Hard drive speed and rendering speed?

[Glenn Chan]

I'm wondering if hard drive speed has a large effect on Vegas' rendering speed. I ran some tests of my own using the following:
A Western Digital 7200RPM 80GB 8MB cache (WD800JB)
This is a modern drive and I believe a top performer for a 80GB drive.

That drive in PIO (priority input mode). PIO cripples drive performance (more on this later).

Quantum Fireballp AS2 - I would know more about this drive if it was my money. This drive is a "few" years old and Quantum doesn't even make hard drives anymore.

Doing a test render (45 seconds of footage, color corrector applied for some of the footage- 1 stream of DV) on my celeron733 (slow...) took on average 1:35 for the Western Digital and 1:37 for the Quantum- a 2 second difference.

In PIO mode instead of DMA, the Western Digital took 3:34 instead of 1:34- a huge difference.

In these tests I rendered a new track to the same drive. I ran them four times and the deviation was about +- 2 seconds.

I ran winBench on all of the drives. It was clearly WD > Quantum >> WD in PIO. WinBench was a bit erroneous though at they all had the same sustained transfer rate. This wasn't a very good test because I don't know if my system is bottlenecking somewhere- maybe on the hard drive controllers. My motherboard is really crappy so it could be happening. Does anyone else have any experiences with hard drive speed affecting rendering speed?

My results show that hard drive speed has a neglible affect on rendering speed unless intentionally crippling your system.

[Edward Troxel]

Yes, you MUST use DMA mode for your hard drives. PIO is simply too slow. It makes a big difference on capturing and printing as well.

As I mentioned in another thread, I have tested with a slow desktop with a 7200 RPM vs a fast laptop with a 5400 RPM and speed CAN make a difference depending on other factory.


If I edit an equal project on each and to a PTT (rendering ONLY the changed sections) the fast laptop is about 4 times faster than the slow desktop.

If I render that same project to a full lenght AVI (i.e. all NON-rendered sections are also copied), the fast laptop takes LONGER than the slow desktop. So I do believe that drive speed makes a difference.

[Glenn Chan]

It doesn't seem like drive speed explains the 4X speed difference. Anyone with a RAID setup should be able to do a good test (RAID versus no RAID). I'd be interested to see what happens.

[Brad Higerd]

A friend helped me build an excellent video editor at an excellent price, and the specs are as follow:

ABIT IC7 MOTHERBOARD
P4 2.8 GHZ HYPERTHREADED PROCESSOR
1 GB COSAIR MATCHED CERTIFIED RAM (Dual Channel at 400MHz each)
4 HARDDRIVES:
2 SATA (80 GB & 200 GB MAXTORS)
2 PATA (40GB & 120 GB WESTERN DIGITALS)
and some other incidental stuff…

To answer your question, most of the render bottleneck in the faster machines still occurs at the processor level. I have over-clocked this system to 3.3 GHz (approximately 1 GHz FSB) without any stability issues, and we still find that the RAM and hard drives are not even close to maxing out (this assumes you are doing a color correction or something which requires a significant alteration of most pixels in every frame). The video drives (all except the 40Gb) can handle sustained read/write speeds in excess of 40Mb/s, and we always test with transfers from and to different drives. Meanwhile, my processor is chugging along between 55%-95% usage with the higher values exerted on the simplest of render tasks. If you do a lot of color correction, and you are using today’s technology (PATA or SATA drives at 7,200 rpm or greater) with the DMA enabled, your hard drives should not be your limiting factor.

In this particular computer, the video card is the limiting factor.

[Glenn Chan]

Out of curiousity, how is the video card the limiting factor of your computer?

[Bogdan Vaglarov]

Yeah, looking to all of the recent benchmarks on the net it's clear that the processors clock speed is the main factor for faster encoding/rendering.

Brad, how do you solve the heating problem overclocking to such an extend? Stock cooler or better cooler, case air flow? I don't want to opt for water cooling but yould like to overclock P4 2.6 or 2.4 to about 3 GHz.

I'm going to make very similar configuration (just opting for the IC7 Max3) and wonder also about the memory.

Does it have big influence on the rendering? Lets say if you run DDR400 or if you overclock the memory too at 250 MHz? Does it matter if you run it synchronious or you step back and use 5:4 or 4:3 mode?

[Glenn Chan]

With really good air cooling you can get up to 3.5mhz with a 2.4c and 2.6mhz with a 2.6c (see the pentium review at ocaddiction.com under reviews).

