Threaded Mode | Linear Mode

feinicks · 16/12/2008 09:25 AM

Vegetano1 Wrote:
feinicks Wrote:
Vegetano1 Wrote:
feinicks Wrote:
Vegetano1 Wrote:i am saving for a geforce 280 its prob the best graphics card there is now,. but i read the 8800gts is still very good,. i have never bought a ATI graphics card,.

GeForce 280 can be considered as the best card for gaming... but for encoding, rendering etc, even the Quadro FX series is excellent...

if you got the $$$'s!! Quadro FX would be great!

whot part does a graphics card do when rendering/encoding,. i thought you needed a good cpu for that,.!>?

a simple example would be Photoshop CS4 which uses GPU rather than CPU to add filters...

essentially, a GPU is a lot more capable of high calculations that are required in encoding/rendering..

you mean the open-gl feature in PS CS4!>?

but not many programs use gpu>!? maybe 3d programs and such,. but for encoding cpu is used,..>?

also i thought the quad core was made to prevent overheating,. so i was told,. but seems quad core cpu's reach hight temperature's also.

yeah.. not as of now.. but GPUs were to blame for that. It was not as simple to switch between GPU and CPU for a program. It would be either Mostly GPU dependent (like games) for major processing or would stick to CPU..

But with CUDA it now way simpler... expect many heavy duty programs to follow suit.

Vegetano1 · 16/12/2008 09:27 AM

feinicks Wrote:
Vegetano1 Wrote:
feinicks Wrote:
Vegetano1 Wrote:
feinicks Wrote:
Vegetano1 Wrote:i am saving for a geforce 280 its prob the best graphics card there is now,. but i read the 8800gts is still very good,. i have never bought a ATI graphics card,.

GeForce 280 can be considered as the best card for gaming... but for encoding, rendering etc, even the Quadro FX series is excellent...

if you got the $$$'s!! Quadro FX would be great!

whot part does a graphics card do when rendering/encoding,. i thought you needed a good cpu for that,.!>?

a simple example would be Photoshop CS4 which uses GPU rather than CPU to add filters...

essentially, a GPU is a lot more capable of high calculations that are required in encoding/rendering..

you mean the open-gl feature in PS CS4!>?

but not many programs use gpu>!? maybe 3d programs and such,. but for encoding cpu is used,..>?

also i thought the quad core was made to prevent overheating,. so i was told,. but seems quad core cpu's reach hight temperature's also.

yeah.. not as of now.. but GPUs were to blame for that. It was not as simple to switch between GPU and CPU for a program. It would be either Mostly GPU dependent (like games) for major processing or would stick to CPU..

But with CUDA it now way simpler... expect many heavy duty programs to follow suit.

SWEET!!

Assassinator · 16/12/2008 04:01 PM

feinicks Wrote:
Vegetano1 Wrote:
feinicks Wrote:
Vegetano1 Wrote:i am saving for a geforce 280 its prob the best graphics card there is now,. but i read the 8800gts is still very good,. i have never bought a ATI graphics card,.

GeForce 280 can be considered as the best card for gaming... but for encoding, rendering etc, even the Quadro FX series is excellent...

if you got the $$$'s!! Quadro FX would be great!

whot part does a graphics card do when rendering/encoding,. i thought you needed a good cpu for that,.!>?

a simple example would be Photoshop CS4 which uses GPU rather than CPU to add filters...

essentially, a GPU is a lot more capable of high calculations that are required in encoding/rendering..

Encoding? Nah, it's currently like 100% CPU. Encoding is pretty much just doing DCTs and other kind of transforms (so pretty much pure number crunching). Don't know how good GPUs are on these transforms.

If you include filtering as part of encoding, then there's like 1 avs filter I know of that uses the GPU. (And for that particular filter, my computer actually runs faster using the CPU version than the GPU version, because I've got a much better CPU). Someone should write an anti aliasing filter for the GPU. All the current anti aliasing filters are pretty slow, and that's something that GPUs are supposed to be good at.

The problem is that the GPU is too limited in what it can do. Some video decoders uses the GPU though.

-PHENOMENON- · 16/12/2008 04:05 PM

ZiNgA BuRgA Wrote:- MSI Radeon4670 512MB GPU [$119]

Lol. Just kidding. Nice system though.

ZiNgA BuRgA · 17/12/2008 07:17 AM

Update: seems the Intel guy was right about the OS installing the CPU drivers automatically - installed a fresh XP64 and the drivers showed up fine. Also, since I didn't strip out Windows Update from that one, I used that to grab the "HD Audio device", which apparently is an ATI device (for HDMI perhaps?), then transferred the drivers to my current install - yay, no more driver issues!
...only problem is that apparently MS disables hibernate if you have ≥4GB RAM... So I set the RAM limit for XP to be 4096MB - hibernate now works (though it's a bit slow) but I've got 3.25GB of usable RAM. Well, guess that's enough, even with my large RAMDisks...

Assassinator Wrote:x264 authors actually suggested using 1.5*cores. So try 6 threads.

