Can I Improve On Video Rendering Time?

[TBoneit]

http://www.cpubenchmark.net/ has benchmarks compring various CPUs. Highend chart @ http://www.cpubenchmark.net/high_end_cpus.html shows a Pentium D 3.2 benchmark @ 933 vs a Core 2 6300 1.86Ghz benchmarks @ 1417

There is a reason the Pentium D is cheaper. As well as being an older design. OTOH with a Pentium D you can turn off the heat in the room in the winter when you are encoding. And it won't overheat since it slows down as it gets hot to prevent CPU damage.

So nobody should upgrade to a Pentium D becuase a Pentium 4 is faster and never overheats?

NOt what I said, I compared Pentium D to Core2Duo. If I had a Pentium 4 and was upgrading to dual core I'd go AMD dual core or Intel Core2duo. In fact I jumped to the dual core AMD 64 X2. From what I've seen here at work the Pentium D are not suitable for anything that runs the CPU at 100% for more than a momentary burst due to heating.

Having said that, Good Luck and look into a extra large copper CPU cooler such as Zalman and some of the others. Artic Silver for mounting the Heatsink to the CPU.

Cheers

[ntscuser]

Well at present I'm getting transcode times at least four times better with Pentium D than I did with Pentium 4 and no overheating issues so far. No settings have been changed. A DVD9 which previously took 25 mins to analyse and a couple of hours to Shrink now takes 6 mins to analyse and half an hour to Shrink. I consider that to be a worthwhile improvement for a relatively modest investment.
EDIT: A 45 min TV programme which prevously took 8 hours to deinterlace and encode with TMPGEnc Plus 2.5 with maximum quality settings now takes 2 hrs 14 mins. It still takes almost as long to compile a DVD with DLP2 as it did before. Ambient CPU temperature is 17C, at full load is 33C

Haven't had an opportunity to re-encode a DVD9 with DVD-RB + CCE in 10-pass mode yet.

[Rudyard]

I have a pentium D and notice the fan goes like crazy! Encoding times vary, soemtimes my PD absolutely flies and will encode some video clips in say 5-10 minutes that sometimes take 40 mins (2 pass vbr encodes). I figured it was dependant on the source...its generally much quicker with AVI than MPEG reencodes but mpeg reencodes vary aswell.

Can I simply replace my processor with a Core2duo or AMD equivalent or do I need a new motherboard?

Also if I have a PD 3ghz, what would be a better Core2duo or AMD processor? Or are quadcore just around the corner and I should wait for that?

I do a lot of video rendering and 3ds max rendering.

[ntscuser]

ICan I simply replace my processor with a Core2duo or AMD equivalent or do I need a new motherboard?

That may depend on what kind of motherboard you have now.

Also if I have a PD 3ghz, what would be a better Core2duo or AMD processor? Or are quadcore just around the corner and I should wait for that?

I think it's extremely unlikely you could use a AMD processr in an Intel motherboard.

[TBoneit]

The fan will speed up and slow down with the processor temperature. If the CPU gets to hot it will throttle down speed to protect the CPU. The facts, fan going like crazy & Varying encode times, that you have described indicate not enough cooling.

Changing the processor to a core 2 duo depends on the motherboard and sometimes even the bios version. Details on the make and model and hardware revision of the motherboard are needed to determine that. I included the hardware version too as sometimes what h/w version determins what can be done. In teh past I have seen motherboards where the old h/w could only go to a certain speed but the newer version of the same model could use a faster processor.

No interchange of processors brand for motherboards since the old Socket 7 motherboards. Making it impossible to use a AMD or Intel in the motherboard was why Intel went to slot 1 and then socket 370 and socket 423 and the 478 and now 775. Since they protected those designs from AMD being able to make a plugin replacement no interchange possible. Not to mention that a Intel chipset won't support a AMD even if you could get it in the scoket.

[painkiller]

Somebodeez:

Now you've done it.

You're gonna make me upgrade to a C2D aren't you?


(sigh mo' money mo' money)

[nlec]

... The Pentium D dual was a stop gap until the Core2Duo was ready and was done by Intel in answer to the AMD dual cores.

Minor nit here... that's not the reason they developed C2D at all. It helped Intel get a leg up on AMD in the price/performance arena, true, but the real reason they abandoned NetBurst technology was they were hitting the wall with power consumption and heat as they tried to up the clock speed. AMD dual cores had zilch to do with the decision.

[jagabo]

When Intel introduced the P4 they bragged about how the netburst architecture would allow them to hit 10 to 15 GHz by 2010.

[nlec]

When Intel introduced the P4 they bragged about how the netburst architecture would allow them to hit 10 to 15 GHz by 2010.

Exactly, and when they had trouble pushing it to 4 GHz without generating tons of heat they had no choice but to abandon it and go in a different direction.

