Vik

Intel Announces SSD DC P4501 Low-Power NVMe SSD With 3D NAND
  • Comments are Closed

Intel Announces SSD DC P4501 Low-Power NVMe SSD With 3D NAND

A month ago, Intel unveiled a new generation of enterprise SSDs based on their 3D TLC NAND and powered by a new generation of Intel NVMe SSD controllers. Intel’s first NVMe controller—introduced with the SSD DC P3700—had a great run, but it’s been showing its age for a while now. The new controller used on the P4500 and P4600 enables higher performance despite the new drives using 3D TLC NAND compared to the P3700’s 20nm planar MLC, and despite the newer controller having only 12 flash channels instead of 18.

More significantly, Intel’s second-generation NVMe controller is much smaller and more power-efficient, allowing it to be used in product segments that Intel’s first generation controller literally couldn’t fit in. Intel’s new SSD DC P4501 is a lower-power take on their new generation of NVMe technology, putting Intel controllers into M.2 and 2.5″ 7mm U.2 SSDs for the first time. With capacities from 500GB to 4TB, the P4501 delivers somewhat lower performance than the P4500 and P4600, but with reduced power consumption and substantially higher density.

Intel has previously shipped the SSD DC P3100 M.2 SSD, but despite using the same 3D TLC, it was not at all in the same league as the P4501. The P3100 is based on the same platform as the consumer Intel SSD 600p, with a Silicon Motion controller, poor sustained write speeds and no power loss protection. The P3100 was pitched as a boot drive for servers, but the P4501 can be a serious workhorse. The P4501 supports the same advanced feature set of its bigger siblings, including full power loss protection, NVMe Management Interface, and support for up to 128 queues for optimal performance even in systems with the highest CPU core counts. The P4501 even comes with a slightly higher endurance rating than the P4500.

Intel NVMe Datacenter SSD Comparison
  P4500 P4600 P4501
Form Factor PCIe HHHL or 2.5″ 15mm U.2 M.2 22110 or 2.5″ 7mm U.2
Interface PCIe 3.1 x4 NVMe 1.2
Memory Intel 384Gb 32-layer 3D TLC
Capacities (TB) 1, 2, 4 TB 1.6, 2, 3.2 TB (U.2)
2, 4 TB (HHHL)		 500GB, 1TB, 2TB, 4TB (U.2 only)
Sequential Read (64kB) (up to) 3290 MB/s 3280 MB/s 3200 MB/s
Sequential Write (64kB) (up to) 1890 MB/s 2100 MB/s 900 MB/s
Random Read (4kB) (up to) 710k IOPS 702.5k IOPS 360k IOPS
Random Write (4kB) (up to) 68k IOPS 257k IOPS 46k IOPS
Maximum Power Read 10.9 W 9.9 W 8–12.5W (U.2) 6–8.25W (M.2)
Write 18.3 W 20.7 W
Write Endurance Random workload 0.7 DWPD 2.9 DWPD 1 DWPD
Sequential workload 4.62 DWPD 4 DWPD 3 DWPD
Warranty 5 years 5 years

The M.2 version of the P4501 will provide capacities up to 2TB on a 110mm long card, larger than normal for consumer M.2 SSDs but typical for enterprise SSDs that include power loss protection capacitors. The 2.5″ U.2 version uses the same 7mm thickness as consumer SSDs and many of the more value-oriented enterprise SSDs, compared to the 15mm thick form factor that Intel’s previous enterprise SSDs have used for thermal dissipation reasons even when the extra volume was not necessary for higher storage capacity. The U.2 version of the P4501 will provide capacities up to 4TB.

The P4501 will support tunable power limits: the U.2 version can be capped at 8W, 10W or 12.5W while the M.2 version will support limits of 6W or 8.25W. Idle power is not a priority for datacenter SSDs, but the U.2 version is specified for less than 5W at idle and the M.2 version is specified for less than 3W.

Intel’s first-generation NVMe controller came to the consumer market in the SSD 750, the first consumer NVMe SSD. It raised the bar for consumer SSD performance, but as soon as truly client-oriented solutions like the Samsung 950 Pro hit the market, the SSD 750’s inconveniently large form factors, high power consumption and lower peak performance started to erode its cachet. The new Intel SSD DC P4501 shows that Intel is in a position to offer a new generation of high-end client NVMe SSDs that will be much better fit for today’s market. With new firmware optimized for client workloads and possibly with cost savings from dropping the power loss protection or using Intel’s upcoming 64-layer 3D TLC NAND, Intel could have a serious contender for the consumer market in the near future.

  • Comments are Closed

Test Driving Futuremark’s PCMark 10 Benchmark

​Futuremark is a well-respected provider of computer benchmark applications. Their PCMark and 3DMark benchmarks have been around for almost 20 years, and provide a good indication of the system performance for various workloads. Today, Futuremark is launching PCMark 10, their seventh major update to the PCMark series of benchmarks first launched in 2002. PCMark 10 builds upon the PCMark 8 platform, adds a few workloads and streamlines the rest in order to present a vendor-neutral, complete, and easy-to-use benchmark for home and office environments. Futuremark provided us with a preview copy of the benchmark, and we took it for a test drive. This piece presents some of our impressions with the benchmark, and points out areas where it excels, and where it could do with some improvement.

