on it, but also not on a few typical “mid-rangers” — their range is almost exhausted. When a buyer turns to the store’s assortment, he sees the Samsung 870 Evo model for 5,500 rubles, Crucial MX500 for 4,000 rubles, and nearby there is an option cheaper than 3 thousand (remember, this was in the summer, although prices could have changed). In this situation, the choice seems obvious: why pay more?

However, understanding “why” comes with experience of use. Basically, the demand for SATA drives is associated with the modernization of old devices, especially laptops, where installing another type of drive is physically impossible. Even replacing the hard drive with the most budget SSD significantly improves performance, even on older devices. The differences between them are noticeable.

Some users really shouldn't pay extra. For most operating system and application software tasks, a few thousand operations per second are sufficient, which is easily achieved by the most affordable SSDs, but remains elusive for hard drives. The differences in SSD performance are only apparent in the details, and while these differences are important to many users, alternative solutions are increasingly moving away from the SATA interface.

As a result, even manufacturers that keep “decent” SATA SSDs in their assortment are increasingly focusing on the production of budget models. For example, Crucial introduced the Crucial MX500 and Crucial BX500 line almost simultaneously. However, the second model sells better, which is explained by its lower price. These are products of different levels, and over the course of their existence they differ more and more, which confirms the changes in the SSD market.

How the Crucial BX500 fluctuated along with the party line

Comparisons between the Crucial BX500 and MX500 initially presented minimal differences, given both series' use of the same 64-layer Micron memory. The main difference was in the controllers: a Silicon Motion SM2258 controlled the MX500, while the BX500 had an SM2258XT. Despite only one letter in the marking, this was of fundamental importance. For example, the SM2258 was equipped with a DRAM buffer, while the SM2258XT did not have this capability. These controllers were low-cost, single-core, four-channel options, with more advanced alternatives having left the market. However, the SM2258 had different firmware with different SLC caching strategies, while the XT series had only one strategy: writing all data through the cache.

This allowed the BX500 to become slightly cheaper due to savings on the DRAM module and a slightly more affordable controller. In addition, the lines were divided by capacity: the MX500 was a 250/500/1000 GB range, while the BX500 was 120/240/480 GB. Thus, those who wanted to purchase the most affordable SSD turned to the junior BX500 line. However, those who were more knowledgeable on the issue did not seek to save at the expense of quality, including in terms of capacity. For example, when choosing a terabyte option, the choice was obvious: MX500. This became especially noticeable at a time when the price per terabyte of flash memory became more and more affordable.

Manufacturers follow market logic without trying to make the world better or worse. Some critics felt that the BX500 should not have been released, but they would also be dissatisfied with the lack of available models for those who focus specifically on affordable SSDs, ignoring other characteristics. And although Intel and Samsung did not initially present models in this class, this did not stop them from expanding their range of more expensive SSDs for more demanding users.

However, a few years ago, the scale of change in the SATA drive market did not seem so significant, despite the clearly visible direction. Micron launched production of 96-layer memory and brought it to the Crucial BX500 and MX500 lines. The maximum capacity has increased to 2 TB in the older line and to 960 GB in the younger one. However, Crucial BX500 models of this capacity did not last long on the market. Micron began mass production of not only TLC memory, but also QLC memory on 64 and 96 layers. QLC memory made its debut in the BX500, offering 1000 and 2000 GB models. This allowed Micron to release budget consumer lines using QLC memory.

Micron also mastered 176-layer TLC, which turned out to be very fast and inexpensive. This allowed the company to offer models with cheaper QLC memory. Currently, Micron produces budget Crucial BX500, P3 and P3 Plus lines on QLC memory. Thus, the BX500 line has been replenished with a 500 GB model instead of the previous 480 GB version using TLC memory. However, the cost of the new modification was not necessarily cheaper. The general conclusion is that the economic benefits remain more on the side of the manufacturer, and the choice for buyers is perhaps becoming less clear.

480 Gb
500 Gb

There are still different assumptions regarding the 240 GB version of the model. It is possible that TLC memory is used internally, however this may vary depending on how long the SSD has been in the store. Currently, purchasing such a capacity SSD makes economic sense, given the higher cost per gigabyte of larger capacity models. Despite this, we do not recommend purchasing this model due to the unknown memory technology inside.

