Athlon64 X2 encoding performance

[Sharktooth]

Anandtech reviewed the A64 X2 and this puppy seems to crunch everything you throw at it...

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=6

[Sirber]

sweet!

[Doom9]

I've been waiting all that time for a 2.6 GHz Athlon64 4200+, now there won't be one so I'm torn between waiting for the X2 or getting the Venice core 3800+. Best thing would be to have two identical PCs though, one with a single core and one with a corresponding dual core so that in the next codec comparison, I can compare single to dual core.

[Angelus]

Extremetech had similar results with the new AMD processor, here's a link to their review of it too:

http://www.extremetech.com/article2/0,1558,1813763,00.asp

they seem amazed by it as well. It seems like my next computer build is going to have to be AMD :D

[b00zed]

In the words of Keanu Reeves, "woah."

I was a bit up in the air as to whether my next media PC would be a P4 or an A64... and now it's settled. The deficit traditionally suffered by AMD's CPUs in audio and video has completely evaporated, the P4 can't touch it at all in most other tasks, and according to THG the X2's power consumption at full load is lower than any current *idling* P4, which is very attractive. Those multitasking tests look very promising too. The only shame is that we have to wait until the second half of the year for these things to show up in volume.

looks like it will soon be time to finally say goodbye to my faithful Athlon XP 2000+ I think :D

[Doom9]

The deficit traditionally suffered by AMD's CPUs in audio and video has completely evaporatedWas there ever one? XviD has been consistently faster on the AMD platform than on Intel. DivX is the other way around.. considering how much money Intel invests to get people to optimize software for their CPUs, that's not a surprise.. generally unless a video software is heavily optimized for the Intel platform, running it on a comparable AMD chip should give comparable results, and depending on the implementation, the AMD should even be faster.

It is to be noted that starting with the Venice core AMD also supports SSE3 so it can take advantage of the "Intel optimizations" unlike previous cores. Hopefully, not too many software will check the CPU type but just query the CPU's capabilities and activate/deactivate optimizations accordingly.

[b00zed]

I had a feeling I'd get pulled up for that. Unfortunately XviD does not represent all there is to audio and video production.

[Sirber]

Originally posted by b00zed
looks like it will soon be time to finally say goodbye to my faithful Athlon XP 2000+ I think :D
You could build a linux server from it :)

[Sharktooth]

http://www.techreport.com/reviews/2005q2/pentiumd-820/index.x

Watch how pentium 4s single and dual cores get pulverized...

[Doom9]

Finally they've shown in in the PC store where I shop. Supposed to be available this week even.. even with some delay, just in time to replace my NForce3 barebone with an NForce4 one and an X2.. the old box will make a nice birthday present for my dad and his reorganized home office (currently the big almost-full-tower) looks really ugly and doesn't go well with the 26" flatscreen TV/PC monitor.

[BangoO]

Hi there,

Forgive me if this is a stupid question, but I need to be sure before spending my $$$ :)

I read that Windows Media Encoder 9 is multithreaded.
I am encoding stuff in WM9, but with VDubMod.
I read that, though VDubMod is multithreaded, its encoding part is not...
So... should I buy an X2 cpu, and will I experience the same improvement as in the benchmarks which use Windows Media Encoder 9 ?

Thx ! ;)

PS: I currently have a P4C running at 3.15Ghz...

[tsp]

yes you will. It only depends on the encoder.

[BangoO]

Thx for the fast answer tsp, now I know what my next cpu will be :D

[Doom9]

This post is made from a dual core machine... X2 4600 :)

[Bootsy]

Hope we will see some benchmarks soon. I'm tempted to buy one myself but I want the larger cache that the 4400 and 4800 offers.

[BangoO]

This post is made from a dual core machine... X2 4600 :)
Nice !
Are you happy with it ? ;)

[Doom9]

Are you happy with it ?I haven't gotten around to transferring all the files so I haven't really tested anything.. for now I just know that my wallet is very unhappy. The NForce 4 based XPC is a bit quieter than the NForce3 based one though, so I'm happy about that.. I'm also happy that I didn't need to put in a floppy to set up Windows.. the last intel machine I got required a floppy (a real one, usb didn't help) to get the ICHR6 driver even though I just had a single regular SATA disk. I may get around to racing the two CPUs (A64 3500+ versus the new one), to see how much an additional 200 Mhz and an additional core will fare.

[Doom9]

I've finally switched the two machines.. too bad my stereo can only handle 5.1, not 7.1. I've had one blue screen while transferring dozens of GBs of raw DVD data I use for testing purposes.. and some other weird stuff.. now using this full time will show how good the whole thing really is (Barebones, while great from a visual and space perspective tend to be a lot less unforgiving when it comes to stability, and I packed a lot of high end stuff into this box.. my last one had to undergo an MB and RAM change to run stable). Anyway, I'm now about to install AviSynth, get my D2Vs working again and start a quick encoding test.

