Please help me compare Dual Quad 2008 Mac Pro, w/ a i7 Quad Macbook Pro
If it now takes me one hour to render a Final Cut Studio "Compressor" project, (2-Pass VBR, mov to m2v), using my:
Early 2008 Mac Pro (MP), Dual Quad Core, 2.8ghz, 12 GB RAM, internal 1tb 7200rpm (non-raid) hard drive, 512mb Nividia GeForce 8800 GT...
How long would the same render take with the new 17" or 15" Macbook Pro (MBP) (Quad Core i7, 2.2ghz, 8gb RAM, internal 500gb 7200rpm drive, 1tb Video Card?)
And how big a difference in the render time would there be if the computer was doing another activity, say surfing the internet, or light Photoshop work, at the same time? Is there a substantial performance bump gained by paying the extra $250 and getting the 2.3ghz i7?
I know in the past comparing mac laptops and towers was like comparing gas and rocket fuel.
But it has been three years since my last purchase. (Early 2008 MBP Pro and MP.)
I need to regularly be able to effectively do some video editing.
Rendering time is key, as is using multicore processing with FCS's Compressor.
Early 2008 Mac Pro (MP), Dual Quad Core, 2.8ghz, 12 GB RAM, internal 1tb 7200rpm (non-raid) hard drive, 512mb Nividia GeForce 8800 GT...
How long would the same render take with the new 17" or 15" Macbook Pro (MBP) (Quad Core i7, 2.2ghz, 8gb RAM, internal 500gb 7200rpm drive, 1tb Video Card?)
And how big a difference in the render time would there be if the computer was doing another activity, say surfing the internet, or light Photoshop work, at the same time? Is there a substantial performance bump gained by paying the extra $250 and getting the 2.3ghz i7?
I know in the past comparing mac laptops and towers was like comparing gas and rocket fuel.
But it has been three years since my last purchase. (Early 2008 MBP Pro and MP.)
I need to regularly be able to effectively do some video editing.
Rendering time is key, as is using multicore processing with FCS's Compressor.
Comments
Generally, I would expect that your Mac Pro is still a tad faster. The disk speed, RAM, and cache are very important for this type of benchmarking. Also, if you are multitasking, you will be wiping out cache data during context switches, which will slow down your render a lot.
Thanks for explaining how multi tasking slows down the render time.
(For the sake of argument, lets eliminate the multi tasking variable from the equation.)
So, the disk speeds, 7200rpm are the same. (no advantage)
The RAM is greater on the MacPro 12gb v. 8gb, (though the MacPro has 800mhz DDR2, while the Macbook Pro has 10ggmhz DDR3)
I am guessing the advantage may still be the MacPro.
How much faster is the 1ghz DDR3 RAM v. the 800mhz DDR2 RAM?
I am not sure how to interpret the difference in cache.
The MacPro has 12MB L2 cache per processor (each pair of cores shares 6MB)
The i7 Macbook Pro has two options:
with 2.2ghz option: 6MB shared L3 cache
with 2.3ghz option: 8MB shared L3 cache
The question is how much faster is L3 cache v. L2 cache
Thanks for the reply. I will have to check and see if the Macbook Pro's at the Apple Store have Compressor installed, and whether I could run that test.
Thanks for explaining how multi tasking slows down the render time.
(For the sake of argument, lets eliminate the multi tasking variable from the equation.)
So, the disk speeds, 7200rpm are the same. (no advantage)
The RAM is greater on the MacPro 12gb v. 8gb, (though the MacPro has 800mhz DDR2, while the Macbook Pro has 10ggmhz DDR3)
I am guessing the advantage may still be the MacPro.
How much faster is the 1ghz DDR3 RAM v. the 800mhz DDR2 RAM?
I am not sure how to interpret the difference in cache.
