Sign in with
Sign up | Sign in
Your question
Solved

Video Encoding: Core 2 beats i7?

Last response: in CPUs
Share
April 16, 2010 1:03:31 AM

OK, so I have 2 systems, a HTPC with Core 2 Duo 3.0 GHZ (E8400), 4GB DDR2 800 RAM and a 9400GT video card. The second is a newer laptop, Toshiba Qosmio with an i7-720QM (1.6GHz + , 6GB DDR3 1066 RAM and a GTS 250M.

When using AutoGK, the HTPC beats the laptop by a few minutes. WTF?

Now I know that laptops are usually always going to be slower than comparable desktops, but I would have thought a generation skip would have given at least a little bit of advantage.

Both machines have mid-range SSDs for their OS drive. I tested encodes to and from the SSD drives as well as placing the SSD as the source OR target drive, just to make sure that storage wasn't a bottleneck.



I'm considering upgrading my desktop, and was trying to decide between a Q9550 or an i7-930. Both processors at at comparable prices at MicroCenter, and the price of DDR2 seems to keep doing up while DDR3 is coming down.


Can anyone explain why the C2D outperformed the i7? Is it the limitations of the application (AutoGK) to utilize the hardware or are these new processors a bunch of smoke and mirrors? (OK, I know they aren't, but still...I was a bit surprised) I guess the "turbo boost technology" isn't as great as advertised..

If you had to choose a processor whose biggest usage was audio/video encoding, which would it be?

Thanks.
a b à CPUs
April 16, 2010 1:07:41 AM

My guess is that the software isn't optimized for the i7.
I run/ran encoding on a Q6600 @3.6 and an i7 and even at stock, the i7 is quite a lot faster. (I'm using Vegas btw)
m
0
l
a b à CPUs
April 16, 2010 2:02:43 AM

If it's single or dual threaded, that's exactly the result I'd expect. With 4+ threads, the i7 should pull ahead.
m
0
l
Related resources
a c 127 à CPUs
April 16, 2010 2:21:48 AM

Even though it could be due to the fact that he is comparing a laptop CPU to a desktop CPU. Laptop CPUs have advanced greatly but are geared more towards efficiency than performance. Its possible that because the laptop CPU runs at a lower clock unless turbo kicks in under certain situations it is going to go slower than the desktop CPU.

Thats my thoughts because the app would have to be highly multithreaded optimized to be able to use a Core i7 to its max and beat a higher clocked CPU.
m
0
l
a b à CPUs
April 16, 2010 2:52:49 AM

AutoGK does have some weird threading issues... that 45% less clock speed for the i7 probably is also part of the problem.


As for the which CPU for your desktop, I think the choice is obvious from CPU price alone... For the price, if you were just going to do a drop in replacement, the Q9550 does look interesting. Don't knock the i5 750 with the lower platform cost and Thuban might also be intriguing if you wait a little more.

m
0
l
April 16, 2010 3:05:26 AM

Although it is a crude test, my friend suggested I launch task manager and/or resource monitor and see what the CPUs were doing. Auto GK is actually an app that takes advantage of 3 other applications to completely encode video to audio. During the audio process, lame only utilized a single processors, but Virtual Dub Mod seemed to use all processors. What was interesting was that the Core2Duo box ran steadily at 90-95% CPU while the i7 machine stayed at the 65-75% range. (i.e....didn't max out the machine)

It sounds like another program might do more (Vegas...am not familiar with but will have to Google) on the i7, and I'm totally up for that.

Basically, I'm looking to take advantage of the current CPU sale at MicroCenter, which has 3 different procs at about $100 cheaper than NewEgg. The 860 is $50 cheaper, the 930 is $95 cheaper, and the Q9550 is $80 cheaper. I want to go with one of those 3, and have read about the differences, so I thought I knew what I was getting into.

Then the "real world" test showed me different. I will give other software a test to see if I can make a difference, I was just amazed that the 4 core HT processor got outgunned by the Core2Duo in my first crude test.

