Dev.SN ♥ developers
Sections
Dev.SN
NezSez writes regarding an article in extremetech:
"SATA Express is SATA and PCIe over cables (preserving backwards compatibility) and NVMe is the next improvement of AHCI with much lower latencies by using the PCIe bus/lanes. Both have been developed to improve access to SSD's which have their own processors on-board and can communicate quicker than mechanical drives. The specifications look good (up to 4 times faster and can scale with improvements of PCIe) but analysts suspect it will only be adopted for small form factors.
This discussion has been archived.
No new comments can be posted.
SSDs driving SATA Express and NVMe innovation
|
Log In/Create an Account
| Top
| 1000 moderator points
| 17 comments
| Search Discussion
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
I'm not a robot like you. I don't like having disks crammed into me... unless they're Oreos, and then only in the mouth.
-- Fry
(Score: 5, Interesting) by wantkitteh on Monday March 17 2014, @08:46AM
I'm hoping by "small form factors" they mean this will replace mSATA in laptops, because if they do they have overlooked the fact that we already plug RAM into our desktop motherboards, so there's no reason we shouldn't do the same thing with high performance persistent storage as well. The majority by volume of 2.5" SSDs is the packaging and the cabled SATA connection shouldn't be able to compete with a direct-to-board interface.
(Score: 1) by NezSez on Monday March 17 2014, @09:45AM
I cannot tell whether by "compete" you mean bus contention, or go head-to-head performance wise.
The idea is to use SATA cables to traffic data over the PCIe bus itself, which *is* the direct to board interface, so yes there is bus contention (and TFA notes the limits of using the PCIe 3 lanes as this could be swamped by the video card). Regarding performance, SATAe is much faster than mSATA (with speeds faster than SATA6gbit)and it scales directly with PCIe (a faster bus equals faster performance in a linear sense I assume). They don't think it will be adopted in enterprise data centers because PCIe is entrenched, but, like you point out, I don't see why this wouldn't also be an improvement over eSATA for external RAID and such unless it would indeed thrash the bus.
No Sig to see here, move along, move along...
(Score: 3, Interesting) by mhajicek on Monday March 17 2014, @10:28AM
Can't be sure but I think he was thinking about plugging the ssd straight onto the mobo kind of like a video card.
(Score: 1) by guises on Monday March 17 2014, @10:52AM
Doubtless he was, but you lose flexibility that way and only gain a tiny impedance advantage. I can't imagine that has a significant impact on speed.
(Score: 3, Interesting) by wantkitteh on Monday March 17 2014, @11:22AM
I wasn't thinking so much about impedance as the drastic reduction in the conductor length (millimeters instead of meters) reducing signal noise.
(Score: 3, Interesting) by wantkitteh on Monday March 17 2014, @11:19AM
I was visualizing another row of slots very much like SIMM slots on the motherboard, effectively segregating the market into performance storage (solid state) and volume storage (mechanical). I'm assuming there would also be cartridge-style hot-swappable SSD units as well, much the same way the SATA backplane layout has enabled standard internal drives to be hot mounted into bays without the use of caddies.
(Score: 2) by wantkitteh on Monday March 17 2014, @11:03AM
I was using the word compete in terms of performance and should have known better than to leave it unqualified here of all places.
With the removal of the intermediary SATA from the mass storage data flow route, I would assume that latency would drop and throughput would rise. Given that today's SSDs can push 500MB/s over SATA3, which itself is limited to 600MB/s, I can't see that increase being a lot, although faster storage would most likely be right behind SATAe in the release pipeline. A single PCIe v4.0 lane can handle just shy of 2GB/s, so there would be an argument in favour of designing multilane SATAe for workstations and servers that need to aggregate the full throughput from multiple SSDs. As such I can't see any argument beyond legacy architecture for SATAe failing to penetrate the enterprise market where performance is preferable over GB:#LOCAL_CURRENCY.
(Score: 3, Interesting) by hamsterdan on Monday March 17 2014, @11:59AM
A faster interface will allow SSD manufacturers to parallelize them more (ie, run 32 or 64 NAND chips in something similar to a RAID-0 array, enabling even faster speeds)
Ditch the 2.5 form factor too, I'm guessing my 120GB drive only uses maybe half or less of its casing.