The MacPro has 12MB L2 cache per processor (each pair of cores shares 6MB)
The i7 Macbook Pro has two options:
with 2.2ghz option: 6MB shared L3 cache
with 2.3ghz option: 8MB shared L3 cache
The question is how much faster is L3 cache v. L2 cache
Speaking as an ex Apple employee, I would let you install compressor, but i believe it was already installed. All product has a default image so when you reboot it would just revert to the image so i don't see why you couldn't install it if you wanted to :P
As for cache and speed not to much the L2, L3 just denotes the proximity to the actual core. Obviously more is better but 8MB vs 6MB probably isn't going to make a noticeable difference when it comes to rendering.
IHow long would the same render take with the new 17" or 15" Macbook Pro (MBP) (Quad Core i7, 2.2ghz, 8gb RAM, internal 500gb 7200rpm drive, 1tb Video Card?)
And how big a difference in the render time would there be if the computer was doing another activity, say surfing the internet, or light Photoshop work, at the same time? Is there a substantial performance bump gained by paying the extra $250 and getting the 2.3ghz i7?
They benchmark very closely. The 2008 MP gets about 10,900 in Geekbench vs 11,100 for the MBP. It's pretty amazing when you think that a 2011 45W $2200 quad-core 5.6lb laptop can outperform a 2008 250W $3500 8-core 41lb workstation.
In real-world tests, they'll come out pretty closely tied though so it won't seem like an upgrade. There is a chance that some architectural changes can improve certain tasks dramatically though. Intel put some special functions in hardware that can accelerate some processes more than you'd expect.
1 hour for a two-pass encode isn't too bad but it depends on the length of the project. A 90-minute ProRes would take about 12 minutes to read entirely each time from a 7200 rpm drive + 4GB write = 25-minute best possible encode time and that's assuming the drive mostly sustains maximum throughput, which it won't.
When you start to multi-task, it adds to the storage workload.
If you got the MBP with a 256GB SSD e.g coming soon:
http://www.micronblogs.com/2011/03/c...ah-it?s-fast/
then you'd be able to speed up the read process considerably. Keep regular backups though. Compressor isn't the fastest encoder either though. Hopefully the next FCS update will get some optimisation.
The 2.3GHz laptop won't offer anything worth the price - you'd be best putting the money towards an SSD.
The main plus with the MBP is not that it gives you a major boost in performance but gives you the same performance as a workstation in your lap but remember that if you do lots of encoding, that fan noise might get pretty annoying after a while.
Compressor isn't the fastest encoder either though.
I would have thought with its multi core support through qmaster, that it would be fast.
What other encoder progs would you reccomend?
Which ones utilize all the processors, 64 bit architecture, etc?
Thanks again.
I would have thought with its multi core support through qmaster, that it would be fast.
What other encoder progs would you reccomend?
Which ones utilize all the processors, 64 bit architecture, etc?
Apple's video encoders tend not to be that fast compared to open source encoders like ffmpeg for m2v and x264 for H.264. The problem is, you can't get round this with FCP because you have to export from Final Cut first to get it into the other program, which means having to read a huge file from disc and write a huge file to disc, then read that huge file into the other program and export a faster encode so you ultimately end up wasting more time only to find out on occasions that DVDSP doesn't like the export.
I suspect a big slowdown in that process is FCP itself though - if you drop a movie file directly into Compressor, that won't be as bad but you can still try out other encoders like ffmpegX:
http://homepage.mac.com/major4/
I would have said VisualHub, which is much easier to use but it isn't developed any more.
The new FCS may be announced at NAB 2011 in 2-3 weeks and is rumoured to have had a big overhaul. Apple may not coincide an announcement with NAB though, they seem to not like doing that sort of thing any more but it's much in need of an update. Some features apparently will require 10.7 Lion due in June so they may bring out both together.
Is there a reason you use 2-pass encoding over 1-pass? Having a second pass lets the encoder know which bitrate it can get away with but it's easier to just pick a higher bitrate if you can get it to fit on the DVD. 2-pass only really helps for squeezing a video to the lowest bitrate.
you have to export from Final Cut first to get it into the other program, which means having to read a huge file from disc and write a huge file to disc
I usually will save a little time/disk space by exporting a reference movie instead. There was supposed to be a send to compressor function available in FCP without having to save even a reference movie, but I was never able to make it work. Certainly that working would be a huge improvement.