Thanks for any more info and the quick replies.
m
0
l
April 16, 2010 3:10:55 AM

Stardude82 said:
AutoGK does have some weird threading issues... that 45% less clock speed for the i7 probably is also part of the problem.


Hmm...seems that Turbo Boost isn't all its cracked up to be. I was under the impression that the i7-720 would run as high as 3.2 GHz, but apparently it is only for short bursts?

I wonder if tweaking the affinity and setting the app to run on only 4 of the cores would allow the turbo boost to provide better results?
m
0
l
a b à CPUs
April 16, 2010 3:31:02 AM

Shadow703793 said:
My guess is that the software isn't optimized for the i7.
I run/ran encoding on a Q6600 @3.6 and an i7 and even at stock, the i7 is quite a lot faster. (I'm using Vegas btw)


Maybe it's like in the good old days when the P4 with HT was slower than without in some apps?
m
0
l
April 16, 2010 7:48:33 AM

Shadow703793 said:
My guess is that the software isn't optimized for the i7.
I run/ran encoding on a Q6600 @3.6 and an i7 and even at stock, the i7 is quite a lot faster. (I'm using Vegas btw)



Which version of Vegas? I don't think I can afford the Pro version, but I'm not sure the basic version would cover all my needs.
m
0
l
a b à CPUs
April 16, 2010 8:12:43 AM

MGHolley said:
Hmm...seems that Turbo Boost isn't all its cracked up to be. I was under the impression that the i7-720 would run as high as 3.2 GHz, but apparently it is only for short bursts?

The more threads that are use the lower that Turbo will clock the CPU. So it's a balancing act. You can use less threads and have a higher clock speed or more threads and a lower clock speed. Which is better depends on the workload.
m
0
l
April 16, 2010 9:17:50 AM

I disabled threads 1,3,5 & 7, thinking that they were the turbo-boosted/hyper-threaded processors, but got slower speed times. I'll try different combinations to see if I get lucky. However, since I don't seem to be bottlednecked by anything but the CPU, I think I'll stay with the LGA775 and get the Q9550. I think if I were building/buying a whole new machine it would make sense, but since 4GB of RAM seems to be sufficient for the task, I'll save the $.

One day, I hope that video decoding/encoding will take advantage of GPU as well as CPU. Then the newer technologies and the more powerful SLI/Crossfire will actually mean something to me.


Thanks again for the help! :) 
m
0
l
a b à CPUs
April 16, 2010 10:00:29 AM

MGHolley said:
OK, so I have 2 systems, a HTPC with Core 2 Duo 3.0 GHZ (E8400), 4GB DDR2 800 RAM and a 9400GT video card. The second is a newer laptop, Toshiba Qosmio with an i7-720QM (1.6GHz + , 6GB DDR3 1066 RAM and a GTS 250M.

When using AutoGK, the HTPC beats the laptop by a few minutes. WTF?

Now I know that laptops are usually always going to be slower than comparable desktops, but I would have thought a generation skip would have given at least a little bit of advantage.

Both machines have mid-range SSDs for their OS drive. I tested encodes to and from the SSD drives as well as placing the SSD as the source OR target drive, just to make sure that storage wasn't a bottleneck.



I'm considering upgrading my desktop, and was trying to decide between a Q9550 or an i7-930. Both processors at at comparable prices at MicroCenter, and the price of DDR2 seems to keep doing up while DDR3 is coming down.


Can anyone explain why the C2D outperformed the i7? Is it the limitations of the application (AutoGK) to utilize the hardware or are these new processors a bunch of smoke and mirrors? (OK, I know they aren't, but still...I was a bit surprised) I guess the "turbo boost technology" isn't as great as advertised..

If you had to choose a processor whose biggest usage was audio/video encoding, which would it be?

Thanks.

Hmm...

