Better Performance for Cheap?

[consultant]

I'm transcoding my first Blu-Ray (Ratatouille - 1hr, 50min, 26GB source file it looks like) to 720P using RipBot264 (with AnyDVD). At 7% complete it is estimating (4.5 hrs or ~ 8 fps)
I've got a quad processor Q8200 (2.33Ghz) with 6GB RAM on Vista 64-bit with but with onboard Intel G45 chipset video card (1GB shared memory I think?). I'm just curious if 4.5 hrs is already pretty good or if a better video card will get me significantly better performance for transcoding video?

[burfadel]

Changng the video card won't make a scrap of difference. The only way a graphics card can slow down encoding (unless you use a decoder that uses the hardware decoding chip, which is unlikely) is if the graphics card is drawing things on the screen to slow, which is unlikely also, even with the G45. One thing to note though, its to run the latest drivers for your intel graphics.

The encoding and decoding is all done by the CPU, so that is your limiting factor since you have enough RAM. The other limiting factor is hard disk speed, but that shouldn't be an issue either.

[consultant]

Ya, well the CPU isn't something I'm going to change since I just bought the computer 6 months ago and did a lot of research as far as best bang for the buck on the CPU - although it wasn't with transcoding 1080P video in mind. But I only buy about one new BD a week so it's no problem to just transcode it overnight. Its just I have about 20 to start with and I'm a bit impatient. I can live with 8FPS 1080 to 720P conversion.

[Ghitulescu]

Oh, yes, the new generation of H.264 codecs and encoders may use the CUDA (nVidia) to speed up things. And I mean speed up.

[DarkZell666]

I believe you can safely overclock your CPU to 2.8Ghz (maybe 3.0 if you're lucky) on stock cooling. That's the only real "faster for cheap" way :) I've done 2.5ghz => 3.2ghz ghz on a Q9300 on stock cooling without touching voltages. It's dead easy, and way worth it. Just pump up the FSB step by step until the computer stops booting. Come back to the previous working FSB value you found and run a stability test. If it doesn't crash after a 2h benchmark test, you're done :) I'll let you read real overclocking articles for more details ... and watch out for the temperature !

[CpT]

I believe you can safely overclock your CPU to 2.8Ghz (maybe 3.0 if you're lucky) on stock cooling. That's the only real "faster for cheap" way :) I've done 2.5ghz => 3.2ghz ghz on a Q9300 on stock cooling without touching voltages. It's dead easy, and way worth it.

If your board supports clocking I second the overclock option.
I have a q9550 @ 3.9 up from 2.8 and the difference is massive. Waiting for a new board to push it further.

[burfadel]

Just be careful if you do overclock, there is rarely such a thing as a 'safe' overclock.

Even if your CPU can go significantly faster there are other considerations. The first is temperature. If you are in a room that varies in temperature, and can get quite warm, a moderate overclock is all the I'd recommend. A moderate overclock is from 333 --> 400 fsb, which is generally considered safe (watch the speed of the memory though, it may need adjusting).

Separate considerations to the load on the CPU due to the overclock is the extra strain on motherboard components. The motherboard chipset can get ridiculously hot with a high overclock, not to mention other components of the motherboard. So now, you now only have more heat from your CPU but you have more heat from your motherboard components you have to consider when you take into account your graphics card and hard drives.

Despite what has been claimed by certain people, I strongly believe heat is a major cause of drive failure. In a good case, the heat from the motherboard shouldn't interfere with the drives, but man cheaper cases the drive heat up as well, and personally I don't think 60C, 70C for a drive will do it any good!

Also consider the extra load on your power supply. It is minimal, but the extra load is there :)

I wouldn't recommend doing as CpT has done, that is an excessive overclock and requires chaning voltages etc., it may mean one day when he turns it off it won't turn back on again! :) thats very much an enthusiast overclock!

So basically, just ramp the fsb up to 400, watch the memory speed, and go from there :) you can then tweak the other settings, and the memory timings. Just hope you have at least 1066Mhz RAM.

Alternatively you could sell your q8200 CPU, and puchase the q9505. They don't cost excessively more than the q8200, and it is the highest CPU you can buy now for your computer without going to socket 1156 Core i7 (the core i7 8xx are socket 1156 whereas the core i7 9xx are socket 1366). With the q9505, again you can go to 400fsb or even a bit more.

[CpT]

As long as he keeps stock voltages there's 0 risk. He'll just cap out sooner. But dropping the memory multiplier will most likely be required if the bios even has the option.

