Q6600 overclock question for better X264 speed

[dvd_maniac]

I have recently upgraded to a Q6600 with a Gigabyte P35-DS3R and 2x1GB Ballistix DDR21066 and Zalman 9700 CPU Heatsink.
I am totally new to O/Cing and was wondering if it is worth the speed increase I would see in X264?
Seeing as I know almost nothing about O/Cing, How hard of a learning curve would it be to get this baby to about 3Ghz?

[Sagekilla]

I suggest you visit a hardware oriented site like www.tomshardware.com (Not the best, but it's a start) rather than asking here. All else being equal, going from 2 GHz to 3 GHz will usually net about a 40 - 50% speed increase. Likewise 2.4 to 3 will give you anywhere up to 25% speed increase.

But really, this isn't the right place to be asking how to overclock.

[dvd_maniac]

But really, this isn't the right place to be asking how to overclock.

Thanx for the advice. I have been looking at other sources for how to.... I just wanted to know if others here thought that the speed increase would be worth the learning curve for an O/C newbie rather than how to O/C.

Sorry if I didn't word it that way.

[Blue_MiSfit]

It absolutely will. The Q6600 is somewhat limited from the factory at 2.4 GHz. A G0 stepping chip should be able to hit 3 - 3.4 GHz on a good air cooler - like your Zalman.

You will see a significant speed increase in video encoding, as that's one area that really is CPU bound, and can scale nicely with 4 cores. Just start ramping up your FSB, and see how high you can go :) Your RAM will be no limiting factor, as even at 800 MHz 1:1, you will be at 3.6 GHz CPU - which is probably faster than possible on air.

Tighten timings, try for 667 MHz FSB (3 GHz CPU), and push that CPU hard!

There is a good thread somewhere in PC Hardware + Software that has a MeGUI benchmark. Good stuff! The octo-core monster that somebody built is a real screamer! Can you beat his record? Probably not! ;)

~MiSfit

[NanoBot]

Hi,

I got the same board and the same cpu a few days ago, and I was able to push it up to 378MHz FSB ( = 3,4GHz ) without increasing the cpu core voltage, which defaults to 2.125V in my case. The cooler is an Asus square pro.

Nevertheless I am not sure if I ever will use it at this speed to encode x264, since the cpu cores are heating up to 62-68 degrees celsius at this speed under full load. The cpu claims that the allowed maximum ( Tjunction ) is 100 degrees celsius, but even 68 degrees looks a little high to me.

When I moderately overclock the CPU to FSB 333 ( 3,0GHz ), the CPU only has coretemps between 48-55 degrees, which looks much better to me. 3GHz is a 25% ratio, 3,4GHz would be ~ 41% ratio, so I think its better to stick with the 25% and keep the cpu cool

C.U. NanoBot

[cogman]

Hi,

I got the same board and the same cpu a few days ago, and I was able to push it up to 378MHz FSB ( = 3,4GHz ) without increasing the cpu core voltage, which defaults to 2.125V in my case. The cooler is an Asus square pro.

Nevertheless I am not sure if I ever will use it at this speed to encode x264, since the cpu cores are heating up to 62-68 degrees celsius at this speed under full load. The cpu claims that the allowed maximum ( Tjunction ) is 100 degrees celsius, but even 68 degrees looks a little high to me.

When I moderately overclock the CPU to FSB 333 ( 3,0GHz ), the CPU only has coretemps between 48-55 degrees, which looks much better to me. 3GHz is a 25% ratio, 3,4GHz would be ~ 41% ratio, so I think its better to stick with the 25% and keep the cpu cool

C.U. NanoBot

ummm, 2.125 volts? that is VERY high, is this the board automatically setting it? I might recommend you manually set it to something no bigger then 1.5 volts (even that is fairly high) Seriously though, you shouldn't play with volts that high on your quad.

For a good OC guide, I would recommend anandtech's CPU board in their forums. People there are usually fairly helpful and there are hundreds of threads made on exactly the same topic you are starting here.

[NanoBot]

Hi again,

the default cpu core voltage of course was a typo, in fact it is 1.2125V on my Q6600. And I also need to correct the voltage used to reach the 3,4GHz ( I am running stability tests with prime95 atm ). The BIOS was configured to use "automatic" chipset and cpu voltages and I did not notice, that the cpu voltage was adjusted by the "automatic" to 1.3125V when the fsb is adjusted to 378MHz. And this cpu core voltage really is the minimum needed to run my Q6600 stable at this frequency, the automatic adjustment hit it to the point.