Lets see what FCS4 has in store...
Is there a reason you use 2-pass encoding over 1-pass?
True enough with a short enough export, a title less than 1 hour, CBR is fine once the bitrate calculation is made.
you can still try out other encoders like ffmpegX
I will look into the ffmpegx option.
I just figured after configuring a cluster using qmaster and then watching all those cores being utilized by Compressor in Batch Monitor, that it was actually being processed pretty fast.
I just figured after configuring a cluster using qmaster and then watching all those cores being utilized by Compressor in Batch Monitor, that it was actually being processed pretty fast.
Yeah, that's a pretty good way to get the most out of multiple CPUs - are you sending different segments to different nodes e.g 1-30 minutes to node A, 30-60 minutes to node B etc? That is a very good way to speed up an encoding but you'd be better with fast storage to do that.
I think it would be a good idea to check if your encoding is being limited by your storage performance. If you duplicate a large movie file (e.g 20GB+) and measure the time it takes to copy, that will give you a real-world estimate of your sustained drive speed.
Anyway, just looking at the options out there: I would go for the 750 GB disk at 5400 rpm instead of the 500 GB at 7200. For sustained R/W, the 750 GB should be a bit faster. Plus, it has more space, will be quieter, and will use less power.
are you sending different segments to different nodes e.g 1-30 minutes to node A, 30-60 minutes to node B etc?
I believe that is what happens automatically after I set qmaster prefs, and then encode using compressor, choosing my Mac Pro's cluster, to perform the encode.
I'm no expert for sure.
Many years ago I just used DVDSP to encode everything, (really slow), then I made my way to using Compressor, but without understand using the multiple cores, (still pretty slow, but better than DVDSP).
Then, finally a few years ago I got to understand Qmaster prefs, and how to get multiple cores involved in the encoding, by turning on/off sharing and setting the "number of instances" to 6 of the 8 cores I have, (even though there were some who recommended only using half the total cores), and then watching happily as Batch monitor would then show the file getting automatically split up into segments, (the size of which I have no control over), which then encode separately by different cores. (Audio never gets split.) When it is done it just reassembles the pieces into an m2v.
All I know is that the encoding time then became only a fraction of what I was used to.
I figured this was the best way to encode. Perhaps there are better ways today.
Do you know if ffmpegx takes advantage of multicore processors?
I know there are other pro programs out there like Cinemacraft Encoder MP (CCE-MP), but at $700 I haven't tried it. Of course there are the hardware encoders, which cost thousands.
My understanding is with CCE-MP there wouldn't be much of a save in encoding time, but, it supposedly does a much better job of encoding at lower bit rates, (for longer movies you want to squeeze into one dvd.)
I still can't encode a decent looking 2 hour m2v with compressor, no matter which settings I try. I think you may need a hardware encoder to get a watchable 2 hour encode that fits on a single dvd. Or at least CCE-MP.
I have a macbook pro 13 and an iMac 27. When the iMac is computing, I'll keep working on the macbook, which doubles as a business laptop. I don't know if it fits your usage model, but in my opinion getting a cheap 13" laptop + a nice iMac is about the same price as a 17" macbook pro + a second display, and it is a much better setup.
Thats is kinda where I have been at. Since March 2008 I have had a Dual 2.8ghz Quad Core MP, and a 15" C2D MBP. I let the MP grind away and use the laptop for everything else. I am own looking to try and get down to only one machine, hence this thread.
Anyway, just looking at the options out there: I would go for the 750 GB disk at 5400 rpm instead of the 500 GB at 7200. For sustained R/W, the 750 GB should be a bit faster. Plus, it has more space, will be quieter, and will use less power.
This is kind of shocking to me?? How can the 5400rpm be faster than the 7200rpm??
