SSD - When we hit the laws of diminishing returns

[taltamir]

Actually they reach the 10gbits with one cable and one diode (and one wavelength) if I read their research site and the AT article correctly. I assume you got the idea of the 4 lasers from one picture on the research site where they state "# Light Peak module with four fibers each capable of carrying 10Gb of data per second."

actually, no i didn't. I got it from some tech articles that said intel will be multiplexing 4 different wavelength lasers (@2.5gb each) to reach the 10GB figure.
I don't recall exactly which site I read it at though, and it could be wrong since we are currently in the "rumor" phase.

 

[Voo]

Well from the AT article:
"Intel didn't need to use WDM (Wavelength Division Multiplexing) to achieve 10 Gbps as demonstrated, and maintains that although WDM could be used in the future to increase bandwidth, there are other ways to scale to 100 Gbps"
(src).
So no WDM.

Also "Light Peak module with four fibers each capable of carrying 10Gb of data per second." states rather clearly in my opinion that each fiber is capable of 10gbps. Iirc there were some 50gbps tests that used WDM with 4 channels.

 

[Makaijin]

I don't know why everyone is going gaga over SSD. They're storage capacity is much to small to make them useful. When we start seeing 1TB SSD that don't cost an arm and a leg then they will become mainstream. Until then 1-2TB regular hard-drives are the only way to go at the moment.

I think you obviously haven't seen an SSD in action. The advantages of SSD drives is for it's speed not for it's mass storage.

Last year madshrimps timed 2 identical dell laptops, one with a typical 7200 rpm HDD and the other a OCZ Vertex and filmed their boot times. They mirrored the hard disk contents, put them side by side, turned them both on at the same time and filmed it. After booting it also autoloaded a bunch of suites like Office and stuff.

Check the article yourself http://www.madshrimps.be/?action=getarticle&articID=923 especially the youtube vids.

If that isn't enough to make you go gaga I honestly don't know what will.

 

[ksec]

I just read Anand replying to a comment somewhere ( i cant find it anymore )
Once you switch to a decent SSD ( Sandforce based ) most of your software loading time will be effectively CPU bound rather then I/O bound.

Good time for SandyBridge then.........

 

[taltamir]

Once you switch to a decent SSD ( Sandforce based ) most of your software loading time will be effectively CPU bound rather then I/O bound.

That... assumes an AMAZINGLY complex (and thus rarely used) and supremely efficient fractional loading software design...
The VAST majority of software out there performs loading as such:
A. Read data into RAM
B. When all data is DONE reading into ram, perform CPU calculations.

This makes total time to load A + B... there is no "bottleneck" in this kind of operations. Bottlenecks tend to happen in more concurrent operations (ex: video encoding, video decoding, file compression, etc) where data is loaded at the same time as it is being worked on. (or rather, split into bite sized chunks which are processed individually as soon as they are loaded)... ex: while encoding/decoding frame 1001 it can be concurrently loading the data for future frames.

This is actually WORSE in most games... NWN2 for example:
It decompress' a zip file on HDD to a temp directory on HDD using a SINGLE CORE ONLY! (concorrent operation thankfully), the resultant files are then loaded into RAM, and only then processed.
so the process is:
A. Read zip file, concurrently processing it with CPU, concurrently writing resultant files to HDD.
B. Read resultant files from HDD into RAM.
C. Process said files.

That being said, its LIKE a bottleneck because even if the operations are concurrent, with standard HDD the CPU portion of the work is a fraction of the time of the HDD portion... while with a top quality SSD the SSD portion is a fraction of the time of the CPU...
so going from 1 minute HDD work and 10 seconds CPU work to 1 second SSD work and 10 seconds CPU work would make it seem as if there was a bottleneck, and that it "shifted". But for a bottleneck to exist the work must be concurrent (so that in case two, for ever 10 seconds the CPU works, the SSD works 1 second and waits 9). That only happens in certain scenarios (video encoding, video decoding, file compression, file decompression, etc).

 

[ksec]

Which basically answer the question, we are no longer bottlenecked by this Gen SSD, and we need software and CPU to improve to get further performance difference...

 

[taltamir]

Which basically answer the question, we are no longer bottlenecked by this Gen SSD, and we need software and CPU to improve to get further performance difference...

only if you actually had a specific scenario such as: "so going from 1 minute HDD work and 10 seconds CPU work to 1 second SSD work and 10 seconds CPU work"
For many scenarios its a sizeable enough amount of work for both that you will notice real improvements with either upgrade.

 

[ModestGamer]

If i hit the power switch on my PC and it basically instantly is ready for use (without hibernate) then I'm fully satisfied.
Yes SSD are much faster and boot times get pretty short compared to normal hdd's but isn't instant-on what we want in the end?

My pc now probably boots faster than my mobile phone. And it don't have one of these "I'm also a coffee machine" phones.

my pc boots in 12 seocnds with a 5400rpm drive on a p3 system. well thats one of my net stations around the house.

Alot of the boot performance is driven by the os itself.

 

[ModestGamer]

Which basically answer the question, we are no longer bottlenecked by this Gen SSD, and we need software and CPU to improve to get further performance difference...

you can hope Ms adress's the situation but to date the less then stellar performance of those OS products is lacking and has remianed so for a very very long time.