For Pentium the best heatsink is probably the Thermalright Sk-900 (something like that- the one that takes 92mm fans). There are a huge variety of fans which tradeoff performance versus noise.

The most comprehensive thermal paste review is over at http://www.thecrucible.ca/reviews/th...conclusion.php. ocaddiction.com has an article on how to apply it.

Having a good case also helps. There isn't any hard answers on which is best. You can try the case and cooling fetish forum at arstechnica.com.

Also get a good power supply. tomshardware.com might have a roundup for these.

Great memory will let you run at a 1:1 ratio to very high speeds (synchronous is fastest of course). Currently the best memory is OCZ PC4000 (the most expensive type). It's not the best value though...

sites like hardocp.com, anandtech.com, ocaddiction.com will have more information on overclocking.

By the way, definitely check out http://www.go-l.com even if only to look at their computers. With Asetek's phase-change cooling solution (fancy name for refrigeration) they will overclock the P3.2c to 3.8mhz!

Water, phase-change, Peltier, and liquid nitrogen are all possible ways of getting a better overclock but they are a case of diminishing returns. Liquid nitrogen cooling (putting the stuff into a container sitting onto the CPU) isn't really feasible, but it's interesting and has been done.

I have no idea how much overclocking speeds up Vegas rendering speed. Maybe someone would like to post up their results.

2- With someone with a RAID please do a more definitive test to show if hard drive speed has a major effect on rendering?

[Brad Higerd]

“Out of curiosity, how is the video card the limiting factor of your computer?”

- not sure exactly how this works. I have been told by several folks that the graphics card is utilized in rendering some of the effects (maybe just 3D transitions and effects like Boris Red or AE produces). It is the limiting factor on my PC because I have not found a really good 8X video card that has a firewire out to feed my “Movie Box” which I then feed into an NTSC monitor for previewing the edit work. It gets choppy when I add effects and keep the resolution at best, and I don’t want to back off the resolution.

“Brad, how do you solve the heating problem overclocking to such an extent? Stock cooler or better cooler, case air flow? I don't want to opt for water cooling but yould like to overclock P4 2.6 or 2.4 to about 3 GHz.”

The system builder specified the following cooler: CPU FAN|AEROFLOW VP4-C7040 VTEC RT (newegg nomenclature). The case has a 400 watt power source and lots of fans and venting. The case remains at or below 32C, and the CPU has never gone above 52C (well within its operating range).

“Does it (overclocking) have big influence on the rendering?”

NO! If the render times are short, you will hardly be able to notice, and if they are long, you will likely leave the room and return at some later time to find that one or two minutes here or there is insignificant. However, if you are always riding your deadlines to the wall, and you like to watch the rendering frame numbers scroll like the fee at the gas pump, then overclocking may be your answer.

I generally don’t do it. The computer is very fast, and I want my system to remain as cool as possible.

“Lets say if you run DDR400 or if you overclock the memory too at 250 MHz? Does it matter if you run it synchronous or you step back and use 5:4 or 4:3 mode?”

The system is stable only when the RAM is synchronized to the CPU. The ABIT BIOS allows for some control over the RAM, however the RAM cannot exceed 400MHz (800 MHz with dual channel), and I don’t think the high speed Cosair RAM can be used below 400MHz either. This limitation is related to the RAM alone. However, the RAM we chose is capable of a latency set to the most aggressive settings possible on the ABIT BIOS, meanwhile the system remains extremely stable. The system rarely ever uses more than 400 Mb of the 1024 Mb available. The advise of my builder was ALWAYS buy good RAM! “Otherwise,” he warned, “your RAM will be the bottleneck.”

[Bogdan Vaglarov]

Glen, thank you for the very usefull links. Recently I found more than expected information on the web. I've been frequent visitor to toms hardware but there is lots more independent reviewrs - I was stunt from the volume of information you can get for just anything.

Brad, thanks for the detailed replay. BTW I was considering exactly the same heatsink (Vantec) for the CPU - rated as better cooler than the stock and quiter in the same time. The case I will buy though (for many reasons) is Antec Sonata which is not so well vented and doesn't give much posibilities of adding fans.

I meant if the memory speed matters for the rendering. I'm sure memory is crucial for gaming but how about video editing, encoding and rendering. Is it really neccessary highest speed and timings?