I think that today I will get a special type of "liquid" thermally conductive paste which is said to decrease the cpu temperatures by several degrees compared to standard paste. ATM I have 68 / 68 / 62 / 62 degrees C at 3,4GHz, lets see, if this special solution is able to cut these temperatures at least a few degrees. ATM I am using a noname standard thermally conductive paste.


C.U. NanoBot

[glen8]

www.overclockers.co.uk

VERY good forum, probably the best on the net

[Yoshiyuki Blade]

Also drop by http://www.xtremesystems.org/forums/and probably http://hardforum.com/ too. I don't know whether its my chip or my mobo, but I'm having trouble getting past 3.0 GHz on my Q6600 without pumping insane voltages. Maybe its the x38 motherboard, but I've seen a few others get some good overclocks at low voltages with the same board I have.

Also, are you measuring your temps with coretemp or speedfan (speedfan's temps are off by 15C unless you have the latest beta)? The max temperature Q6600s are specified to is 71C, but make sure you're measuring the core's temps, coretemp and speedfan are by far the most accurate for this.

On a sidenote, I'm getting a really weird bug in x264 on my quad core system. It seems to encode the first few seconds of the video fine, but then it starts going whack (encoding fps skyrockets for no apparent reason) and eventually crashes. Anyone experienced this? I'm on Vista x64, but I've used Vista x64 on a previous system just fine.

Update: Removing 2 sticks of RAM solved the problem. I had 8 gigs of RAM, which is a ton for normal usage O_O. Now I have to find out whether its the mobo that can't properly utilize 8 gigs of RAM (4 x 2048), whether one or more of the sticks is defective, or if its a problem with x264.

[burfadel]

What room temperature is your computer sitting in? A warm room will significantly reduce the safetyness of overclocking. If you overclock to say 40%, and it looks good temp wise at 22C, if your rooms hits 28C - it will warm up due to the computer!, then the temperature may start to be of a concern!

Its also vitally important to have good case airflow. The best CPU air cooler in the world will not help if all its going to do is spread that heat around the inside of the case, to affect your HDD's (the main cause of failure of HDD's is from temperature these days so I've read), your RAM, graphics card etc etc.

To your question of whether you can get faster encoding, like others have said thats a definite yes! - just be sensible and keep your eyes on the temps. Also keep an eye on HDD temp.

The Core 1/2 temp is apparently different to the CPU temp? If you use PCWizard it will show you both, and it will also show you your hdd temp.

[NanoBot]

Hi,

first of all, the usage of the "liquid" thermal conductive reduced my core temperatures to 62/62/58/57, so I can really recommend this.

My first test was to reencode a 1920 x 1080i h.264 ts from a DVB rip to a lower bitrate. I did the same with the same x.264 settings a while ago using my old cpu.

A64 x2 3800+ @ 2,4GHz ~ 5fps
Q6600 @ 3,4GHz ~ 15fps

To get an idea how much the difference between a dualcore and a quadcore would be I will do another test by using the taskmanager to assign only two of the four cores to MeGui.

Nevertheless, at least here in germany the choice between an dualcore (e.g. E6600) and a quadcore (Q6600) should be easy: The E6600 is priced about 150€ (tray), while the Q6600 ist priced about 215€ (tray)


C.U. NanoBot

[Yoshiyuki Blade]

Hi,

first of all, the usage of the "liquid" thermal conductive reduced my core temperatures to 62/62/58/57, so I can really recommend this.

My first test was to reencode a 1920 x 1080i h.264 ts from a DVB rip to a lower bitrate. I did the same with the same x.264 settings a while ago using my old cpu.

A64 x2 3800+ @ 2,4GHz ~ 5fps
Q6600 @ 3,4GHz ~ 15fps

To get an idea how much the difference between a dualcore and a quadcore would be I will do another test by using the taskmanager to assign only two of the four cores to MeGui.

Nevertheless, at least here in germany the choice between an dualcore (e.g. E6600) and a quadcore (Q6600) should be easy: The E6600 is priced about 150€ (tray), while the Q6600 ist priced about 215€ (tray)


C.U. NanoBot

Sounds like very good results :D. My old athlon x2 4400+ OCed to 2.53 GHz averaged 1.8 fps (I encode at insanely high settings and at 1280x720), and now my Q6600 does ~3.2 fps at stock 2.4 GHz. Im currently doing tests on the memory to find out what causes instability with all 4 slots of memory used up (with 8 gigs).