What about a third party WD 750gb 7200rpm drive?
I get the power usage and heat, but for video editing I have always opted for the 7200rpm drive over a larger and slower one.
Is this wrong? Can you point me to some benchmarks?
If this is the case, why ever get the faster rpm drive.
Then, finally a few years ago I got to understand Qmaster prefs, and how to get multiple cores involved in the encoding, by turning on/off sharing and setting the "number of instances" to 6 of the 8 cores I have, (even though there were some who recommended only using half the total cores)
That helps you avoid hitting memory/storage bottlenecks. Each process can use at least two processors so half the core count is a good guide.
If you have 100mbits source and 6 encodes, it will try to read > 600 mbits/s from the drive = 75MB/s and then the drive starts to slow things down. You really do need to balance the IO with the CPU usage.
All I know is that the encoding time then became only a fraction of what I was used to.
I figured this was the best way to encode. Perhaps there are better ways today.
Do you know if ffmpegx takes advantage of multicore processors?
The setup you have is really the best way to do it. While there are faster encoders, doing segments concurrently is faster than anything and other encoders often don't give you that flexibility - consumer encoding apps like ffmpegx certainly don't.
My understanding is with CCE-MP there wouldn't be much of a save in encoding time, but, it supposedly does a much better job of encoding at lower bit rates, (for longer movies you want to squeeze into one dvd.)
Someone here reviewed CCE vs Compressor:
http://www.dvinfo.net/forum/distribu...-encoders.html
It's a fairly old post but has some interesting info and there's a comparison of a few here:
http://digitalcontentproducer.com/vi...oder_shootout/
CCE seems to get quite a positive response.
I still can't encode a decent looking 2 hour m2v with compressor, no matter which settings I try. I think you may need a hardware encoder to get a watchable 2 hour encode that fits on a single dvd. Or at least CCE-MP.
If you don't use DL discs just now, that would give you some extra room, even though they cost a bit more. 2 hours into < 4.4GB would give a max bitrate of 4.8Mbps. For SD resolution, that should be ok but it's probably just Compressor failing to deliver good enough encoding here. If possible, you can crop the footage to a narrower aspect to reduce the info needing encoded.
I wish there was more of a push to get away from DVD format for distribution. Even just being able to author 720p 4-8Mbps H.264 to a DVD disc would do just fine. Blu-Ray is obviously the format of choice but needs everyone to buy new drives. SD cards would be my own preference as they can be fitted directly into a TV easily but cost is too high and no-one is willing to make the investment to make it happen.
This is kind of shocking to me?? How can the 5400rpm be faster than the 7200rpm??
What about a third party WD 750gb 7200rpm drive?
In the same trace on the disk surface, the denser disk (750 GB) will contain 50% more data than a 500 GB disk of the same size. So, at the same rotational speed, it will read 50% more data in the same duration. In other words, it will be 50% faster for sustained reads and writes. The rotational speed difference between 5400 and 7200 rpm is 33%, so the net speed benefit in this case is 17%.
Note that this is for sustained reads. For data latency, the higher rpm disk will be better. However, video is a sustained r/w process.
Thanks for the thoughtful comments. Sorry, I have been away a couple of days!
I need to understand this...
If you have 100mbits source and 6 encodes
Please tell me what you mean by "100bits source..."
6 encodes I'm guessing is 6 processes, or instances in Compressor.
it will try to read > 600 mbits/s from the drive = 75MB/s and then the drive starts to slow things down.
I understand this. (just not the original 100bits.)
You really do need to balance the IO with the CPU usage.
"I/O" - read and write speeds... "CPU usage" - Please explain this a little bit more.
Thanks for the CCE MP review links.
If you don't use DL discs just now, that would give you some extra room, even though they cost a bit more.
I wish. The money isn't even the most important concern. I just wish they worked! The DL DVD's I burn just aren't recognized in all set top DVD players. Many of the only slightly older models don't work. And, a lot of people still have their 5-10 year old, top of the line (then), DVD players.
