View Full Version : Hard drive configuration?


Bret Corbin
February 12th, 2004, 02:50 PM
I have recently purchased a 2nd hard drive for my PC which I will use for capturing/editing video using Vegas 4 +DVD. I have not yet purchased Vegas but will do so soon. The drive capacity is 250GB. Does anyone and/or everyone have any suggestions on partitioning this drive? Is there a reason to split it into two drives or 3 or whatever. I am brand new to NL video editing and have been slowly adding hardware to my home PC to make it an editing suite. The information I have so far I have read on these forums and in D.S.E.'s book on Vegas

Also, I currently have 512MB of Ram. Is it worth it, and if so how, to add another 512MB.

I think that's it for now.

Thanks in advance!

Bret Corbin

David Mintzer
February 12th, 2004, 02:52 PM
No reason to partition that capture drive---particularly if thats all you are going to use. Vegas loves memory so the more the merrier----it will help with dynamic ram previews although 512 certainly is enough to run Vegas without a hitch.

Rob Lohman
February 12th, 2004, 02:54 PM
I'm partitioning my drives to keep things organized, but it isn't
demanded by anything as long as you use NTFS filesystem.

Edward Troxel
February 12th, 2004, 03:04 PM
I see absolutely no reason to partition the drive. In fact, partitioning CAN slow you down - especially when accessing files from two different partitions. Just leave it as one big drive and use folders for organization.

Bret Corbin
February 18th, 2004, 08:37 AM
Thanks for the input. I will install the drive as a single drive this weekend and buy Vegas within the next couple of weeks.

Glen Elliott
February 18th, 2004, 01:16 PM
Optimaly you'd want 3 drives- one with the program (Vegas), one for your raw footage, and one for your output/render drive.

Ram limits the length of ram renders you can achieve. Also high performance ull (ultra low latency) definitly helps speed up ram rendering. I'm running a gig of dual channel Corsair PC3200 ull Ram...however, like others haven mentioned, it's by no means necessary. Thats one of the great things about Vegas- it's scalability.

Glenn Chan
February 18th, 2004, 01:35 PM
Glen, I've tested lowering memory timings on my RAM and it makes no measurable difference! Even assuming there is a 1% performance boost, it definitely isn't worth it to get low latency RAM.

Memory bandwidth does make a slight difference (2-3%), but you need to overclock to take advantage of it (or get a 800FSB instead of 533FSB processor) and PC4000/4200 RAM is quite expensive. It's much more bang for your buck to spend more on cooling, get a 2.4 or 2.6"C" to overclock (not the faster grades), and get normal RAM and run at a memory divider.

Optimaly you'd want 3 drives- one with the program (Vegas), one for your raw footage, and one for your output/render drive.

I tested this out, hard drive speed doesn't make much of a difference. The best way to improve performance is to not have a hard drive bottleneck. Use internals, check that they aren't in PIO mode, and maybe watch out for fragmentation. And have enough space- not enough will drop performance significantly.

2- Don't partition the drive unless it helps you to organize. Store only video on that drive and fragmentation shouldn't be a problem.

3- More RAM? You don't really need more RAM, but it can help out RAM rendering. You don't need to use RAM rendering though.

Glen Elliott
February 18th, 2004, 01:52 PM
PC 3200 matches the bus of the P4 3.0ghz...and surely dual channel ull ram will produce the ram render in shorter time than much slower ram.

I agree that hardrive speed doesn't affect video editing too much unless it's so slow that it affects the read and write performance that Vegas needs to display all the footage correctly on the timeline. Though...once rendered that doesn't matter.
The idea of 3 drives is the cpu's ability to access all drives simultaneously. Instead of having to read, write, seek, etc on one drive- there is a drive for each. While drive 1 is being used to run the app, drive 2 is being used by drive 1 for source material, and drive 3 will be open and ready for any writing of new files.
Hardrive speeds *do* however have a negligable effect on all programs- thus the fact a Raid 0 will load aps faster than a single drive.

Gints Klimanis
February 18th, 2004, 04:35 PM
>I see absolutely no reason to partition the drive. In fact, >partitioning CAN slow you down

I'm experimenting with making a faster drive using partitions.
The outer portions of the disk offer greater bandwidth. Partitions are located from outer to inner, so the outer partitions will have a higher data transfer rate. I'll post my results when I'm done.