On the other hand, it was decided to review half-terabyte SSDs with a SATA interface. Still, it is worth noting that in some cases it is not recommended to purchase such SSDs if there is a choice. However, not all users have such a choice, so we decided to take a closer look at popular models from Moscow retail stores as part of our series of articles, including this one.

Crucial BX500 480 GB

Dembel chord of the line. Here, the 176-layer Micron B47R TLC memory is used, which is controlled by the Silicon Motion SM2259XT controller. He, we repeat, does not need such memory; he cannot really use it. But it's better than QLC.

There's nothing more to discuss. Typically, Crucial SSDs come with a standard three-year warranty limited to 120 terabytes. However, in our area, the official manufacturer's warranty is often not relevant, since Crucial SSDs enter the country through parallel imports, where importers only provide a one-year warranty. This is another reason to prefer not so much established brands as new companies offering more affordable prices and order in matters of warranty. Predicting the lifespan of a device designed with an emphasis on aggressive minimization can be challenging, even for the most advanced technologies.

Crucial BX500 500 GB

The above especially applies to this modification. If the previous version was at least built on TLC memory, then this version definitely uses QLC, which makes it of lower quality. Second-tier brands can also sell such a device, but there is most often a lottery where you can get lucky (to a certain extent). Crucial's 480GB BX500 model usually offers the «good» option in this lottery, while the 500GB model always has the «bad» option.

It's pretty good that they have slight differences in capacity, otherwise it would be quite difficult to distinguish these lines: appearance, specifications — everything is too similar. At one time, for the 1000 GB model, the guaranteed “resource” was increased compared to the 960 GB version, but now even this change is not there — it was decided that buyers will have to make do somehow.

As for the internal components, everything is clear with the memory: it uses 176-layer QLC memory Micron N48R. But the SM2259XT2 controller from Silicon Motion is something new for us. It looks like it's even cheaper than the old version since it's only a two-channel solution. In practice, however, this should be enough: no one produces QLCs with a capacity of less than 1 Tbit (YMTC immediately switched to 1.33 Tbit), and SSDs with low capacities are in demand, so only four memory devices are used here. If there were four channels this wouldn't happen. In addition, the board layout is simpler, the chip itself is smaller and cheaper — in general, this is not about speed. To be fair, Maxiotek also (for the same reasons) replaced the four-channel MAS0902 with the two-channel MAS1102, and Phison originally introduced the two-channel S11 (a product that was announced to much fanfare five years ago, but the four-channel S13 failed to materialize, and the S17, announced three years ago, begins to gradually appear). This is a general trend. Of course, an updated controller may also have its benefits in the form of faster internal operations or support for a faster interface with modern memory. But combining it with QLC doesn't make much difference. Now we need to see how this all works in practice.

Testing

Testing methodology

Here we use a test bench based on an Intel Core i9-11900K processor and an Asus ROG Maximus XIII Hero motherboard on the Intel Z590 chipset, which gives us two ways to connect an SSD — to the “processor” PCIe Gen4 lines and the “chipset” PCIe Gen3 lines, but neither the first, neither the second one will be useful to us today. Unlike the “chipset” SATA600 controller, all participants will need it.

Samples for comparison

Crucial MX500
KingSpec MT-512

To simplify and improve the performance picture, we will choose two main drives for this test. One of them will be the Crucial MX500 — this is a good example that differs only in one letter and price. This will be our example of what can be considered good. While another drive, the KingSpec MT-512, which uses a bufferless dual-channel Maxio MAS1102 controller and the same Micron N48R QLC memory, will show us what can be considered bad. Let's see if Crucial can offer something better than this example.

Filling with data

Hasty clearing of the SLC cache with a corresponding drop in speed, as already mentioned, is a feature of the MX500 firmware. However, the “failure” here is only up to 300 MB/s, which is not important. In fact, the 500s of the previous version wrote data outside the cache even more slowly. And here the speed is also restored.