2 hours into < 4.4GB would give a max bitrate of 4.8Mbps. For SD resolution, that should be ok but it's probably just Compressor failing to deliver good enough encoding here.
I don't like the way a 4.8Mbps bitrate looks. I never have. Perhaps CCE-MP/Bitvice, might be good at 4.8 with 10-20 passes... And if so, is it worth it.
Assuming again, a hardware encoder is the only way to get a store bought dvd quality on a 4.7gb disk.
If possible, you can crop the footage to a narrower aspect to reduce the info needing encoded.
Man I never even thought of that one! More black, less encoding, makes sense...
I wish there was more of a push to get away from DVD format for distribution. Even just being able to author 720p 4-8Mbps H.264 to a DVD disc would do just fine.
Hallelulah, if I may.
My 720p mkv's at that bitrate look so much better than store bought SD DVD's.
I hate I have to down convert 720p output to SD DVD just so my clients can watch it easily.
In the same trace on the disk surface, the denser disk (750 GB) will contain 50% more data than a 500 GB disk of the same size.
Got this much...
So, at the same rotational speed, it will read 50% more data in the same duration. In other words, it will be 50% faster for sustained reads and writes.
Still with you... If BOTH drives are the same speed. A 7200 rpm 750gb drive reads 50% more data than a 7200rpm 500gb drive in a single rotation... Therefore larger drive, same speed faster.
The rotational speed difference between 5400 and 7200 rpm is 33%, so the net speed benefit in this case is 17%.
OK, I'm getting lost here. Is this 17% this 33% subtracted from the 50%?
The question was how can a 5400rpm drive be faster than a 7200rpm one?
Is 7200rpm only faster when the two drives in question are the same size?
Is a 5400rpm 720gb HD then actually faster than a 7200rpm 500gb HD? By 17%
Note that this is for sustained reads. For data latency, the higher rpm disk will be better. However, video is a sustained r/w process.
Please explain data latency. Does it not pertain to video?
Disk latency is important when you have files, or fragments of files, located in different physical locations on the disk surface. A higher rpm disk will get to them more quickly. Video data, however, is not usually fragmented.
Please tell me what you mean by "100bits source..."
6 encodes I'm guessing is 6 processes, or instances in Compressor.
If your source footage is encoded in a format like ProRes, it can have 100Mbit/s bitrate. This means to decode one stream of video, it transfers that much data in real-time. So, if you have too many streams being read at once, you can create an IO bottleneck. This is worse than if you don't max out the CPUs.
"I/O" - read and write speeds... "CPU usage" - Please explain this a little bit more.
Ideally you want to max out your CPU capability while minimising your I/O so you don't create a bottleneck. Generally try to use the fewest Compressor processes that max out your CPU. If you have source files encoded in Apple Intermediate format, 6 at once may create a bottleneck. Here are some examples for simultaneous uncompressed streams capable with RAID:
http://kb.promise.com/KnowledgebaseArticle10065.aspx
If your source footage is ProRes standard, 6 processes from a single 7200 RPM drive should be ok but experiment with using 4, 5, 6 processes on the same job and check which gives the best results.
I wish. The money isn't even the most important concern. I just wish they worked! The DL DVD's I burn just aren't recognized in all set top DVD players. Many of the only slightly older models don't work. And, a lot of people still have their 5-10 year old, top of the line (then), DVD players.
Yeah, you'd think with DVD being such a ubiquitous format that compatibility issues wouldn't crop up but they certainly do.
I don't like the way a 4.8Mbps bitrate looks. I never have. Perhaps CCE-MP/Bitvice, might be good at 4.8 with 10-20 passes... And if so, is it worth it.
Assuming again, a hardware encoder is the only way to get a store bought dvd quality on a 4.7gb disk.
The benefit of hardware encoders is mainly the speed, not so much quality. In one of the above links, they tested out the $25,000 hardware encoder and found it not much better quality than Compressor. CCE was rated as the best.
