Solid State Disk again...

[Crystal]

Just a link to a new test : http://www.hyperossystems.co.uk/07042003/HD4tests.htm

I think those tools could really be revolutionary for sampling technologies.

Unfortunately nothing appends in this field.


What a pity.

[Tobias Erichsen]

Hi,

just found a manufacturer who has just announced solid-state-drives
with up to 160GB capacity...

http://www.adtron.com/newsroom/25fb-Solid-State-Disk.html

Tobias

[Crystal]

Hi Tobias, thanks for the link.

That flash technology seems twice slower than the DDR one in terms of MB/s, but permits a large storage.

I’m curious about the seek time.

And the price...


That drive could possibly be still pertinent to buffer our samples permanently.

(And delivering us from loading times and networking... my dream...)

[janila]

The sustained transfer rate doesn't practically matter at all as long as it's comparable to a normal hard drive. The seek times and burst rates make these products interesting. We would have to wait for the sampler manufacturers to catch up as samplers don't offer sample buffers small enough to be able to load huge templates on one or two gigs of RAM.

Here's a link (http://www.tomshardware.com/2006/09/20/conventional_hard_drive_obsoletism/) to Tom's Hardware test of the Samsung 32GB solid state hard drive from last september. It gets interesting on the seventh page. The performance is stellar for our needs but 32 gigs certainly isn't enough and the price doesn't make you smile.

[cmdratz]

Hi,

just found a manufacturer who has just announced solid-state-drives
with up to 160GB capacity...

http://www.adtron.com/newsroom/25fb-Solid-State-Disk.html

Tobias
Anyone know where we can buy a solid state drive of 80gb or more, right now though?:n:

[Crystal]

Hi Janila, interesting link.

Hi Chris

About capacities :

There is two approaches : SSDs are so fast that if we use them to stream samples we can reduce the size of samples’ buffers that we load into the motherboard’s RAM. And so we can load more of them in it.

But it’s volatile. And needs huge SSDs.


My idea is to use SSDs to put on them only samples’ buffers, and still stream classically with hard drives.

Then buffers are permanent. And actual SSDs size enough for our actual symphonic libraries.


Power on your PC.

Open your software.

And play all instruments you want instantly...

On a single PC.



That’s what I call a productive machine...


It would be interesting to find a comparison between motherboards’ RAM and actual RAM drives access time (I’m not sure flash drives can fight).

HyperDrive4 shows a 0,001 millisecond seek time in their test.

Anyone knows the seek time of actual motherboards’ RAM ?

[Craig F]

Anyone knows the seek time of actual motherboards’ RAM?
Technically, there is no "seek" time. Memory chips have direct access to any location. The spec you need to look at is "access time". It is in tens of ns (nanoseconds).

DRAM uses page/burst methods to speed access to adjacent memory cells. DRAM is not as fast the CPU, so there is latency involved. The first memory access has higher latency because subsequent locations are automatically decoded. This is because memory is typically accessed in blocks. So this is a cheap way of speeding up overall access.

Bottom line is, it’s FAST.

[cmdratz]

My requirement is reliability: for instance, I have a well-cooled 400GB 133GB per platter Seagate Barracuda 7200.8 ata-100 sample drive and I installed my main sample library (32GB) on that drive before anything else. Now that the first platter is filling up with other stuff, I am getting dramatic dropouts at the same latency where I had none before. For instance, I used to be able to load 8 instruments and play them all at once multiplied over many keys on my controller keyboard with no dropouts; now I often have trouble with even 1 normal-sized instrument having dropouts when I play 1 key at a time. The drive seems to do all other tasks just fine, just not the sample streaming. I am now freezing one instrument at a time with disk streming disabled (hardly my original plan, but it works okay as I have 3GB of RAM). If I invest in a different sample drive, I want something solid (no pun intended) that I won't have to worry in any way about primarily for my existing 32GB library, and yet not too small, so that I can also run something like the Sonivox Complete Symphonic Collection off of it later. RAM is much more reliable and for much longer under heavy strain than a plattered harddisk, hence my interest in a solidstate hard drive of about 120GB. Then I could use the 400GB drive as a mass storage drive, next to my AID-0 double 40GB system drive, and of course, the 120GB solidstate sample drive.:|:

[Crystal]

Hi Craig.


Technically, there is no "seek" time. Memory chips have direct access to any location. The spec you need to look at is "access time".


Yes, I know, I just copy/past the term from the HyperDrive test, which is inappropriate for RAM, you’re right.

Nanoseconds ?

So 0,001 millisecond is... 1000 nanoseconds, am I wrong ?


I finaly find here (thanks to specify it’s about nanoseconds... I can google it, now...) that RAM « vary from 9 nanoseconds to 70 nanoseconds » : http://searchmobilecomputing.techtarget.com/sDefinition/0,,sid40_gci523855,00.html

100x faster...

Well it seems that Sata bottleneck won’t realize my dream for now...




I have a well-cooled 400GB 133GB per platter Seagate Barracuda 7200.8 ata-100 sample drive and I installed my main sample library (32GB) on that drive before anything else. Now that the first platter is filling up with other stuff, I am getting dramatic dropouts at the same latency where I had none before.

There’s something strange here. There’s no reason that your main 32GB library bugs because you installed something else on your drive, if it’s your problem... And a 120 GB library won’t drop anyway on a 400GB drive if it doesn’t with 32GB. Performances collapse on plate drives (except SCSIs), but not so fast... I personally never use the last 30% of my samples or audio drives, which are critical.
You really have some room with a 400GB drive.

[Craig F]

So 0,001 millisecond is... 1000 nanoseconds, am I wrong?

Correct.

1 ms = 1,000 us = 1,000,000 ns
or in Euro speak
1 ms = 1.000 us = 1.000.000 ns

ms = milliseconds 10^-3
us = microseconds = 10^-6
ns = nanoseconds = 10^-9

[cmdratz]

Hi Craig.
I finaly find here (thanks to specify it’s about nanoseconds... I can google it, now...) that RAM « vary from 9 nanoseconds to 70 nanoseconds » : http://searchmobilecomputing.techtarget.com/sDefinition/0,,sid40_gci523855,00.html

Additionally, some graphics memory (fast GDDR4) runs at as low as 1 nanosecond.

There’s something strange here. There’s no reason that your main 32GB library bugs because you installed something else on your drive, if it’s your problem... And a 120 GB library won’t drop anyway on a 400GB drive if it doesn’t with 32GB. Performances collapse on plate drives (except SCSIs), but not so fast... I personally never use the last 30% of my samples or audio drives, which are critical.
You really have some room with a 400GB drive.
I have a little over one platter (133GB) full on that drive, but the 32GB library should be on the outside of platter 1 as I have not moved it, and it is never written to. Does a plattered disk write on the outside of all platters first or does it fill up one platter at a time and then move on to the next? Could defragmenting the drive move the library to the inside of a platter, thereby slowing it down by ~half? One of the reasons I got that drive in 2005 was to allow enough data on the drive for my needs and still remain ~70% empty.

[Crystal]

The way a platter drive is filed up is quite complicated I guess, and not linear, physically.

I’m not sure , but I won’t be surprise if several platters are in use for the same file.

However I’m sure about one thing : the notion of ‘beginning’ and ‘end’ of the drive is still pertinent and linear in terms of performances. If you copy/past some files on a virgin drive, they’ll be allocated on the fastest part of the drive first, to the slowest part of the drive, no mater where is it physically.
Personally I split my disk in two letters to put samples on the first one, and nothing else to do not fragment it, and use the second letter for archives. And even take care to put most exigent libraries first, typically, percussions, and specially pianos.