The same memory and SM2259XT — and a completely different result. However, for an SSD on this controller it is generally a bit atypical — one gets the feeling that Silicon Motion (perhaps just under pressure from Micron and other customers) implemented a direct recording mode in TLC. The usual behavior for the XT was “cache for all free cells” (that is, about a third of the capacity on an empty device) with a subsequent drawdown somewhere up to 50-60 MB/s. Here this only happens at the very end — when the cache, which was initially limited in capacity, is “expanded”. But the bulk of the volume is written at speeds above 100 MB/s. It’s not exactly a huge achievement, but it’s better than before. On the other hand, the cache has decreased — so in some places this approach may slow down a little.

But it was worth reworking the firmware earlier — now the BX500 on the B47R (and TLC in general) is already a thing of the past. QLC paired with the SM2259XT2 takes us back to the old days. And even worse — here the speed outside the cache is already dancing around 30 MB/s. So, if we evaluate this particular scenario, the previously BX500 was 3.5 times worse than the MX500, and the new BX500 is even 2.5 times worse than the old one. That is, the spread approaches a whole order.

But it could have been even worse — the same N48R memory paired with MAS1102 differs by an order of magnitude from the MX500 in terms of test execution time. You can be glad that the BX500 is at least a little faster. But this is some kind of sad joy — and the modifications on TLC did not shine at all, but QLC, and even in small quantities (we have already said more than once that it is better to look at this memory somewhere from 2 TB, and even then these will not be diamonds at all, as a rule) the speed sinks below the baseboard. Well, this plinth turned out to be a little higher than that of the Chinese — so what?

Maximum speed characteristics

Low-level benchmarks in general and CrystalDiskMark 8.0.1 in particular have long been limited by SLC caching and can only evaluate the cache. However, the information provided by manufacturers about the performance level of devices is also limited by these limits. Therefore, checking is always useful. Caching work is being done with the goal of maximizing cache utilization in real-world use cases and demonstrating high speeds despite decreasing memory costs.

This is all clear. SLC caching is called that for a reason — it works in one-bit mode, regardless of the memory “density”. Therefore, everything is beautiful — and thus conducive to saving. Except that in mixed mode, echoes of the presence or absence of a DRAM buffer are visible, but with sequential access the latter is not so critical.

With arbitrary, the difference is naturally noticeable. However, so far there is nothing criminal — after all, requests for real application and system software do not exceed the level of 2-3 kIOPS, and almost any SSD can provide this, and most often with a two to three times margin. For this reason, long queues are not relevant — there is no reason for them to line up when requests are processed faster than they are received.

Similar layout. Only you can write even faster — so far all operations are limited by the size of the cache, of course. It’s simple — we write wherever we want, I just adjust the address translation table, but when reading, we need to “honestly” display the contents of the requested block on the interface, but not the first one that comes across.

Such operations have a much greater impact on the speed of real software than the previous ones: “long” queues have nowhere to come from in practice — but blocks other than 4K bytes are very common. The number of operations per second on «large» blocks is slightly reduced, but they themselves are larger — so the resulting speed in megabytes per second is higher. Therefore, whenever possible, everyone tries to work this way. But, again, as long as we test the cache, nothing bad happens. Silicon Motion SM2259 is naturally faster than entry-level controllers, but the software requirements are more modest than the capabilities of even the latter. Memory can introduce dissonance into this symphony, but CDM still cannot “get through” to its real characteristics.

And this also fully applies to recording. Here, sometimes you can even get confused — which platform is more budget-friendly.

But here it’s almost impossible to confuse who is who. Even in greenhouse conditions, the difference between platforms can reach several times. But, once again, we repeat, relying on these same “hothouse” conditions is not justified at all — then you can hit reality very painfully.

Working with large files

Indeed, although low-level utilities can show high speeds, in practice it is always difficult to achieve such indicators. CrystalDiskMark, for example, works with small pieces of information within a single file. In today's environment, this information is usually stored in the SLC cache during testing, which does not always reflect real-world use cases. Additionally, actual file writing involves file system overhead operations such as changes to the MFT (Master File Table) and file system logs, which means writing data to different places and in different block sizes.

For a more accurate and practical performance assessment, you can use tools such as the Intel NAS Performance Toolkit. This tool allows you to test the device not only on an empty disk, where the cache size is maximum, but also in more realistic scenarios, when there is almost no free disk space.

Working in one thread is the most common (146% of cases), but also the most difficult scenario. However, it is difficult for drives with completely different interfaces — the SATA ceiling is too low for any other noticeable problems to arise. However, if the active implementation of QLC in low-capacity SSDs continues, this may be possible in the future.

