Video editing machine

[extremetaz]

Howdy all, great idea this new section.

Here's the spec:
Core 2 Quad Q9450
8GB (4x2GB) OCZ DDR2-800
Gigabyte GA-EP45-DS5
Spinpoint 160GB 16MB master
6x Spinpoint 500GB 16MB for 2xRAID5 arrays
2x Gigabyte 8800GT 512MB (NX88T512HP)
3x Viewsonic 22"
Thermaltake 750 Toughpower
Thermaltake VH6000SWA

Now, here's the questions.

1)The mobo has hardware RAID on board and 6SATA ports for it - I want to set up 2x 3HDD RAID5 arrays (capture data and render data storage). Can I run two arrays off the one controller?

2)The processor is a quad... for no particularly good reason really as I think I'm correct in saying that there aren't any video editing packages (final cut or premier grade, not pinnacle type stuff) or DVD authoring packages out there that can actually make use of multiple cores when transcoding. So am I likely to find that my new build at 2.66GHz is actually going to render slightly slower than my current 3GHz P4 unit?

3)I selected 8GB RAM in 4x2GB DIMMS as I assume that this is the most logical setup for use with 4 cores? yes/no

4)I'll need to be running Vista on the new machine in order to make best use of the RAM, the Cores and the GC's? yes/no

5)Finally, how decent is the stock cooling on the processor above? I'm not massively worried about noise but the machine will be running high loads for extended periods so it needs to work well. Worth investing in some aftermarket bling?

Cheers all,
Seán

 

[Shadow703793]

Ok, looks like you aren't gaming, so why not get a 8800GS or a 9600GSO to run the 3 LCDs and save some money? If gaming consider getting 2*4850 in CrossFire.

Imo, consider getting the 640GB WD AAKS drivers. I see you are running RAID5 so this may not be an option for you due to cost.

Consider getting a PC Power & Cooling 750 or a Corsair 750TX if possible. Imo, these PSUs are a bit higher quality quality than the Thermaltake PSU.

=========
As for your questions:

1. I am not sure if you can do that with the on-board RAID controller, but if you get a good quality PCI/PCIe RAID controller this is possible.

2. No. Clock for Clock the newer C2d/C2Q CPUs are much faster than the old P4s.

3. If you don't have a 64 bit OS forget going more than 4GB. Having a quad/dual core CPU has nothing much to do with how much RAM you need/have. If you are running a 64bit OS and have the budget and the need for 8GB go for it.

4. Depends on which Vista. If you want to use all 8GB you will need Vista x64. I personally use XP x64 on builds for video/image editing rigs as it doesn't hog too much RAM.

5. Im a fan of running cool so I personally put after market HSFs on all my builds even if not OCing. For HSFs, I recommend the XIGMATEK S1283 (+ bolt in kit) for OCing or not.

Side note: Assuming you have a x64 OS, and you want 8GB, remember to optimize your page file.

 

[thomaslompton]

Final cut and Premier grade authoring programs will use all four cores in some aspects of the program i.e final rendering etc.

And to 2nd the statement before, you definately need a 64bit OS.

You may not need two separate raid arrays. Render storage typically is not as intensive as capturing the data because you usually render in some sort compressed format and therefore needs less space and less time to write out the data. However, a good storage controller will allow you to create two concurrent raid arrays, but you may loose out on overall performance.

 

[einstein4pres]

1) Recent Intel chipsets with raid allow "Matrix RAID", which is similar to what you're talking about. Try looking into that.
2) No. There was an architecture change between the P4 and the Core which means that there will be greater performance per clock per core. Also, I expect that
3) It's certainly the cheapest way to get 8GB. There's no difference between 2x4 or 4x2 though.
4) No, you need to run a 64bit OS. There are 64 and 32 bit versions of Vista, and non-Vista 64 bit OSes (Linux, Mac OS, others).
5) If you are not overclocking the stock cooler is sufficient.

 

[extremetaz]

Morning all - and firstly cheers for the thought food... very helpful.

That Matrix RAID system seems to primarily be concerned with allowing the performance of RAID 0 with the redundancy of RAID 1, on only a 2 drive set. When it mentions two volumes on the array it means two logical volumes on one physical array off of one controller.

What I want it two physical RAID5 arrays off the one intel controller. At the moment it looks like I can from what I can gather - I was just hoping for confirmation. I'll probably just go ahead with it anyway and I'll report back later.

As for using the 8800GT rather than the 9600GSO, the former is touted as being better suited to HD video playback - which is probably where I'll be heading ultimately... although not for another wee while.

Very happy to hear that about the clock cycle efficiency - would have been a right kick in the nuts to find that I couldn't make the whole process any faster.

Now - RAM and such things so.
Here's the thing, Premier will run on 64-bit no hassle AFAIK, however, Encore (which is the REAL power hog) is only certified for 32-bit - and given that it's the transcoding process within Encore that I'm really looking to boost, I'm making it the priority.

So, 32-bit OS means 4GB RAM max yes?
...But only about 3GB actually practically usable

...and that means messy ram sets in order to spread that equally over both channels

...so I'm probably best just to go for a 2x1GB matched pair of high quality ram with decent timings then yes?

Thing is then that 2GB really doesn't sound like a whole lot to be using in a new machine, particularly as I've been using 2GB for the last 5 years in the old on (works fine, was just another place to improve) - although I assume that the jump to DDR2 will make all the difference?


The hard drives:
These WD AAKS units, is there a particular reason to go for them other than the additional volume?

To be honest, the only reason I stayed with 500GB unit was because they've been around for a while now so the manufacturing process for them is more robust and they'll be more reliable (just because I'm employing RAID doesn't mean I want to have to use it... ) but if the word on the block is that the bigger drives are just a reliable, or more so, then I've no bother stretching to them.


And finally the cooling:
I've one vote for stock and one vote for aftermarket - so here's some more detail for ye.

At the moment the machine on a render will run at 100% load on the 3GHz P4* for up to 4 hours at a time - I would expect that the new machine will run at least 50% faster so I'd expect the longest run to be about 3hours at 100% load. Cooler verdict?

*given the architecture of that core, the heat as you can imagine is intense (aftermarket cooler & aircon piped to the case to hold at 50C at full load).

 

[extremetaz]

You may not need two separate raid arrays. Render storage typically is not as intensive as capturing the data because you usually render in some sort compressed format and therefore needs less space and less time to write out the data. However, a good storage controller will allow you to create two concurrent raid arrays, but you may loose out on overall performance.

This is true - and it maybe my old school thinking getting in the way of modern progress so here's why I was going with 2 logical arrays - let me know if I'm just wasting my time.

In the old days... ...HDD's and the busses they were on had to be correctly configured so as not to bottleneck the system, so the way to set up a machine for fastest transfer was to have a seperate path for each drive on a dedicated channel. This meaning that data tranfer was all in one direction and there was no switching on the bus. So my older machines all had dual HDD RAID1 arrays (I'm a redundancy fanatic), each built on seperate physical PCI controllers to allow for max data transfer rates between the captured data and the RAM for transcoding, and then subsequently back out of the RAM and into the rendered data array for storage.

I don't want to build a system where I am transferring captured data to the ram from the same source as I'm sending the subsequently rendered data back to hence avoiding any switching on the bus and keeping the data flow nice, fast, and smooth.

Is it the case that these days the performance difference will be less noticeable?