I think it would be a waste to spend a lot of money on something like CCE though to try and get store-quality that uses 8.5GB discs onto 4.7GB ones. Depending on the number of discs you burn, you might be cheaper splitting the 2 hour movies over 2 discs.
As with anything, if it saves you money in the long run, it may be worth the $695:
http://www.cinemacraftusa.com/cinemacraft_encodermp.php
Maybe contact them to see if you can get a demo of it first.
Man I never even thought of that one! More black, less encoding, makes sense...
Noise reduction should help with quality too and not just from an immediate aesthetic point of view - delivery format encoders minimise space by compressing the differences between frames in groups. If a picture is grainy e.g shot in low-light, adjacent frames have more differences to encode. Noise removal helps reduce the differences between frames. This has an extra performance penalty however.
The question was how can a 5400rpm drive be faster than a 7200rpm one?
It's generally not the case in my own experience but some higher capacity drives use higher density platters using e.g perpendicular recording so you can find that a newer 5400 rpm drive will outperform an older 7200 rpm drive. Sometimes it's assumed that a 500GB drive will use a lower density configuration than a 750GB but it can use 2 platters instead of 3 with the same density.
I actually find very high capacity drives (1-2TB) to be pretty slow. The fastest seem to be 7200 RPM in the 250-500GB range i.e fewer high density platters at a high spindle speed.
But SSD beats all of them quite easily.
Thanks so much for the education!
This is worse than if you don't max out the CPUs.
I see how you calculated the I/O bottleneck, depending on the bitrate of the codec used. (I use Apple Pro Res 422 (not HQ or LT.)
But, how can I determine whether or not I am maxing out the CPU? Slower render times?
I just bought the 17" 2.2ghz i7 Quad Core on Thursday.
Let's say I set up Compressor to do a dedicated encode, (that is no other activity on the machine.)
I only have a maximum of 4 cores now, (unlike the 8 in the Mac Pro), so source footage with 100Mbit/s bitrate, I/O would no longer be a problem as with 4 occurrences it would only total 50MBs.
I know they recommend using only 1/2 the cores for an encode. But if the machine was dedicated shouldn't I use all four?
Depending on the number of discs you burn, you might be cheaper splitting the 2 hour movies over 2 discs.
This is exactly what I do to get reasonable quality... I do not know why "they" just can't give us the ability to burn a DL DVD that will play with legacy DVD players. It should be a software update...
But SSD beats all of them quite easily.
I am now considering getting an SSD for the first time, and installing it into the Optical drive slot. The Vertex 3, when it finally comes out, any day now, sounds pretty amazing.
But, even if I get an SSD, it would only be for OS X and Apps.
My video files would be on regular HD's. (I certainly couldn't afford enough SSD's GB's to use them as scratch disks for editing.)
I suppose once I am finished editing, and am in the encoding stage, I could save my FCP output onto the SSD for Compressor use. Would that speed encoding up much?
I am assuming this is the best use of an SSD. Any thoughts?
But, how can I determine whether or not I am maxing out the CPU? Slower render times?
Your activity monitor shows how much CPU a process uses. If it hovers round 350-400% on a quad core, that would be using about as much as you will get from it.
I know they recommend using only 1/2 the cores for an encode. But if the machine was dedicated shouldn't I use all four?
Your Macbook Pro i7 CPU is multi-threaded so it will show up as 8-cores. Each video process can use two threads so 4 is fine. Like I say though, use the fewest that max it out. Start with fewer processes and work up e.g try with 3 then 4 then 5 etc. You can test on a smaller file e.g a 5-10 minute clip to see which thread count gives you the best results.
I suppose once I am finished editing, and am in the encoding stage, I could save my FCP output onto the SSD for Compressor use. Would that speed encoding up much?
I would expect some speedup from the fact that you aren't reading and writing to the same drive but the output of 4GB or so is not going to be a slow process compared to reading in 50GB of source data - it would be better to have the larger source files being read from a faster drive.
As you say though, it's not really feasible to get your source footage on the SSD.