And you should be able to overclock the memory rising the voltage. There are also memories working on 433, 466 and 500 (the OCZ PC4000 which Glen noted is especially good - overclocked you can run it even over 260 MHz - 10 above the nominal 250 giving the best performance record of all)

Sorry for taking you a bit of the topic of this tread.

[Brad Higerd]

The way the IC7 ABIT BIOS manual explains the overclocking, you are really overclocking the FSB and not the RAM. If the 200 MHz dual channel DDRAM (acting as 800MHz normal clocking) can really be brought to 250MHz by a voltage increase, you are not gaining that much on a system that can feed data through the dual channels at a rate much higher than the processor can compute (that’s exactly what our testing has shown). I don’t doubt overclocking the FSB has an effect on the RAM, but it has been explained to me and demonstrated on this computer that the latency settings and the type of RAM has an even greater impact. Unfortunately, high speed RAM is not cheap.

While we are on somewhat of a tangent, we are still addressing the question of hard drive speed. Given today’s technology, processor speed still remains the most likely bottleneck to any NLE.

All of our tests were conducted using Vegas 3 and 4. The computer that is being discussed is used exclusively as a video editor. No games, internet, or general office use is employed, and all unnecessary background processes have been eliminated. It runs much faster after tweaking, and most of the tweaks can be found at:

http://www.slashcam.de/artikel/Tips/TWEAKS_for_Windows_XP_for_Video_Editing__v_1_0_.html

I read somewhere that this article was swiped from someone else’s work, but I cannot confirm.

[Glenn Chan]

RAM frequency is linked to the front side bus speed. If you raise FSB frequency from 200mhz to 250mhz then your RAM will run at 250mhz (AKA DDR500- the highest rating for RAM). Your system might get unstable if you push FSB frequency further. If RAM is your bottleneck then you can:
A- Use a memory divider (3:2, 5:4) so that the RAM runs at a lower frequency. The frequency it runs is a ratio to the FSB frequency. The fastest setting is 1:1 (normal).
B- Relax the memory timings. This will make the RAM slower, but the benefits of this may be overshadowed by an increase in speed from faster FSB/RAM frequency.
C- Increase the voltages to your RAM. This can really damage your RAM and will void your warranty on nearly all RAM (only certain OCZ brands will let you keep the warranty).

I have no idea what the effect of higher FSB frequency, RAM frequency, RAM timings, or memory divider has on Vegas' performance. You might be able to take a good guess from benchmarks on various computing sites including xbitlabs.com and ones that I already mentioned.

[Brad Higerd]

Glenn,

I am more than happy with the speed of this particular machine, and as stated, the RAM speed is not an issue. My builder did an excellent job specifying parts. By my very rough calculations, my processor would have to be capable of about 7 GHz (leaving the FSB at 800 MHz) before RAM would become an issue.

As far as changing timings, we have tried, and the system became very unstable. Video editing software (at least every trial version we tried) demanded that our system be stable before they would be productive.

Maybe you will have better results with something other than 1:1.

Brad

[Glenn Chan]

http://www.xbitlabs.com/articles/mem...p4-mem_13.html
Hmm Xbitlabs did a thorough test of various memory combinations and processors (overclocked too) and said:

As for using DDR500 SDRAM in the synchronous mode during FSB overclocking to 250MHz, we really doubt that it makes sense. This overclocker memory is much more expensive than ordinary DDR400 SDRAM, while the performance it provides is just a little higher than that of DDR400 SDRAM with 2-2-2-5 timings.

The conclusion is that it makes more sense to use cheaper enhanced latency RAM with a memory divider than overclocking RAM (i.e. PC4000).

[Bogdan Vaglarov]

Yeah, I started asking myself (and you) the question about the memory because this new DDR500 (there is even PC4200 now) is darn expensive.

In fact I suspected that this high band new memory is virtually almost the same as DDR400 with tight timings. With DDR500 you can rarely see something better than 3-4-4-8 T1 and I thought if you push a good DDR400 you just have to lose the timing to achieve similar effect.

If you have the time here is another informative review on memories: www.tweaktown.com/document.php?dType=review&dId=547

The main thing though is if really fastest memory and agresive timings help video editing/rendering/encoding? For gaming you'll achieve the extra frames but is that the case with video work?

[Bogdan Vaglarov]

Great article Glenn - after reading it many questions have their answer. My opinion in the previos post is not exactly true it seems.

Anyway its tuff to decide on which type and class of memory to buy. I would recomend anybody wondering about the impact of the memory on the system to read the whole article on xbitlabs (see the link from Glenn and jump to page 1).