Re: Dual processor system



In message <w5eDe.11179$Ag3.357@xxxxxxxxxxxxxxxxxxxx>, Dorothy Bradbury
<dorothy.bradbury@xxxxxxxxxxxx> writes
>> Why? For HDTV editing the hard disks are not going to be a bottleneck.
>
>You have to get data off the HDs, and data sets can be very large.

Most video editing is done in a streaming manner; the user sets up
start and end points and a script and the data is streamed off the disk
in frame-sized chunks, processed and streamed back to either the final
file or to a temporary file for further passes -- xVid, a common codec I
use a lot has an optimising 2-pass setting. A conventional 7200rpm SATA
hard disk can deliver data easily at a sustained 20Mb/sec assuming the
dataset is not too fragmented but it takes a lot of CPU to process and
render video at that sort of a rate. Really for this sort of work the
CPU is going to be the bottleneck. Lots of RAM will help -- enough to
get a complete project onto the motherboard would be really nice...

As an example, consider producing a DVD on a PC with one of the common
low-end tools like Sonic's MyDVD or the Nero suite (I've just done one
today, coincidentally). It takes a couple of hours on a machine like
mine to process 2Gb of non-MPEG2 video data and recode it to be written
onto a DVD. CPU utilisation was at 100% (Barton 2500+ overclocked to
2.3GHz, 1Gb RAM) for the entire time but the hard disk light only
flickered occasionally. If it was transferring 2Mb/sec I'd be surprised.

>Access speed yes re track-to-track re multi-user data transfers where
>the bulk of the time is spent on electromechanical & rotational latency,
>ie, where the higher IOPS of SCSI 15.3k-rpm outweighs raw SDTR.

If latency and speed were a problem and money wasn't I'd personally go
for a pair of spindle-synced 15k SCSI 3 640-LVD drives on a PCI-X
controller. Database appications and such where random access of disk
data are common really benefit from spindle-sync but more and more stuff
like indexes are being kept on solid-state discs with latency measured
in microseconds, or just kept cached in 128Gb of RAM on the server
motherboard (amazing to think that a terabyte of RAM can be bought for
less than 100,000 quid nowadays...)
>
>However a fast I/O system for video work is important - your RAM
>does not go a long way re data-set, cache, app & O/S in Windows.

That's why I suggested a server-grade motherboard which will have RAM
capacity above the titchy 3Gb found on most low-cost commodity mobos
today. I know that when the first Apple servers came out they were
snapped up by some people working on video as workstations as they could
be maxed out to 8Gb very easily.

>If you Stripe for performance then it is a very good idea to also
>Mirror that Striped set - so if 1 drive fails you don't lose all data.
>As opposed to just using cheap onboard Striping re data risk.

I'm not copacetic about striping and RAID generally -- I've heard too
many horror stories from sysadmins about "failsafe" RAID systems that
did not fail soft. When they did go wrong (and there is much more to go
wrong in terms of hardware) they tended to go catastrophically wrong. If
the RAID controller goes wonky it can write 0xDEADBEEF over all the
disks in its domain, or corrupt the second drive in a mirrored set when
trying to rebuild a new drive's data after a fail. Great idea in theory;
I want to live in Theory, everything works in Theory.
>
>Also backing up perhaps 100-200GB is somewhat slow - it is either
>to a USB 3.5" HD or a spendy DLT drive & multiple pricey tapes.

Backup costs time and money. More money, less time. More time, less
money. Tape robots are nice. USB2 external hard drives or SOHO NAS boxes
are cheap. Sensible backups are probably the best compromise -- I
occasionally watch an author friend backing up his day's work from his
Apple laptop to his PDA, his mobile phone, his wife's Apple in the next
room, the external 3.5" FireWire drive, his colocated server and finally
his iPod. If he could he'd output it on punched tape. I've offered to
track down an ASR33 for him and he's tempted...

>
>If the video editing will be mainly CPU bottleneck, then a single
>Raptor for the working disk would be fine (SDTR is high enough).
>
>Apple based solutions are expensive re "someone elses budget" :-)

There are some good video software tools for Apple but the high-end
market is either Sun or customised rack-mounted PC blades doing a lot of
parallelisation to improve workflow, offloading the rendering for a
cluster of workstations.
--
Note: this address does not accept HTML posts -- text only please.
Robert Sneddon
.



Relevant Pages

  • Re: Hard disk capacity
    ... >Edition, 256MB Shared DDR SDRAM, 40GB HARD DISK ... Is this enough disk space for typical applications? ... If the computer does not have RAM, what types of files will you be unable ... The disk space requirements for application programs are included ...
    (microsoft.public.windowsxp.hardware)
  • Re: Bit
    ... and that 4GB 32-bit address space must be used to address video ... which allows for 128GB of RAM. ... programmer and hardware. ... it would just be filling up the disk now. ...
    (microsoft.public.windowsxp.general)
  • Re: Inter-process communication - does data hit disk?
    ... > You should probably investigate the socket API further. ... > It certainly will hit disk if you write enough data to require the ... >> the complicated business of sending commands to the hard disk, ... > is going to decide to write some of your data out in order to free RAM ...
    (comp.os.linux.development.apps)
  • Re: PC very sluggish after installing new hard drive
    ... How much RAM memory?Try Ctrl+Alt+Delete to select Task ... How large is your hard disk and how much free disk space? ... Is the hard disk formatted as fat32 or NTFS? ... > drives. ...
    (microsoft.public.windowsxp.perform_maintain)
  • Re: Windows could not start because the following file is missing or corrupt: System32DriversNtfs.sy
    ... I used the memtest program to test RAM. ... >> on the hard disk which is now fixed, ...
    (microsoft.public.windowsxp.general)