[TBoneit]

Which brings up back to the AMD dual cores that were using less power and running cooler than the Intel dual cores. I still think that influenced Intel to rethink. As I understand it one reason for the heating was they did something with single core processors and some logic glue to make the dual core thus leading to the heating. Now I've seen some things that say they are doing the same to rush a Quad core out. I think I'll wait until they get it right with the second gen quad cores.

I still reccomend Intel Core 2 Duo over AMD these days. Not so sure where to go with quad core right now.

[ntscuser]

I've had a couple of apps actually run slower with dual core than with P4. One of them is DVD Lab which now takes around 16 mins to compile a DVD that previously took around 13 mins. Another is WinRAR which is now around four times slower extracting a file than it was before even though archiving is around four times faster! Weird!

[jagabo]

I've had a couple of apps actually run slower with dual core than with P4... Weird! :shock:

Not weird at all. Different architectures excel at different things. Especially if the code isn't optimized for the architecture.

[ntscuser]

I've had a couple of apps actually run slower with dual core than with P4... Weird!

Not weird at all. Different architectures excel at different things. Especially if the code isn't optimized for the architecture.

I'm using ver 3.70 of WinRAR with multi-thread enabled and a file which previously took 4 mins to extract now takes 12 and a half mins. A quick Google shows other dual-core users have the same problem.

[TBoneit]

Just for the heck of it did you try setting it to single thread and see what happens. I have used winrar and a dual core AMD and didin't notice a difference.

It depends on what you do the most. If most of your time was spent extracting winrar archives then this would be bad. However that should be a very small percentage of your usage.

If it is speeding up the things that lways took a long time such as encoding video, then it is a worthwhile jump.

I know that going from a 3Ghz HT intel to the AMD dual core 4200 has improved encoing times. Windows boots much faster. Spy Sweeper used to take a long time loading and initializing. Now it and the AV load much faster and I can start using the computer in about a minute. with the AV and AS loaded.

it used to take 3 ior 4 minutes to load everything with the 3Ghz. Same amount of ram in both. 2Gb.

[ntscuser]

Just for the heck of it did you try setting it to single thread and see what happens.

Yes, no difference. Using an older version of WinRAR without multi-thread made no difference either, still 3 times slower than on the P4. The CPU meters barely move, one of them occasionally peaking above 10%. The dual-core XP hotfix from MS didn't help either

I have used winrar and a dual core AMD and didin't notice a difference.

It depends on what you do the most. If most of your time was spent extracting winrar archives then this would be bad. However that should be a very small percentage of your usage.

Agreed but the shorter the period spent extracting, the less chance of it going wrong and the safer I feel about deleting the original archive. I can usually rely on a neighbour to bump something heavy in the building about a minute before the end of the extraction process and have to start over again

Even with just one of the CPU's running at 3.2GHz, there should be an appreciable improvement over the 2GHz CPU I was using before. To add insult to injury, the new PC uses DDR2 memory (albeit single channel) whereas the old PC was only DDR1

[jagabo]

[quote="ntscuser"]

The CPU meters barely move, one of them occasionally peaking above 10%.

That's describing a situation that is I/O bound, not CPU bound.

[ntscuser]

The CPU meters barely move, one of them occasionally peaking above 10%.

That's describing a situation that is I/O bound, not CPU bound.

If that's the case Jagabo, what can I do to improve it? I'm extracting from HD master to HD slave, exactly the same as I was on the P4 machine. Same drives, same (primary) IDE channel but newer IDE cable as supplied with the mobo. Separate IDE channels are not possible with the new case (Foxconn) as the optical drives are mounted too far away from the hard drives.

This may or may not be related but decoding newsgroup files is also much slower than it was on the old PC, causing my newsreader to intermittently pause the downloading.

[jagabo]

Well, it sounds like you shouldn't be bandwidth limited. I wonder if WinRar has some kind of scheduling bug?

As an experiment, try this: open the archive and set the processor affinity (via Task Manager) so that WinRar only uses one core. Then extact the file. See if that makes any difference in the extraction time.

Does WinRar let you manually specify the number of threads? If so, see if there's any difference between one and two threads.

Badly fragmented drives could cause a slowdown with this type of operation.

[ntscuser]

Well, it sounds like you shouldn't be bandwidth limited. I wonder if WinRar has some kind of scheduling bug?

As an experiment, try this: open the archive and set the processor affinity (via Task Manager) so that WinRar only uses one core. Then extact the file. See if that makes any difference in the extraction time.

Does WinRar let you manually specify the number of threads? If so, see if there's any difference between one and two threads.

Badly fragmented drives could cause a slowdown with this type of operation.

You were right the first time Jagabo, there was indeed a bottleneck in the data transfer system Merely writing across drives whether internal or external was taking an age. After fiddling with BIOS settings (and not really knowing what I was doing) Windows reinstalled all IDE devices and drivers. After a further reboot, write speeds are now back to where they were before, something like a gigabyte a minute. File extraction across drives is now also back to near writing speed.