Thecus Announces $130 and $200 Entry-Level NASes For Home Users
  • Comments are Closed

Thecus Announces $130 and $200 Entry-Level NASes For Home Users

This month at Computex, Thecus announced two new entry level-level NASes for consumers. The new N2350 and N4350 are budget-priced NAS products designed for the basic storage tasks. The systems feature a low-power Marvell Armada dual-core CPU and 1GB of DDR4 system memory. The N2350 features two drive bays with support for RAID 0, 1, and JBOD. The larger N4350 increases the capacity to four drive bays and adds support for RAID 5, 6, 10.

Slinging NAS appliances to home users is a difficult challenge. More than half of potential customers don’t know the category exists, and most of the others think the systems are small form-factor PCs with a few extra drive bays. So offering basic boxes at low prices is an important aspect to expanding the bottom of the market, where consumers may not be interested in paying a high price tag for a storage appliance.

Both NASes ship with the company’s latest ThecusOS 7.0 graphical operating system. And while the underlying CPU isn’t going to be especially powerful, it’s enough to offer additional functionality such as external (Internet) file access. Not to mention additional applications/functions that can be installed through the OS itself.

The Thecus N2350 ($130) and N4350 ($200) have already started to ship to distributors. Both provide a low-cost entry point for users to experience NAS without spending a lot of money, and both carry a 2-year warranty.

Chris Ramseyer contributed to this report

AMD Announces Radeon Pro WX 3100 & WX 2100: Entry-Level Polaris Pro Cards
  • Comments are Closed

AMD Announces Radeon Pro WX 3100 & WX 2100: Entry-Level Polaris Pro Cards

A couple of months back AMD announced their third (and presumably final) Polaris GPU, the diminutive Polaris 12. Designing for early-level video cards and laptops, in the consumer space it’s the heart of the Radeon RX 550. Now AMD is now bringing their smallest 14nm GPU to the professionals with the announcement of the Radeon Pro WX 3100 and WX 2100.

The latest Radeon Pro WX cards serve to finish fleshing out the professional lineup, replacing the old Oland-based FirePro W2100. Like its predecessor the new cards are primed for the entry-level market, getting desktop users access to AMD’s pro driver set and certification for a minimal cost, but with performance to match.

AMD Workstation Video Card Specification Comparison
  WX 4100 WX 3100 WX 2100 W2100
Stream Processors 1024 512 512 320
ROPs 16 16 16 8
Boost Clock 1201MHz ~1220MHz ~1220MHz 630MHz
Memory Clock 6Gbps 6Gbps 6Gbps 1.8Gbps
Memory Bus Width 128-bit 128-bit 64-bit 128-bit
VRAM 4GB GDDR5 4GB GDDR5 2GB GDDR5 2GB DDR3
TDP 50W 50W 35W 26W
Display Outputs 4 3 3 2
GPU Polaris 11 Polaris 12 Polaris 12 Oland
Architecture Polaris Polaris Polaris GCN 1.0
Manufacturing Process GloFo 14nm GloFo 14nm GloFo 14nm TSMC 28nm
Launch Date Q4 2016 06/2017 06/2017 08/2014
Launch Price (MSRP) $299 $199 $149 N/A

Both cards are based on the same Polaris 12 GPU. In fact the only difference between them in terms of specifications is the amount of memory bandwidth and the cards’ TDP. Otherwise the two cards are almost identical.

By the numbers then, the Radeon Pro WX 3100 is a “full fat” Polaris 12 card. So all 8 CUs are enabled, as is the 128-bit GDDR5 memory bus. AMD pegs the performance of the card at 1.25 TFLOPS, which works out to a GPU boost clock of about 1220MHz. The card is paired with 4GB of GDDR5 memory, clocked at 6Gbps, for a total memory bandwidth of 96GB/sec.

Comparatively speaking then, the Radeon Pro WX 2100 retains the same core GPU configuration of 8 CUs and a 1220MHz boost clock, but it drops half of the memory bus. The end result is the same theoretical compute/texturing/geometry performance, but only 2GB of GDDR5 and a resulting 48GB/sec of memory bandwidth to do it with.

Otherwise, as you’d expect for entry-level cards, the power consumption of both cards is rather low. The WX 3100 peaks at 50W, and the WX 2100 is even lower, at just 35W. I do have some doubts that AMD could shave 15W just by taking away 2GB of memory, so it may be that the WX 2100 is a bit more prone to throttling.

In any case, the low TDP means that both cards are being offered in the same single slot low-profile design. Though notably, these aren’t AMD’s only low-profile cards, as the more powerful WX 4100 is also a low-profile configuration.

What this low-profile layout affords the user in terms of display outputs then is quite interesting. Like their other pro cards AMD has gone for an all DisplayPort configuration, but rather than using a single port type, AMD is using a mix of full-size and mini DisplayPorts. As a result both cards offer 2x mini DisplayPort 1.4, and another full-size DisplayPort 1.4. Going with a pure DisplayPort configuration is what we’d expect to see, as it allows AMD to drive home the fact that the card can potentially drive up to three 8k monitors. Otherwise the odd mix of ports appears to be due to the fact that the Polaris 12 GPU can only supply 3 display headers (though it has 5 display controllers internally, should you ever use MST).

AMD’s target market for these cards then is very much the entry-level market, where users need access to the Radeon WX driver and support stack, but not large amounts of performance. So less oil & gas, and more simple media and CAD. These new cards will be going up against NVIDIA’s existing Quadro P400 and P600 cards for the low-end of the market.

Wrapping things up, the WX 3100 and WX 2100 should be available from AMD’s usual distributors later this month. The WX 3100 will run for $199, while the WX 2100 will go for $149.