Rob Lohman
February 18th, 2004, 04:58 PM
I think it's actually the otherway around. Harddisks drop off in
their speed the futher you get on the drive (and the futher
outwards you go), so I'd say it beging on the inside and then
moves outside?

edit: removed some nonsense...

Gints Klimanis
February 18th, 2004, 07:03 PM
Given that the platter has a constant density and a constant spin rate, the disk bits on the outer tracks travel faster. The angular velocity is the same at every point on a disk radius, but the outer points are traveling a greater distance in the same amount of time. However, the hard drive makers manipulate sector size in zones, as you'll notice in the stepped response from a disk speed
tool. Have you ever noticed that the hard drive sustained read and write tests aren't constant ? That's because different parts of the disk are faster than others. My plan is to use this tool to identify the disk zones and allocate partitions accordingly on some disks.

Anyway, partitions are allocated from outer tracks to inner tracks,
unlike CDs and DVDs which are inner to outer. I notice that my disks seem to vary in read bandwidth by a factor of two over the disks I've tested.

Gints Klimanis
February 18th, 2004, 07:05 PM
Check out a disk speed tool such as HD Tach. I have a registered version of 2.61, but here is 2.70 that will provide free read speed tests, both bandwidth and access time.

http://www.simplisoftware.com/Public/index.php?request=HdTach

You'll notice a stepped response on the graphs. The steps identify the disk zones.

Gints Klimanis
February 18th, 2004, 07:28 PM
HD Tach 2.61 draws a read speed graph with greater detail than 2.70, the version currently available on the web site. On 2.61, It's much easier to see the stepped response that identifies the disk zones.

Glenn Chan
February 18th, 2004, 07:39 PM
PC 3200 matches the bus of the P4 3.0ghz...and surely dual channel ull ram will produce the ram render in shorter time than much slower ram.
I tried this out. My system is a 2.6C 800FSB with *single channel* 1X512MB PMI RAM (no heatspreader). There is no measurable different between the RAM timings 3-4-4-8 and 2-3-2-5.

RAM bandwidth makes a small difference, but applies only when overclocking.

I agree that hardrive speed doesn't affect video editing too much unless it's so slow that it affects the read and write performance that Vegas needs to display all the footage correctly on the timeline. Though...once rendered that doesn't matter.
The idea of 3 drives is the cpu's ability to access all drives simultaneously. Instead of having to read, write, seek, etc on one drive- there is a drive for each. While drive 1 is being used to run the app, drive 2 is being used by drive 1 for source material, and drive 3 will be open and ready for any writing of new files.
Hardrive speeds *do* however have a negligable effect on all programs- thus the fact a Raid 0 will load aps faster than a single drive.

Tried this out too. 2% difference with having more drives in a good case scenario, 0% on long complicated renders. On renders where the CPU is not a bottleneck (my test was turning JPEGs into DV and applying color correction), I tried a RAM disk (use RAM as hard drive) and performance bumped up 9%. A RAM disk is extraordinarily fast and should indicate the best performance you could hope to achieve.

RAID performance should be less than 9% a RAM disk gives. In real world renders (reading DV footage, not JPEG) the difference would be even less. On a long render I measured no difference between a RAM disk and a normal hard drive.

Links:

http://www.dvinfo.net/conf/showthread.php?s=&threadid=18841 Memory timings and bandwidth

Hard drive speed tests:
http://www.dvinfo.net/conf/showthread.php?s=&threadid=15637
http://www.dvinfo.net/conf/showthread.php?s=&threadid=18784

The one tweak that DOES work is overclocking, and it works well. It's somewhat involved, but if you're going to tweak then what else would you expect?

My computer (which isn't a great setup for overclocking) can overclock 20% but I have to disable hyperthreading so that it doesn't reboot/freeze. Performance gains should be about 16% most of the time (depends on render).

Gints Klimanis
February 18th, 2004, 07:48 PM
Glenn, Thanks for the detailed response. It sounds like you've put in a lot of good research. In particular, I liked the RAM disk idea.

