This article is more than 1 year old
XPoint: Leaked Intel specs reveal 'launch-ready' SSD – report
Roses are nice, but SSDs from Chipzilla (if these specs are true) are DRAM-NAND gap-filler
Leaked Intel Optane SSD specs show much higher endurance and lower latency than NAND SSDs and good, but not great, IO performance.
M.2 format 16GB and 32GB caching Optane drives have already been announced, but no add-in-card/SSD-type drives yet.
However, the Chinese TechBang website says it has seen a data sheet for an Intel Optane SSD, the DC P4800X, and says the product is ready for launch.
Ignoring all of Intel's previous XPoint hype and treating this as just another solid-state drive, we can see that its IO performance is not radically better than existing SSDs – in fact it's inferior to many – but its latency is lower than other SSDs, at less than 10 µs, and its endurance is far longer, at 30 drive writes per day.
Its quality of service with 4KB and a random queue depth of 16 is 99.999 per cent, with reads completing in less than 150µs and writes in less than 200 µs.
We don't know its price so cannot judge its price/performance, but we can certainly say it reacts faster than NAND SSDs, has higher write performance and lasts much longer.
The data sheet specifies its capacity as 375GB and provides much performance and interface information. It says it is a half-height, half-length low profile add-in card, for example, and uses an Intel controller with Intel firmware:
Copy of Intel DC P4800X data sheet
We have tabulated some data from this data sheet and compared it to existing Intel, Samsung and WD NAND SSDs also using NVMe interfaces. All the numbers are "up to" values.
Product | Capacity | Media | Random Read IOPS | Random Write IOPS | Seq Read | Seq Write | Latency | Endurance | |
---|---|---|---|---|---|---|---|---|---|
Intel | DC P3608 | 1.6, 3.2, 4TB | 20nm HET MLC NAND | 850,000 | 150,000 | 5GB/sec | 3GB/sec | 20μs read/ write | Up to 21.9PB writen. Up to 3DWPD over 5 years |
DC D3700 | 800GB, 1.6TB | NAND MLC | 470,000 | 95,000 | 2.1GB/sec | 1.5GB/sec | 10DWPD @ 800GB | ||
DC P4800X | 375GB | 3D Xpoint | 550,000 | 500,000 | 2.4GB/sec | 2GB/sec | <10 µs | 30 DWPD. 12.3PB written. | |
Samsung | PM1725a | 1.6, 3.2, 6.4TB | 3D V-NAND 48-layer TLC | 1,050,000 | 160,000 | 6.4GB/sec | 2.9GB/sec | Read/write 90/20 µs. | 5 DWPD 5 years |
960 PRO | 512GB, 1TB | V-NAND 48-layer 21nm MLC | 440,00 | 360,000 | 3.5GB/sec | 2.1GB/sec | 1.2/2PB written or 5-year warranty | ||
WD | Black | 256, 512GB | 15nm TLC NAND | 170,000 | 134,000 | 2.05GB/sec | 512GB/sec | 160TB written, 5-year warranty |
A glance at the table shows that Samsung's PM1725a has a 1,050,000 random read IOPS rating, making the XPoint SSD's 550,000 look slow. However, Optane's 500,000 random write IOPS looks good, being double or triple other SSDs. Several drives in the table exceed the Optane product's sequential IO bandwidth; we can say it's respectable but not in the lead. Intel's own DC P3608 exceeds it comfortably.
The Optane drive's endurance is outstanding, up to 10 times better in drive writes per day compared to the other drives in our table. This is no device whose weak write performance/endurance means it has to be used for read-centric applications.
Is this leak genuine? Tom's Hardware found a conformity certificate on Intel's website, which is no longer available.
We have asked Intel about this DC P4800X and expect a firm "no comment". If we get anything else, we'll let you know. ®