My q9550 E0 stepping hits 3.8 on stock cpu voltages. @ 3.9 it requires a small bump and above 3.95 my board gets flaky due to the fsb speeds.
As long as you don't touch any voltages there's 0 heat increase. Heat is Not generated by speed/mhz. Its only generated by voltage.

Tbh I don't think that board even has clocking options. You could try and google "overclock +board model here" and see what turns up. I had a q8300 that hit 425fsb on stock volts and a stock cooler = 3.1ghz or so. I sold it b4 trying it in a better board but it had more in it for sure.

Before clocking do some googling on the subject. Its very easy these days if your board supports it but things like locking your pci-e and pci bus are important. Google first, clock later.

GLHF

[kypec]

As long as you don't touch any voltages there's 0 heat increase. Heat is Not generated by speed/mhz. Its only generated by voltage.
I strongly disagree with your opinion on this matter. Do you know what generates most heat in semiconductors [or simply gates]? It's the loss caused by switching from one stable state to another; ON to OFF, 1 to 0 or vice-versa. The higher the operating CPU frequency (clock) the more transitions occur in same time period hence producing more losses -> more heat generated over the same time period.

[DarkZell666]

I strongly disagree with your opinion on this matter. Do you know what generates most heat in semiconductors [or simply gates]? It's the loss caused by switching from one stable state to another; ON to OFF, 1 to 0 or vice-versa. The higher the operating CPU frequency (clock) the more transitions occur in same time period hence producing more losses -> more heat generated over the same time period.

I'll have to second that remark, bumping the FSB up ~80mhz alone pushed the thermometer 10°C further up in my case.

NB : Let's not make this thread an overclocking thread, there are dozens of guides out there, it was just a random idea to get a better speed-for-money compromise :)

[burfadel]

Yeah thats right :) temperature wise, 400mhz for a 333mhz cpu is generally the 'safe' (I use that world very lightly) limit to cover a vast range of scenerio's. Some can't even get to 400, in which case the 'safe' amount is the maximum for stability take around 10mhz to be on the 'safer' side.

[CpT]

Do you know what generates most heat in semiconductors [or simply gates]?

I am well aware gates and their functioning. But your statement only tells me while you may understand electronics, you've done very little overclocking.

For example bumping a cpu from 2.6ghz up to 3.0ghz if the voltage is set to its correct default ammount "Not Auto" will not increase the actual cpu temps. While yes the gates are opening and closing faster this is happening at the nanometer level. And generally speaking the actual speed increase of the gates when overclocked only a few hundred mhz is relatively small.

Instability comes into play when the chip is no longer able to function at the requested speed running on current voltage. That's when bumping the voltage is required and then and only then is when any real and measurable temperature increase will take place.

I'll have to second that remark, bumping the FSB up ~80mhz alone pushed the thermometer 10°C further up in my case.

If your voltage is set to auto in the bios its possible that when upping the FSB it automatically bumped the cpu voltage. But 80mhz isn't enough to sneeze at. And a 80mhz fsb bump would in no way raise your case temps by 10c.

1. The cpu and or motherboard northbridge voltage were left on Auto and not set specifically to their correct stock settings. -Even then 80mhz does not equal 10c case temp increase.
2. There's other circumstances that came into play. Eg your aimb temps were higher than normal, your case fan speeds are adjustable and they were left lower than normal.

In any case 80mhz does not equal a 10c case temp increase. This is how rumors get started and how new overclockers get filled with incorrect information.

Anyone actually interested in overclocking I recommend you head over to xtremesystems (http://www.xtremesystems.org/forums/forumdisplay.php?f=56) or the oc forums (http://www.overclockers.com/forums/)

Also note, when I buy cpu's @ places like newegg I buy several of the same chip in search of the best stepping or batch. After testing them all I keep the best one or two and sell the rest. Out of all the chips I've never once seen a 10c case temp increase from a 80mhz overclock. Nor have I ever witnessed any major temp increase from overclocking while using/setting in the bios the correct stock voltages. At the very most .5c to 1c and generally that's due to aimb temp changes.

If you're considering overclocking go read up on it. Its hands down the best way to get the most bang for your buck.
But I strongly suggest that you question everything you read. There's loads of info about clocking in the web but 60% of it is complete bs.

[DarkZell666]

@CpT : I do remember overriding the voltage manually to 1.25V (which is it's default), though there were other voltage settings that I did leave on auto indeed. Thanks for the heads-up, I'll read more based on your post a bit later on :)