Here are a few possible reasons off the top of my head:

1. 6GB of RAM on a Dual Channel Processor relegates the entire thing into Single Channel Mode. The 720QM is a Dual Channel Processor (not Tri-Channel). For best performance... pair memory in groups of two.
2. Clock speed. The 720QM has a maximum turbo speed of 2.8GHz which means that it is most likely to perform between its 1.6GHz clock speed and 2.8GHz max speed and may not be hitting the max Turbo speed. If you launch CPU-z while running AutoGK it might shed some light as to its operating speed under load.


As for AutoGK well it is Multi-threaded:
m
0
l
April 16, 2010 10:13:27 AM

ElMoIsEviL said:
Hmm...

Here are a few possible reasons off the top of my head:

1. 6GB of RAM on a Dual Channel Processor relegates the entire thing into Single Channel Mode. The 720QM is a Dual Channel Processor (not Tri-Channel). For best performance... pair memory in groups of two.
2. Clock speed. The 720QM has a maximum turbo speed of 2.8GHz which means that it is most likely to perform between its 1.6GHz clock speed and 2.8GHz max speed and may not be hitting the max Turbo speed. If you launch CPU-z while running AutoGK it might shed some light as to its operating speed under load.


As for AutoGK well it is Multi-threaded:
http://images.tweaktown.com/content/3/1/3177_17.png



Good to see it is multi-threaded. It was strange. By disabling all but one of the processor threads in task manager, I was still able to get nearly as good times as when all threads were enabled. 29 minutes as opposed to 27 minutes from previous attempts.


So the laptop shipped with 6GB of RAM. There are only 2 slots, and one with Hyundai RAM (WTF?) according to CPU-Z with max bandwidth of PC3-8500F (533MHz) and the second with Samsung RAM and same max bandwidth. I'm not quite sure how that all plays out in the end. I've never had mixed RAM like that before in dual channel config. I suppose it is possible that the RAM is from the same supplier with different NAND (is that right technololgy?) sources? Regardless, I could see how that might affect the system somehow. Just not technical enough to understand it immediately. I could see how this might be confused as a triple channel kit in a dual channel system. The end result of mismatched dimms could have negative effects even if they got the number of channels right. :) 


Thanks for the benchmark numbers. Could you list the source?
m
0
l
April 16, 2010 10:39:55 AM

I decided to test another program, Handbrake, to see if I would get better results. It seemed to push the processor at a similar CPU usage standpoint (70-80%), but CPU-Z seemed to stay fairly stable at 1700 MHz with a x13 Multiplier (133 MHz Bus Speed).

AutoGK bounced all over the place, sometimes pushing the multiplier as high as x21 with 2700 MHz Core Speed. I will have to keep running a few more tests. At the end of the day, I realize that laptop does not equal desktop, but I hope I figure out something to give me confidence. Retailers frown on "try before you buy" processors.
m
0
l
a b à CPUs
April 16, 2010 11:04:18 AM

I have two Desktops. Core 2 Quad 2.33 GHZ 8 GB RAM. The other is a i7 2.8 GHZ Lynnfield 4 GB RAM. Both have Radeon 4870's for graphics. Both machines run Win 7. Corel Videostudio X3 on both machines encoding the same file started at the same time. The i7 was a solid 40% faster. The Core 2 Quad ran at 95% across all 4 Cores while the i7 was around 30% across all cores. The i7 runs at a lower percentage as this chips design has reduced latency on its instruction delivery. It can do more with lower clock speeds than the Core 2. In your case I think the Laptop is inhibited by the likely use of shared controllers and a chipset designed for things like low power consumption over raw performance. Also to be considered is the HDD on the laptop may not be as fast as your desktop.
m
0
l
a b à CPUs
April 16, 2010 12:34:13 PM

Are you sure all the components are the same?
m
0
l
a b à CPUs
April 16, 2010 2:18:20 PM

MGHolley said:
Which version of Vegas? I don't think I can afford the Pro version, but I'm not sure the basic version would cover all my needs.