That's actually more general than just x264 (many apps will recommend 1.5*cores) - but using more threads should stress the CPU more, as it has to perform more thread switching. It goes at around 40-70% now. I'll probably stick to 3 single threaded encodes - a lot more stable.

Assassinator Wrote:Actually, would you mind doing this encoding speed test I made?

Download the package » here (341KB)

Comes with x264 r1057, some batch files and an avs file.

There's practically no motion in that sample, so it's not really the best to test with, but at 1.8GHz (9x200): (multi encode seems to stick the CPU at around 45-80% (weirdly, the speed goes down over time, and so does the CPU usage of all the encodes - could be the motherboard automatically slowing down the CPU do to heating))
[Image: clipboard01le3.jpg]

^ Like, as you can see, the multi-thread encode averages slower than the single-thread, but I performed it after the single one (and the speed scaled down far more aggressively).

Assassinator Wrote:So how do I check my cpu temperature? With some program?

The motherboard usually comes along with some apps to view temperature. Since mine doesn't work, I have to boot into the BIOS to see them (by then, the CPU may have cooled a bit). There may be other apps to view temp, but I haven't checked.

Vegetano1 Wrote:also i thought the quad core was made to prevent overheating,. so i was told,. but seems quad core cpu's reach high temperature's also.

The idea of going multi-core was partly because of the "MHz barrier", but a quad core doesn't reduce heat. A quad-core 2GHz does have less power consumption (ie TDP) than an 8GHz single core, for example.

feinicks Wrote:But with CUDA it now way simpler... expect many heavy duty programs to follow suit.

Likely to take some time, especially with ATi holding a fair bit of the graphics card market still (and it has it's own version of GPGPU stuff).
CUDA is actually rather slow - it gains most of its computing power from its ability to perform mass operations at the same time - good if your apps can run in parallel, but unfortunately, the majority of computer applications are serial and it's not always easy to convert a serial problem into parallel.

Assassinator · 17/12/2008 08:41 AM

ZiNgA BuRgA Wrote:
Assassinator Wrote:Actually, would you mind doing this encoding speed test I made?

Download the package » here (341KB)

Comes with x264 r1057, some batch files and an avs file.
There's practically no motion in that sample, so it's not really the best to test with, but at 1.8GHz (9x200): (multi encode seems to stick the CPU at around 45-80% (weirdly, the speed goes down over time, and so does the CPU usage of all the encodes - could be the motherboard automatically slowing down the CPU do to heating))

As for the lack of motion, it's kinda difficult to make motion out of nothing with avisynth alone. I guess I could upload a source file for you to encode, but that's annoying to both upload and download.

As for the speed going down over time, it's probably the avs script. It's more complicated later on. On hindsight, I guess that might not be a very good idea, since some of your cpu power will be spent on avisynth instead of all on the encode. And since avisynth doesn't take advantage of multiple cores, the avisynth portion of the encode will under perform, and that might make the test slightly biased against multi-core computers. But if it were a real video source instead, depending on how it's decoded, the same problem could also exist.

If you want to know if your encode slows down due to your motherboard, just go and kill everything in the avs script except for the 1st and last lines.

ZiNgA BuRgA Wrote:

^ Like, as you can see, the multi-thread encode averages slower than the single-thread, but I performed it after the single one (and the speed scaled down far more aggressively).

Ok, after looking at your screenshots... I realized how hard I screwed up my command lines. Lol, the multi-thread and single thread batch files I gave you are identical. Both are --threads auto Facepalm

. Heh, remember to double check the stuff before uploading it. And that's probably why you got similar speeds for both.

Replace the x264 command line with this, and try again.

Code:

x264 --crf 20.0 --threads 1 --progress --output "ColorBars.mp4" "ColorBars.avs"

Anyway, here are my speeds.

Multiple threads
[Image: EncodeSpeedMulti.png]

Single thread
[Image: EncodeSpeedSingle.png]

So the difference in speeds between us is quite a bit less than I expected. (Heh, I expected you to completely donkey rape my speed). But you did underclock your cpu, so I guess it's capable of more than that. It's the core2 laptop by the way. My computer (desktop) is much slower, not just because it's slow, but also because the older cpu doesn't have as many of the more advanced instruction sets that can speed up the encoding. I'll do this on my comp and upload the results sometime tomorrow.

As for the "insane" encode, the speed slowly goes down steadily to something really poo poo. Not just avs script, I think it's partially due to how the fps is calculated. Not an instantaneous amount, some sort of average, and the start is faster because there's not as many frames before, which means less analysis. Since your screenshot is at 0.9% (at that time, the displayed fps is much faster than the average), It's kinda senseless to compare our results.

EDIT: Something extremely weird here....

Look at the x264 info displayed after the encodes, and compare yours with mine. The average quantizer and the metrics all don't match up. Not only that, they are quite very different... and wee're using the same settings and source...

I can imagine that using a different amount of threads might cause a slightly different result, as shown in my 2 screenshots, but the difference is way to large to be accounted for just by that...