[burfadel]

Sounds like very good results :D. My old athlon x2 4400+ OCed to 2.53 GHz averaged 1.8 fps (I encode at insanely high settings and at 1280x720), and now my Q6600 does ~3.2 fps at stock 2.4 GHz. Im currently doing tests on the memory to find out what causes instability with all 4 slots of memory used up (with 8 gigs).

1T/2T timing :scared: <-- only because it affects the speed a fair bit. I take it you've increased the RAM voltage? one thing to note though, the standard DDR2 voltage is supposed to be 1.8v, but there are DDR2 module (pc6400 modules say), that require up to 2.0 at stock speeds! overclocking these modules means that you'd have to give more voltage than seems necessary.

I take it also that the RAM modules are exactly the same type? (brand/version etc). You can get away with basic mixing, as long as the two modules on the same DDR2 channel are the same. Just helps to avoid timing issues etc.

[Yoshiyuki Blade]

1T/2T timing :scared: <-- only because it affects the speed a fair bit. I take it you've increased the RAM voltage? one thing to note though, the standard DDR2 voltage is supposed to be 1.8v, but there are DDR2 module (pc6400 modules say), that require up to 2.0 at stock speeds! overclocking these modules means that you'd have to give more voltage than seems necessary.

I take it also that the RAM modules are exactly the same type? (brand/version etc). You can get away with basic mixing, as long as the two modules on the same DDR2 channel are the same. Just helps to avoid timing issues etc.

Yeah, theyre all the same and are rated at 2.1v, which I've set them to manually in the BIOS. DDR1000 @ currently at 750 MHz (1:1 ratio) timings are 5-5-5-15 2T.

[burfadel]

Yeah, theyre all the same and are rated at 2.1v, which I've set them to manually in the BIOS. DDR1000 @ currently at 750 MHz (1:1 ratio) timings are 5-5-5-15 2T.

Should be able to squeeze out CAS 4 running them slow like that! I take it 1T timing doesn't work at all?

There are several other timing options as well, its most likely one of these settings that need to be increased slightly. Many of those options may not be available under the BIOS unfortunately. I use Ntune, only because I have an Nforce chipset, I don't know of any good programmes to set the other DRAM timing options on Intel chipsets.

[Yoshiyuki Blade]

Should be able to squeeze out CAS 4 running them slow like that! I take it 1T timing doesn't work at all?

There are several other timing options as well, its most likely one of these settings that need to be increased slightly. Many of those options may not be available under the BIOS unfortunately. I use Ntune, only because I have an Nforce chipset, I don't know of any good programmes to set the other DRAM timing options on Intel chipsets.

People usually prefer tweaking via BIOS. It takes a lot of restarts and all, but it works well. Anyhow, I still don't understand why x264 crashes with 8 gigs installed. XviD seems to work fine under this configuration.

[Sagekilla]

People usually prefer tweaking via BIOS. It takes a lot of restarts and all, but it works well. Anyhow, I still don't understand why x264 crashes with 8 gigs installed. XviD seems to work fine under this configuration.

Perhaps x264 (I don't know why it wouldn't though) isn't compiled to support more than 4 GB (8 GB in this case) properly?

[burfadel]

People usually prefer tweaking via BIOS. It takes a lot of restarts and all, but it works well. Anyhow, I still don't understand why x264 crashes with 8 gigs installed. XviD seems to work fine under this configuration.

Thats true, I do too :) However, there aren't too many bioses that actually have have every configurable DRAM option available! How many are there for DDR2 anyways? I know there's more than 10

[akupenguin]

x264 doesn't ever use more than ~300MB (1080p 16ref multithread esa ...). However, I have heard some reports that Avisynth crashes if you have too much memory, since it tries to use some fraction of your total memory but doesn't actually support more than 2GB. Nothing I can do about that.

[Sagekilla]

x264 doesn't ever use more than ~300MB (1080p 16ref multithread esa ...). However, I have heard some reports that Avisynth crashes if you have too much memory, since it tries to use some fraction of your total memory but doesn't actually support more than 2GB. Nothing I can do about that.

Really? My x264 actually uses up 700 MB RAM encoding @ 864x480

[akupenguin]

And you're using Avisynth input, right? Avisynth being a dll, it counts as part of the same process.

[squid_80]

Hmmm but Yoshiyuki Blade says xvid encoding works fine. The easy test would be to add setmemorymax(1024) to the start of the script and see if it still crashes.

[Yoshiyuki Blade]

Thats true, I do too :) However, there aren't too many bioses that actually have have every configurable DRAM option available! How many are there for DDR2 anyways? I know there's more than 10

I have no clue haha. I usually like to keep most of the things default and get a modest overclock.

