The SSDs available in the market are becoming faster everyday, and the SATA-600 interface is becoming a performance-limiting factor. Because of this, newer high-end models are using the PCI Express interface. In this review, we will compare two high-end SSDs from Kingston: the HyperX Savage, which uses the 2.5” form factor and SATA-600 (600 MiB/s) interface, and the HyperX Predator, which uses the PCI Express 2.0 x4 interface (2 GiB/s), both with 480 GiB capacity. Let’s see what is the performance difference between these two models.
The HyperX Savage is the most high-end model from Kingston using the 2.5-inch form factor. It uses a SATA-600 interface and is available with capacities from 120 GiB up to 960 GiB.
The HyperX Predator, on the other hand, is an SSD that uses the M.2 2280 (80 mm long) form factor and a PCI Express 2.0 x4 interface. It can be bought as a simple M.2 card, to be installed in an M.2 slot, or with a PCI Express x4 adapter card, needed if your motherboard does not offer an M.2 slot with PCI Express x4 connection. This product is available in 240 GiB and 480 GiB capacities.
In our tests, we will compare the performance of these two units, both with 480 GiB of capacity.
Before proceeding, we highly suggest that you read our “Anatomy of SSD Units” tutorial, which provides all the background information you need to know about SSDs.
In the table below we compare the units tested.
|Manufacturer||Model||Model #||Nominal Capacity||Form Factor||Interface||Price|
PCI Express 2.0 x4
We researched the prices on the day that we published this review. In the table below, we provide a more in-depth technical comparison between the two drives.
|Kingston HyperX Predator||Marvell 88SS9293||2 x 512 MiB DDR3L-1600 Kingston D2516EC4BXGGB||8x 64 GiB Toshiba TH58TEG9DDKBA8H|
|Kingston HyperX Savage||Phison PS3110-S10||512 MiB DDR3L-1600 Nanya NT5CC256M16CP-DI||16x 32 GiB Kingston FQ32B08UCT1-C0|
[nextpage title=”The HyperX Predator 480 GiB”]
The Kingston HyperX Predator 480 GiB comes in a small box. Besides the M.2 SSD unit installed in a half-height PCI Express 2.0 x4 expansion card adapter, there is a bracket to install the card in half-height cases, a manual, and a case sticker.
The product also comes with a serial number for you to download and install the Acronis True Image HD software, which allows you to copy all the contents of your boot drive to the new SSD, in order to use it as a boot drive without reinstalling the operating system.
Figure 3 shows the solder side of the adapter card. Keep in mind that this card can also be installed in PCI Express x8 or x16 slots.
The M.2 card can be easily removed from the adapter card. If your motherboard has an M.2 slot with PCI Express 2.0 x4 connection, you can install the HyperX Predator directly in this slot. Remember, however, that some motherboard has an M.2 slot that is only compatible with SATA connection; in this case, the HyperX Predator will not work, and you must use the adapter card.
[nextpage title=”The HyperX Predator 480 GiB – Components”]
Removing the sticker that covers the component side of the M.2 card, you can see four flash memory chips, the controller chip, and one DDR3L memory chip used as a buffer.
At the other side of the card there are another four flash memory chips and another DDR3L chip.
The controller used by the HyperX Predator is the Marvell 88SS9293.
The unit brings two DDR3L-1600 chips with 512 MiB capacity each, model Kingston D2516EC4BXGGB, which work as data buffer.
The NAND flash memory chips are from Toshiba, model TH58TEG9DDKBA8H.
[nextpage title=”The HyperX Savage 480 GiB”]
The HyperX Savage 480 GiB also comes in a small box, shown in Figure 10. Besides the SSD itself, there is also an adapter to install the unit on a 3.5-inch bay and a frame to fit the drive in laptops that require a 9.5 mm thick drive, as this unit is only 7 mm thick. The Savage also comes with a case sticker and a serial code to redeem the Acronis True Image HD software.
The HyperX Savage comes in a metal casing. On the bottom cover, there is a sticker with the information about the unit, as seen in Figure 12.
Opening the drive, you see the printed circuit board. On the component side, you will find the controller chip, one DDR3 memory chip that works as a buffer, and eight flash memory chips.
[nextpage title=”The HyperX Savage 480 GiB – Internal components”]
On the solder side, there are another eight flash memory chips.
The controller used by the HyperX Savage 480 GiB is the Phison PS3110-S10.
There is a DDR3L-1600 512 MiB memory chip from Nanya, model NT5CC256M16CP-DI, which works as a data buffer.
The NAND flash memory chips are Kingston FQ32B08UCT1-C0 parts.
[nextpage title=”How We Tested”]
During our testing procedures, we used the configuration listed below. The only variable component between each benchmarking session was the SSD being tested.