ZiNgA BuRgA · 19/12/2008 03:55 AM

Well, I thought my CPU died earlier on. Turns out that the CMOS reset button doesn't do everything, and had to short the CMOS reset pins.
Now CPU gets far too hot - Ima try getting some thermal grease to see if it fixes that >_>

ZiNgA BuRgA · 22/12/2008 07:04 AM

Well, got myself a Xigmatech HDT-S1283 cooler today.  Gosh, that thing's a huge piece of metal (though to enthusiasts out there, it's probably nothing) - never seen anything that big, and surprised the motherboard holds it up on its side.
As usual, installing the CPU cooler is one of the hardest things, and this thing took me (literally) 2 hours to finally push in - you push one side in, and the other pops out >_>  Got a few nice cuts too. (or it's just me who sucks at installing coolers)

Also bought some Arctic Silver 5 thermal grease, but turns out the cooler had its own, so that's a waste of $9 there...

Performance wise, this thing isn't bad - Q6600 at stock, idles at 35-40C.  Under load (three instances of x264 running for about an hour) temps across the cores are around 50-60C, which still appears to be higher than what others are getting for the Q6600 on load (around 40C I've heard?).  And today isn't particularly hot here in Brissy.
Oh well, I don't plan to OC, though I expected some better results >_>  At least it doesn't get as loud as the Athlon X2 S939 fan running under load (the fan can actually go up to like 7000RPM :o ).

Whilst I was there, also got myself a Logitech Desktop 350 Keyboard/Mouse - the first keyboard I've had which actually has any extra keys (only Media, Mail, Web, Calc on the left, and Play/Pause, Vol-, Vol+, Mute on the right).  I'm hardly using those keys though - guess I'm too used to not using them...

Also discovered that the graphics card fan makes the most noise in my PC.  Might try downclocking the GPU and disconnecting the fan to see if it gets hot (I don't really use it much anyway).

@Assassinator:
Old Athlon X2 4200+ @ 2.2GHz:
[Image: 66ydlc.jpg]

Current Q6600 @ 2.4GHz (stock)
[Image: 2j65weo.jpg]

Yeah, seems not very much of an improvement. But I did happen to use DVDShrink today, and did notice quite a bit of a gain.
On old PC, analyse took 10mins, encode took 30mins. New comp, analyse = 3mins, encode = 10mins. DVDShrink does use multi-threading, which explains part of the difference, but it's still an impressive gain.

Vegetano1 · 22/12/2008 07:30 AM

holy shoot!! that's a cooler> !O_O!

and that's why i tried to find a cooler with backplate and screws,. ;p those push-in-turn-cut installments are crappy! ;p

feinicks · 22/12/2008 07:37 AM

ZiNgA BuRgA Wrote:Well, got myself a Xigmatech HDT-S1283 cooler today. Gosh, that thing's a huge piece of metal (though to enthusiasts out there, it's probably nothing) - never seen anything that big, and surprised the motherboard holds it up on its side.
As usual, installing the CPU cooler is one of the hardest things, and this thing took me (literally) 2 hours to finally push in - you push one side in, and the other pops out >_> Got a few nice cuts too. (or it's just me who sucks at installing coolers)

hmm... I saw a rig on the net (for some enthusiast meet or something..) that has a pipe system hooked to it. The pipes go around the processor, the RAM, the Northbridge, southbridge, GPU and along the mobo. The Piping is connected to a liquid Nitrogen tank. So yeah.. yours is not the biggest cooler..

an example... http://img383.imageshack.us/img383/628/pc2ke3.png

ZiNgA BuRgA Wrote:Also bought some Arctic Silver 5 thermal grease, but turns out the cooler had its own, so that's a waste of $9 there...

lol.. didn't it say on the box under Contents?

ZiNgA BuRgA Wrote:Performance wise, this thing isn't bad - Q6600 at stock, idles at 35-40C. Under load (three instances of x264 running for about an hour) temps across the cores are around 50-60C, which still appears to be higher than what others are getting for the Q6600 on load (around 40C I've heard?). And today isn't particularly hot here in Brissy.

AMD chips are supposed to run cooler than Intel... I think. I have a thread somewhere here about that...

ZiNgA BuRgA Wrote:Whilst I was there, also got myself a Logitech Desktop 350 Keyboard/Mouse - the first keyboard I've had which actually has any extra keys (only Media, Mail, Web, Calc on the left, and Play/Pause, Vol-, Vol+, Mute on the right). I'm hardly using those keys though - guess I'm too used to not using them...

lol.. my keyboard has some extra buttons but it needs drivers to be installed... anyways, I use combinations shortcuts or "Run" to launch most direct programs like calc or word etc... may get a Razor presently though...

ZiNgA BuRgA Wrote:Also discovered that the graphics card fan makes the most noise in my PC. Might try downclocking the GPU and disconnecting the fan to see if it gets hot (I don't really use it much anyway).

Yeah it will... even casual usage of the graphics card heats it.

Threaded Mode \| Linear Mode New Computer
Author	Message