Moreover, the reverse evolution of budget controllers only contributes to this. A normal occurrence for a SATA SSD in multi-threaded mode is to hit the interface, at least when reading. But XT2 paired with four QLC crystals cannot do that. On the contrary, the speed decreases compared to single-threaded mode.

You can only write quickly to the cache. Moreover, for the new BX500 this is also less fast than is customary. When there is little free space, the cache size decreases and the test file is guaranteed not to fit in it. After which the difference between QLC and TLC becomes clear — no matter how bad the old BX500 was, lagging behind the MX500 by half, but the MT-512 and the new BX500 are 3-5 times slower. And 100 free gigabytes for five hundred is, rather, a lot. They usually buy them when money is really tight (otherwise, at the current price level, it’s better to look at 1-2 TB), and then they use them for the tail and the mane. Therefore, such a picture may even be considered overly optimistic.

In multi-threaded mode we see the same thing. No wonder — the problem here is precisely in the memory, against the background of which the shortcomings of the controllers do not come to light. When you manage to fit into the SLC cache, you can see that the SM2259XT2 is even worse than the SM2259XT. It would seem — much worse. But then there was a knock from the bottom.

What could have changed here? Globally — nothing. What changed? Nothing too. A brilliant match between theory and practice.

The only interesting thing is that the KingSpec MT-512 manages to bypass the old version of the Crucial BX500 even if there is not enough free space. Well, this is another confirmation that the developers at Maxiotek eat their own bread for a reason, managing to squeeze everything possible out of controllers in the budget segment. Silicon Motion would do well to learn this — since the controller was being slightly redesigned anyway, it made sense to do more than just trim the existing one. However, this all has only theoretical significance. For testers it’s interesting, for users it’s almost useless. These SSDs are slow — all three. And when caching cannot fully work, it is very slow. It’s better not to mess with such products — even if they are cheaper than the rest of the “normal” ones.

Comprehensive performance

Indeed, PCMark 10 Storage is one of the best comprehensive benchmarks for assessing storage performance. This suite of tests includes several use cases that cover various aspects of data storage, including booting the operating system, copying files, and other common tasks.

Of all the tests included in PCMark 10 Storage, the Full System Drive test is generally considered the most useful. It provides comprehensive performance assessments, including a wide range of use cases. The remaining tests in this suite are subsets of this full test and may be less representative or less interesting in determining real-world performance.

PCMark 10 Storage also provides latency information, which can be important to understanding the overall performance of the drive under real-world operating conditions. While averaging the metrics to create a single number may be somewhat synthetic, these results still provide useful information for comparing the performance of different storage devices in different use cases.

Let us remind you that the best hard drives in this test slightly exceed 300 points on external tracks, while on internal tracks they fall within 250. Therefore, previously, the second generation WD Green was considered objectively the main brake — knocking out about 360 points. The third version has significantly accelerated, since it is built on the same platform as the WD Blue SA510, aiming for the middle class. And the vacated position was taken by the new BX500. It is clear that it is also faster than hard drives. Even the best. Not to mention laptop hard drives, where even a hundred points is an unattainable ceiling. But objectively — a bad result. Moreover, we repeat, few people spoke well about the old BX500. But it turns out that this was exactly what happened.

Total

Understanding current market trends does not always lead to clear conclusions. Often, while we win in one aspect, we lose in another. For example, the 480GB Crucial BX500 may not be considered the best choice, but the new 500GB BX500 model is even less efficient. Despite this, both versions sell massively.

In light of constant changes in the assortment, it remains either to recognize the choice as limited, or to pay attention to products from other manufacturers. Competition in this market segment forces users to look for better deals in terms of price and quality. But choosing a more affordable product sometimes does not guarantee the optimal quality of materials and components. For example, the Micron memory dies used in Crucial SSDs are typically more reliable than products from smaller manufacturers.

This mixed development in the SSD world reflects market trends. Buyers, in their quest to lower prices, sometimes overlook other important factors needed to make informed decisions. Despite this, the NVMe storage market offers a variety of options for buyers of different interest levels and budgets. On the other hand, for SATA drives there is only one direction of development, and the examples discussed only reflect this trend.