Hmmm but Yoshiyuki Blade says xvid encoding works fine. The easy test would be to add setmemorymax(1024) to the start of the script and see if it still crashes.

That actually seemed to do the trick. Thanks a lot! So what exactly was the cause of the problem? avisynth or x264? I'm using the Cef's build r709 and AVS 2.5.8 Alpha 2.

EDIT: Seeing that the encoding process is currently using 1.4 gigs of memory (x264 using ~400MB?) I guess avisynth was using more than it could handle without setmemorymax(1024). I think before it went up to 1.8 gigs before ultimately failing O_O. Also x264 does not seem to be utilizing 100% of the CPU. It hovers around the 70s and 80s, dips to the 50s for a second or two, and occasionally goes up to the 90's and 100. Are there any tricks to optimize quad core?

[Sagekilla]

And you're using Avisynth input, right? Avisynth being a dll, it counts as part of the same process.

Aha, that explains things quite a bit.. Still, I don't really care if my x264.exe process is eating up 400, 700, or even 1.5 GB of memory since I have more then enough memory for even 1080p + 16 refs (Just tested this out to prove it to myself that I could.)

[foxyshadis]

Also x264 does not seem to be utilizing 100% of the CPU. It hovers around the 70s and 80s, dips to the 50s for a second or two, and occasionally goes up to the 90's and 100. Are there any tricks to optimize quad core?

Avisynth is the bottleneck here, too. If it's already using SetMTMode, there isn't a lot more you can do.

[Yoshiyuki Blade]

Avisynth is the bottleneck here, too. If it's already using SetMTMode, there isn't a lot more you can do.

Ah ok. Yeah as a solution, I started to do 4 encodes simultaneously to make use of the available resources. My CPU usage is at 100% all the time now.

[burfadel]

What the resource handling of x264 like? Does x264 take cycles from the avisynth part of the encoding chain? Ideally, x264 would not take any cycles from avisynth, and as a priority thing would act sacrificial when other programmes request CPU cycles instead of those programmes taking the cycles away from the avisynth chain. Then, have a logic inside x264 that uses up the cycles that avisynth is not...?

[Sagekilla]

What the resource handling of x264 like? Does x264 take cycles from the avisynth part of the encoding chain? Ideally, x264 would not take any cycles from avisynth, and as a priority thing would act sacrificial when other programmes request CPU cycles instead of those programmes taking the cycles away from the avisynth chain. Then, have a logic inside x264 that uses up the cycles that avisynth is not...?

Yes, I believe it works something like that, but aku probably knows much better then I do.

But, Avisynth IS frame serving x264, so it wouldn't be a good idea at all for x264 to steal cpu cycles from Avisynth, or it wouldn't have any video frames to be encoding in the first place!

[akupenguin]

Avisynth and x264 run in separate threads. It's completely up to the windows scheduler how cycles are divided between them, x264 doesn't and can't make any such choice.

[squid_80]

It could give the avisynth thread a higher priority...

[steve77]

This could help your decision:

http://forum.doom9.org/showthread.php?t=122318

It's a comparasion of dual core and quad core processors, overclocked and factory std.

Overclocking is worth it... but it does shorten the lifespan of you processor. I used to be able to run my E6300 @ 3.25GHz, now I have to put mad voltage to get to 2.8Ghz.... I just run stock volts @ 2.4 now.

[burfadel]

This could help your decision:

http://forum.doom9.org/showthread.php?t=122318

It's a comparasion of dual core and quad core processors, overclocked and factory std.

Overclocking is worth it... but it does shorten the lifespan of you processor. I used to be able to run my E6300 @ 3.25GHz, now I have to put mad voltage to get to 2.8Ghz.... I just run stock volts @ 2.4 now.

Overclocking can be safe if you overclock to reasonable values. Overclocking a E6300 (1.86Ghz) to 3.25Ghz you should expect the life expectancy to decrease. As soon as you get to the point where you need to increase the voltage of the CPU you're in the life-shortening area. Increasing the voltage of the RAM is generally ok up to 2.1V, 2.2V, the reason being that most DDR2 modules don't conform to the DDR2 spec of 1.8V. Some DDR2 modules from well known brands at 800Mhz Cas 5 require more than 1.8V - and some more generic brands can easily do 800mhz Cas 4 as the specified 1.8V.

Anyways, keep overclocks within reasonable values, and have very good cooling - this is important. An overclock with the Core 2's of 20 percent is considered quite safe (can be down at stock), just because it can overclock 75 percent doesn't mean its a good thing to do so!