The HyperX Predator was installed at a PCI Express 3.0 x16 slots using the adapter card, while the HyperX Savage was connected to a SATA-600 port.
- Processor: Core i7-5960X @ 3.5 GHz
- Motherboard: ASRock Fatal1ty X99M Killer
- Memory: 16 GiB DDR4-2400/PC4-19200, four G.Skill F4-2400C15Q-16GRR 4 GiB modules
- Boot drive: Kingston M.2 SM2280S3 de 120 GiB
- Video display: Samsung U28D590D
- Power Supply: Corsair CX750
- Case: NZXT Phantom 530
- Operating System: Windows 7 Home Basic 64-bit using NTFS File System
We adopted a 3% error margin in our tests, meaning performance differences of less than 3% can not be considered meaningful. Therefore, when the performance difference between two products is less than 3%, we consider them to have similar performance.
[nextpage title=”Compressible Data Test”]
As mentioned in the previous page, me measured the performance of each drive using the CrystalDiskMark 4 program. In this version, the software performs sequential and random reading and writing with 4 kiB blocks, first with a queue depth (QD) of 32, and then with a QD of one. So, it does not only test the performance with a single task, but also the performance with simultaneou read and write requisition, mimicking a scenario such as the one found in database servers.
Also keep in mind that CrystalDiskMark 4 uses a different measuring methodology from CrystalDiskMark version 3, so data obtained with different versions are not comparable.
First, we ran CrystalDiskMark in “All 0x00 Fill mode”, where the data writen on the drive are only zeros, in order to measure the SSD performance with compressible data.
On the sequential read test with a queue depth of 32, the Predator was 178% faster than the HyperX Savage.
On the sequential write test with a queue depth of 32, the Predator was 89% faster than the Savage.
On the random reading test with 4 kiB blocks and QD 32, the HyperX Predator was 38% faster than the Savage.
On the random writing test with 4 kiB blocks and QD 32, the Savage was 11% faster than the Predator.
On the simple sequential read test (i.e., queue depth of one), the Predator was 28% faster than the Savage.
On the simple sequential writing test (i.e., queue depth of one), the Predator beat the Savage by 105%.
On the simple random reading test, the Predator was 68% slower than the HyperX Savage.
On the random write test with 4 kiB blocks, the Predator was 11% slower than the Savage.
[nextpage title=”Incompressible Data Test”]
On this test, we left CrystalDiskMark at standard mode, with random, non-compressible data.
On the sequential reading test with a queue depth of 32 and non-compressible data, the HyperX Predator was 180% faster than the HyperX Savage.
On the sequential write test with a queue depth of 32, the Predator was 91% faster than the Savage.
On the random read test with 4 kiB blocks and a queue depth of 32, the HyperX Predator was 25% faster than the HyperX Savage.
On the random write test with 4 kiB blocks and queue depth of 32, the Savage was 11% faster than the Predator.
On the simple sequential read test (i.e., queue depth of one), the Predator was 71% faster than the Savage.
On the simple sequential write test (i.e., queue depth of one), the Predator was 104% faster than the Savage.
On the simple random read test with 4 kiB blocks, the Predator was 24% faster than the HyperX Savage.
However, on the simple random write test, the Predator was 7% slower than the Savage.
As we mentioned at the beginning of this article, you need to keep in mind that the Kingston HyperX Savage 480 GiB and the HyperX Predator of the same capacity are not direct competitors. Besides using different form factors, the price of the products is also very different. While the HyperX Savage is targeted to the home market, being a high-end “traditional” 2.5-inch SATA-600 (600 MiB/s) SSD, the HyperX Predator is an SSD aimed on the performance segment, for enthusiasts, high-performance workstations, and servers, using a PCI Express 2.0 x4 (2 GiB/s) connection and an M.2 form factor, with an optional adapter card to be installed in a PCI Express slot.
That said, we could easily see that the performance of these products is completely different. Even though the HyperX Savage 480 GiB provides an excellent performance, the HyperX Predator 480 GiB presents a performance level that is, in some cases, three times higher.
Some points are clear in our tests. First, while the HyperX Savage presents different performance levels with compressible and non-compressible data, the Predator had a consistent performance on both tests, which proves that its controller does not rely on data compression to achieve higher performance.
Another important result from our benchmarkings is that the HyperX Predator achieved a really impressive performance in the test with a queue depth of 32, which suggests it shows all its potential when it is under a heavy workload. So, applications that make use of smultaneous read and write requests will benefit more from this unit.
In other words, the HyperX Predator 480 GiB has an exceptional performance, but only storage-intensive applications will take advantage of all of its potential. For applications where the amount of simultaneous read and write requests is low or average, the HyperX Savage 480 GiB is a better choice, thanks to its lower price point, yet still maintaining a high performance level.