I wouldn't have considered this to be the basic problem Jagabo if you hadn't mentioned it, so many thanks for the help I also hope this is of help to anyone else with slow transfer or extraction rates.

[jagabo]

Maybe your drives (or one of them) had fallen out of DMA mode, into PIO mode. That would slow them down a lot.

[ntscuser]

We've at least discovered that programs such as WinRAR and DLP are limited more by data transfer rates rather than CPU speed. I suspect they work even better with a SATA drive than with PATA but haven't tried it yet. Some of the benchmark speeds I've seen for WinRAR are complete nonsense as they don't relate to real world operation. There is a slight speed advantage with WinRAR in dual processor or hyperthread mode in theory but it makes very little difference in practice.

[TBoneit]

Your NG burps are pr0obably back to speed?

Keeping in mind that your Antivirus can impact speed of data transfer as it has to monitor everything going to/from the drives.

I've seen Norton cause signifigant delays when downloading drivers and such for repairs here at work. After IE has done the download then it sits at 99% during the write to disk phase as it writes from the temp to permanant file. I am of course guessing that it is norton Av doing it however since installing a new copy of windows on a machine at home trying AVG, I noticed when I was downloading all the latest TEMPGEnc products for example that wasn't happening. So I'm guessing either norton is slow doing that type of scan and AVG is either fast, ignoring it, or poor at that sort of scan. Since I only go to the makers sites for software I'm not worrying about it.

One other possibility I also have Spy Sweeper installed at home and work and it may not have played nice with Norton.

[ntscuser]

Your NG burps are pr0obably back to speed?

As far as I can tell, yes Compile time with DLP2 has also dropped from 16 mins to 6 mins for a particular DVD. CPU meter now peaks at 63% when muxing is in progress compared with 24% before.

[nlec]

Does anyone know of any encoder performance benchmarks between P4, C2D and AMD? I have a P4 with hyperthreading, and while hyperthreading is not full dual processors, it does have its advantages.

I'm really tempted to go for a C2D system myself, but what I really want to see is a comparison between MPEG2 encoders running on a hyperthreaded P4 vs. C2D.

[jagabo]

Hyperthreading hardly helps video encoding at all. A dual core system will be significatly faster with most MPEG2 encoders.

Lots of benchmarks on different CPUs here:

http://www23.tomshardware.com/cpu.html

I have a hyperthreaded 2.8 GHz P4 and a Core 2 Duo E6300. I encoded a five minute DV video to MPEG2 with TMPGEnc Plus, same settings on both computers. The P4 took 8 minutes and 8 seconds. The C2D took 4 minutes and 33 seconds.

[manono]

Comparing my current Intel E6320 (at 3.25) with my previous AMD X2 4400+ Toledo (at 2.47) with my one-before-that Intel P4 (non-hyperthreaded) 2.66GHz (at 3.0), the E6320 is half again as fast as the 4400+ (50% increase) and 3 times as fast as the old P4. This is in CCE encoding. 7.8RT vs 5.2RT vs 2.5RT. This is the fastest it'll go. Reencoding a 2.35:1 16:9 NTSC DVD using the most basic AviSynth script.

There's no question at all that for encoding purposes, Core 2 Duos are the way to go at the moment, and maybe for some time to come.

[jagabo]

I encoded the same 5 minute DV file with CCE earlier but I was having some colorspace problem with the P4 system so I didn't report the numbers. CCE on the C2D took 75 seconds (~4x realtime).

[RabidDog]

Yes I am leaning towards an C2d .. would have gone for x2 but the c2d seem too good ..PLUS.. they are generally overclockable .. looking at a 6320 rather than 4300 (I might even do more than 2 passes!). Another point in all this is of course that some softs are optimized for intel or even C2d whilst other softs know nothing of dual treble or octuple cores!

[nlec]

Hyperthreading hardly helps video encoding at all. A dual core system will be significatly faster with most MPEG2 encoders.

Lots of benchmarks on different CPUs here:

http://www23.tomshardware.com/cpu.html

Thanks for that link. So Tom's Hardware is still good for something after all. :P

Yep, processing time is basically cut in half C2D vs. P4 540. Probably uses half the juice, too. I have lots of MPEG2 encoding to do so the difference is substantial. It's 4 times faster on a Core 2 Quad. No contest.

[drtalk]

.......uh guys? This has turned into a debate of Intel vs AMD. I STILL need a quick and EASY solution for my original posting here.
There's a lot of computer jargon that I've found on the net that apparently answers my question, but I have no clue what the **** they're talking about!

I joined here because all of you are so much more knowledgeable individually & collectively. If ever I needed an answer quick, this was my first and usually only stop. So again I ask: "WHAT DO I NEED TO RENDER EDITED VIDEO (WITH OR WITHOUT FILTERS & EFFECTS) IN A QUICK FASHION? I would like a simple explanation please if you can provide one for me. Consider all factors such as RAM, Processor speed, video card, etc, etc.