Anyway, I'm trying to speed up the processing of multigigabyte
uncompressed video files: DV -> uncompressed, uncompressed->uncompressed, DV->MPEG2 and uncompressed->MPEG2. The RAM disk is quite expensive at currently $170 or so per *quality* PC3200 DDR Gigabyte, but would be quite useful on smaller MPEG2 encodes. I went with a RAID array with a $100 controller and several $50 (after two rebates) 120 GB 7200 RPM Western Digital drives.

Gints Klimanis
February 18th, 2004, 07:56 PM
Warning on using HD Tach:
DO NOT USE THE WRITE TEST ON DRIVES WITH USEFUL DATA.

The read test is safe, but the write test will annhilate all data on the drive.

Edward Troxel
February 18th, 2004, 09:22 PM
Let's do a "for instance": Suppose you partition a drive into two partitions and have your video files on one partition and then render to the second partition. In this case, you will slow down the drive because the head will be constantly moving back and forth between the partitions.

Glen Elliott
February 18th, 2004, 09:57 PM
XP automatically stores larger files on the outer edge of the hd. Ever wonder why theres large chunks waaayyy out on the end while defragging? ....

Gints Klimanis
February 18th, 2004, 10:01 PM
Defragging has to do with your defragger. In Norton, I notice that the middle of the disk is left empty. The most frequently used files are at the end of the disk while the infrequently used files are at the beginning of the disk.

Gints Klimanis
February 18th, 2004, 10:06 PM
Edward,

True. That usage of partitioning will not be wise.

For two partitions, a better way would be to keep your infrequently used files in the slower partition while keeping your
uncompressed video or source DV files in the faster partition.

Another usage would be for a temp directory for WinZip or
the disk compilation phase of CD or DVD burning.

One could reduce boot and app launch time by keeping Windows
and Programs Files on the faster partition.

Edward Troxel
February 18th, 2004, 10:24 PM
I'll agree that partitioning CAN have its uses but I would generally say that it is not necessary.

Bogdan Vaglarov
February 19th, 2004, 01:58 AM
I'm also wondering how to make my partitions on my new system.

BTW I've used Norton defrag on my Win Me notebook and to me it seems that the swap file and the frequently accessed files are always in the beginning of the disc. I don't say that means they are on the inner physical side. I doubt Norton will chose to put them on the slower 'end'. The non-frequently used files are gathered on the end.

Now on the disk space and partition usage.

Currently I can't afford putting 3 drives in my box (sweet small factor cubic). So my plan is to make the following set up.

HDD 1 (system & data) 160GB
May be 3 partitions:
1a. system 10GB;
1b. Rendering to MPEG2 (from 2a)
1c. the rest - data, docs, etc.;

HDD 2 (video, backup) >=160GB
May be 3 partitions:
2a) video capture, and related media
2b) DVD authoring (temp for burning the ready project) - 5 GB (is it enough for single DVD?);
2c) 5GB for system disk drive image (backup)

Now writing this even I've got confused. Where the actual editing should be done?
Generally my idea is to render to different disc, then to author again to different disc. Writing to the DVD shouldn't matter.
Any opinions?

Rob Lohman
February 19th, 2004, 06:12 AM
Sounds pretty okay to me. If your drives are empty or have
free space you can make partitions from within windows by
opening Computer Management under Administrative Tools.

If your harddisks are full you basically have the following options:

1) destroy the data on it (usually not good) and simply repartition / reformat the drive

2) move the data to some other drive / system / tape / media before doing the repartition / reformat

3) get yourself Partition Magic Pro. This program will allow you to move, resize, split / merge partitions without loosing the data on it. Ofcourse backupping up valuable stuff is needed because something can always go wrong with such drastic things!!

Gints Klimanis
February 19th, 2004, 03:02 PM
I benchmarked a number of disks, ranging from the 40GB Seagates at work to my 120 GB and 180 GBs at home. I
haven't benchmarked my 250 GBs though. My conclusion is
that two partitions would offer a disk that can do about the data bandwidth in the first partition as the last on the 40 GBs. That's
double the bandwidth and well worth it.

There is less of a performance variation (only about 30%) in sequential read speed on the 120 and 180 GBs. I'm wondering if this is due to their connection to a my Rocket RAID controller card (not in RAID confuguration, though) rather than the motherboard IDE connectors.