 

[Emulex]

half duplex sata -> consumer drives
dual ported sas (full duplex,lvd, switchable) -> enterprise mlc,slc

it's time to ditch sata and at least go to 6gbps sas DP

 

[ksec]

Couldn't we fix this problem by having two Half Duplex SATA connection to the SSD?

While creating another separate connection like HSDL from OCZ would be ideal. It is not very economical to do so.

 

[taltamir]

you can hope Ms adress's the situation but to date the less then stellar performance of those OS products is lacking and has remianed so for a very very long time.

Last I checked, MS's windows 7 was the first OS to get support for TRIM... followed very closely by linux.

Linux TRIM support: Feb 2010
Windows 7 release with built in TRIM: 22 October 2009
RTM of win 7: 22 July 2009
Win 7 implementation of trim?: early in the beta process.

Anyways, I give MS a harsh time myself, because they do a lot of bad things... but there is no reason to take a dump on them when they are actually doing right. And MS was first to market with TRIM support which provided stellar performance to SSD adopters.

 

[bunnyfubbles]

I'd just like to add that while SSDs offer untouchable speed by disk drives when it comes to random access, their real world sustained transfer speeds can easily be infringed upon by disk drives, especially performance RAID setups.

Working with large video files I've noticed some of my more humble disk drives actually rivaling my SSD in sustained write speed.

Personally I'd love to see at least 500+MB SSD drives @ 500+MB/sec sustained reads/writes in real usage scenarios become the standard, although it looks like we're still a bit away from achieving that.

True, a common user can easily tell the difference from the jump from HDD to an average SSD, but that doesn't mean things like Revo drives are useless. Heck, the same could be said of CPUs; most users don't need more than a fast dualcore and wouldn't be able to tell the difference between such a CPU and a quad. However for those that can actually utilize the power can see huge benefits from such things.

 

[Voo]

I'd just like to add that while SSDs offer untouchable speed by disk drives when it comes to random access, their real world sustained transfer speeds can easily be infringed upon by disk drives, especially performance RAID setups.

Well yeah obviously sustained writes with several modern drives in Raid0 are faster (and if it's just because they're not bound by a single SATA3gbps link), nobody disputes that.
Although I'm not sure WHY that'd be so interesting.. sequential r/w is mostly needed for mass storage stuff and SSDs are still way too expensive for that anyhow.

I don't see why we've got to decide between two technologies and not just use both, HDDs are perfectly fine for mass storage where random access is unimportant, so I couldn't care less about using a SSD for that or not, even if they weren't much more expensive.

 

[taltamir]

I'd just like to add that while SSDs offer untouchable speed by disk drives when it comes to random access, their real world sustained transfer speeds can easily be infringed upon by disk drives, especially performance RAID setups.

A single SSD is faster than ANY single spindle drive on the market in sustained sequential reads.

An ARRAY of RAID0 spindle drives can outstrip a SINGLE SSD... but so what... an equal sized array of SSDs would beat the spindle drives by that much more.
And the hypothetical here is "when an SSD costs less then a same sized spindle drive"

Personally I'd love to see at least 500+MB SSD drives @ 500+MB/sec sustained reads/writes in real usage scenarios become the standard, although it looks like we're still a bit away from achieving that.

The C300 SSD already MAXES OUT the SATA2 interface, no spindle drive can do that. It supports SATA3 which allows it to go 100MB/s faster than it can on SATA2 (total speed of 340 MB/s sequential read speed)
Your spindle drives do not outperform SSDs... you spindle drive RAID0 array might outperform ONE SSD... but that just means you should use a RAID0 SSD array.

 

[MagickMan]

Use S3 sleep. I know that the HTPC crowd has been doing it for a while and notebooks do it, but I only recently fiddled with it on desktops to great success. I've tried it on several systems (socket 1156, 775 and an Atom ITX) and they draw around 1-2W from the wall in sleep according to my Power Angel (like a Kill-A-Watt). Waking it up basically takes as long as your monitor switching modes or turning on.

This. I use sleep all the time and almost never completely shutdown my machine. Just tap the space bar and it's instantly awake and ready to go.

 

[ryrynz]

I don't know why everyone is going gaga over SSD. They're storage capacity is much to small to make them useful. When we start seeing 1TB SSD that don't cost an arm and a leg then they will become mainstream. Until then 1-2TB regular hard-drives are the only way to go at the moment.

For good reason actually.. only a few years later they were VERY much mainstream, it didn't take 1TB prices to fall to non arm and leg prices, many people including myself started on 128 went to 256, 512 and 1TB models and beyond. This didn't quite pan out how you expected, the market boomed.. and now you know why everyone went gaga huh?

 

[UsandThem]

For good reason actually.. only a few years later they were VERY much mainstream, it didn't take 1TB prices to fall to non arm and leg prices, many people including myself started on 128 went to 256, 512 and 1TB models and beyond. This didn't quite pan out how you expected, the market boomed.. and now you know why everyone went gaga huh?

Nobody in this thread went "gaga" in over 10 years.