I use Vegas Pro 9: http://www.newegg.com/Product/Product.aspx?Item=N82E168...
(been using it since v6)
Be warned: This is not exactly the most user friendly program. It will take a bit to get used to.

One more thing: I have heard that Windows 7 will not allow Turbo for the mobile CPUs until it is set for "Max Performance" mode in power management,etc. Not sure if this is true, but this is some thing you can look into
m
0
l

Best solution

a b à CPUs
April 16, 2010 9:21:23 PM

My Results with a Desktop Grade Core i7 920 @ 3.8GHz (One of my Encoding Servers).

Looks like it's an AutoGK issue. It appears to not like HyperThreading. Try Disabling HyperThreading in the BIOS :) 


Share
a b à CPUs
April 17, 2010 5:34:15 AM

So I went ahead and also tried it on a Phenom II X3 720 and this is what I observed:


It does appear that AutoGK does not like Hyperthreading.
m
0
l
a b à CPUs
April 17, 2010 7:37:10 AM

MGHolley said:

I'm considering upgrading my desktop, and was trying to decide between a Q9550 or an i7-930. Both processors at at comparable prices at MicroCenter, ..............................

Can anyone explain why the C2D outperformed the i7? Is it the limitations of the application (AutoGK) to utilize the hardware or are these new processors a bunch of smoke and mirrors? (OK, I know they aren't, but still...I was a bit surprised) I guess the "turbo boost technology" isn't as great as advertised..

If you had to choose a processor whose biggest usage was audio/video encoding, which would it be?

Thanks.


Obviously the I7. The laptop I7 is running at 1.6 Ghz. The C2D is 3ghz. The hard drive is likely slow on the laptop despite the SSD stuff. This is not a surprising result. Like jimmysmitty said, lappys are tuned for temp/power, not performance.
m
0
l
April 17, 2010 11:04:41 AM

Best answer selected by mgholley.
m
0
l
April 17, 2010 11:05:02 AM

I'm not ready to discount the laptop being underpowered just yet. After all, this laptop is a high end desktop replacement. 1066 FSB with RAM and CPU to match. AutoGK is undoubtedly the main problem, but it still irks me that a 3 years newer processor still under performs. I know that its not an apples to apples comparison, but I still think I should be able to tweak something to make this work out. I also wish Intel or Toshiba offered a little better control over the mobile processor to make it work for different applications. I wonder if running an app in XP mode would adjust the usage of the processor. Microsoft couldn't have been that smart, could they?

I will play with HyperThreading disable in BIOS and MSConfig. On first pass, I was able to get it down to just one processor in TaskManager/Resource Monitor, so I'm curious to see if I can get it down to just 4, with each core not hyperthreading but performing at the maxed out 2.8 GHz I'm promised. lol.

Found this interesting: http://www.notebookcheck.net/Review-Intel-Core-i7-Proce...

Perhaps in a couple of years I can breathe some new life into the notebook with a 920XM and the newer SSD. Only time will tell.

Thanks to all who replied. I think I have enough information to make a purchase. I also have plenty to tweak in the hopes of getting more out of my laptop.

Have a great weekend everyone....
m
0
l
a b à CPUs
April 18, 2010 9:02:41 AM

You are missing the point. The i7 gains its advantages from its really low latency. As the Core 2 Quad has a higher latency its works harder for less results. The rules are changed a little bit on how a CPU performs. Also to be considered is the i7 boosts a single core which has proven most useful for gaming. The i7 is what the Core 2 was to the Pentium 4.
m
0
l
May 1, 2010 7:22:40 AM

Well, my new i7-960 desktop machine just rendered a 120min movie in 21 minutes @166fps.

It blew away the Q9550 in my test machine, so I'm glad I listened to the hype. Didn't even bother loading up the perfmon util....with results like that who cares if its maxed out.
m
0
l
!