Gints Klimanis
February 19th, 2004, 03:06 PM
Bogden,

If you're making that many partitions, put your high bandwidth
stuff (system 10G and DVD temp) on the lowest partitions.
Recently, I've learned that the outer tracks on a disk offer the greatest bandwidth and that partitions are allocated starting with
the outer disk zones. I'll try to find a web article. My storage specialist friend tells me that this is common practice among disk specialists. It's too bad that this great info doesn't see more sun.

Bogdan Vaglarov
February 19th, 2004, 04:19 PM
Thanks for the replays.

Gint, although I understand your point it's still confusing what you are saying.

So it's clear the outer physical zonez on the disc allow for higher speeds.

The question is where is the logical beggining of the disk. Till now I always asumed that the beggining (represented by bar or table) showed by the software is where the beginning of the physical drive (hence the outer zones).

For CD/DVD I asume same - the graphical representation by the software shows the begining at left but everybody knows the physical beginning of this discs is on the inside tracks.

I don't think you should really care which is the physical beggining of any disc - the OS or utility will show it always on the left side of the bar/graph.

I don'tknow how the nations writing from right to left will implement this my strategy though (just kidding).

Gints Klimanis
February 19th, 2004, 04:54 PM
Hard disk partitions are allocated from outside to inside. CDs and DVDs are just the opposite, inside to outside. Using the WindowsXP tool or Partition Magic, the first partition will appear on the left side.

Use the HD Tach tool to help you identify speed regions.
http://www.simplisoftware.com/Public/index.php?request=HdTach

It's quite clear from the graph that the outer tracks or "beginning"of the drive offer about twice the bandwidth of the inner tracks from this graph:

http://www.simplisoftware.com/Public/index.php?request=HdTachUsage

As mentioned in another thread, it's not so easy to see the steps in the zone speeds on the HF Tach 2.70 graph. 2.61 offers a sharper graph that shows the speed zones better. email me if you really want that revision.

Gints Klimanis
February 19th, 2004, 05:02 PM
Although my 40 GB drives showed a 50% dropoff in bandwidth on HD Tach sequential read speed, my other drives 120 and 180 GB didn't show as large a dropoff. I'll test my 250 GB tonight.

Bogdan Vaglarov
February 19th, 2004, 09:19 PM
Gint, I know well HD Tach - it's used in from many testing and review sites.

English is not my native language so I was trying to make it transperent because it really sound a bit mixing this way.

In other word what you see on the left side of the graf (no matter HD Tach, Partition Magic or Win XP disk manegement) is the faster region which heppans to be physically the outer disc region. Did it become more confusing.

Anyway I think we are talking about same thing.

In this context I think partitioning and the speed of the partiqular partition will depend on one more fact.
Modern big drive reside 2 physical plater each with it's own head.

If your partition 1 is smaller than the first plater it will be served by 1 head. If partition 2 happens to ocupy space from both platers it might be faster as the data can be scatered on the to platers and served by 2 heads.

Thats why the recent trend is that bigger drives are faster and also people do not make partitions but leave the drive as a single partition. With NTFS you don't lose space as the clusters are always 4kB. In Fat 32 partition above 32 GB will use 32kB for clusters which is huge waist of space.

Anyway I whish you successful testing. My current aim doesnt include uncompressed video so I'm fine without top speed.
Cheers, Bogdan

Bogdan Vaglarov
February 22nd, 2004, 11:06 AM
Sorry everybody, apparently my therory was wrong.

After a bit more research I found that actually disk platers are using both sides for writing. Usually the first side of the first plater is reserved from the drive and not used for data.
So when the first partition is made half is written on one side of the first platter and second half is written on a side from the seccond platter. There are also many read/write heads flying over these surfaces. Eg. 10GB partition would be written as 5GB on the side B on platter 1 and 5GB on lets say side A on platter 2.

In addition the outer faster turning tracks are with lower density and the inner tracks are layered with higher density to compensate for the lower speed. So virtually the drives are designed to be equally fast on the whole surface area. The benchmarks show slightly different picture as we see.

Glenn Chan
February 22nd, 2004, 05:22 PM
Outer tracks can be packed with more information.

www.storagereview.com has more information on hard drives that anyone would want. They're the best site in my experience, but they are sometimes wrong. They say in the reference guide that a 8MB buffer doesn't make a big difference- in practice it makes a 30% difference. They also say that RAID doesn't boost performance much (10%), but their own tests with better RAID controllers show much better performance.