[Doom9]

Alright, here's the first report from my X2 4600. x264 revision 270, cli version.

3598 frame PAL DVD source, scaled to 640x352. Settings used: ABR, rdo turned on, 6, 3 ref frames, 2 b-frames, adaptive b-frames and pyramid order turned on, as well as weighted prediction and adaptive dct and 8x8 i-blocks (so high profile).

Naturally, I've used MeGUI 0.2.0.4 for all the tests as it displays FPS and lenght accurately (I'll have to improve upon the FPS as right now it shows the last FPS value received.. I'll make it the average in the next version).

Athlon 64 X2 4600 (2 cores @ 2.4 GHz, 512KB cache each)

1 threads: 6m29s = 9.24 fps, relative speed: 112.23%
2 threads: 4m06s = 14.62, fps relative speed: 177.64%, relative speed gain over 1 thread: 58.22%

Athlon 64 3500+ (1 core @ 2.2 GHz, 512 KB cache, Manchester core)

1 threads: 7m17 = 8.23 fps, relative speed 100%
2 threads: same as with 1 thread.

So, in terms of the PR rating and just looking at x264 performance does the processor deserve its rating? If we consider that most applications are not parallelized to make use of dual core CPUs, then I'd have to say no. Is the money still well spent? I guess that depends on how frequently you're multitasking and which applications you're using.

In my case, using 1 thread my overall CPU load is 51% (x264.exe running in idle mode), so there's plenty of room to do even I/O intensive stuff. The CPU load was consistently above 80% using 2 threads on the dual core CPU. Except for MeGUI running x264.exe, I also had a browser, windows explorer and winamp running (with winamp playing an mp3 online radio @ 128kbit/s)

On a sidenote for those not familiar with AMD's ratings.. the 2.4 GHz single core chip with 512KB cache has a 3800+ rating.

[Blue_MiSfit]

as the script kiddies say... w00t 4nd r0x0r :) I need to get a dual core, but last december I spent a bunch getting a socket 754 board and a 2800 (big mistake)... So I wont be upgrading for at least a year now...

hindsight is always 20/20 :mad:

[tonygib]

Out of interest Doom, I assume thats all using 32 bit code, ie both the OS and the encoder/codec?

[Doom9]

hindsight is always 20/20A year is a long time in the computer world. I got my last box about a year ago and I do not recall even having heard of dual core chips then. But I knew the socket 939 was more futureproof so I spent the extra cash.. basically I could've just upgraded my SN95, but I also wanted PCI Express, 7.1 HD audio, the additional USB ports and lower noise of the SN25P.

And all that is using the regular 32 bit version of Windows. There is not even a 64 bit version of x264 as far as I can recall. I guess I could go hunting for drivers and if everything is available, give the 64 bit version a shot.

[tomos]

any idea of encoding speed for divx6 @ 1080 res using telecide and decimate? thats the reason i'm getting one of these puppies :) my 2500+ sadly isn't up to the job :(

[Doom9]

Only DivX Helium contains SMP optimizations ;)

And we Euros don't even have 1080 sources to work with (well, extremely few and I'd need to get a PCI-E DVB-S card to capture something since the SN25P is PCI-E only (not that I'd want it any other way.. I wished DVD burner makers would release more SATA drives so I could disable IDE entirely and use only the SATA-II ports)).

I've been studying the Anandtech benchmarks and made a startling discovery: the dual core tests, the FX-57 tests.. they don't all use the same media encoding benchmarks or use the same sources.. that is extremely deplorable as only using the exact same scenario really allows you to properly compare CPUs.

[sandman1687]

doom, are you happy with the SN25P? I've heard mixed reviews, and was thinking about getting one for my next system.

[BoNz1]

And all that is using the regular 32 bit version of Windows. There is not even a 64 bit version of x264 as far as I can recall. I guess I could go hunting for drivers and if everything is available, give the 64 bit version a shot.

Hi, there is a 64bit version of x264. However, I think it is only useable with linux due to the same issues as the xvid 64bit port. Although, I would be very interested in seeing how fast it is using 64bit code and multithreading :p.

[Leo 69]

@ Doom9

Have you tried overclocking your X2 ? I'm just curious how far it can go... I'm sure it may work absolutely steadily at 5000+ (and even beyond) without increasing the Vcore. Wouldn't it be nice? :)

[tomos]

i think you can get the core's on a 4400 to 2.6-8 on air fairly easily. not sure about the 4600+

[Revgen]

I'm going to buy a 4800+ when and if MSI finally releases a BIOS update for my board. I wonder if the 1mb of cache on the 4800+ makes a difference or not. I think it will when it comes to encoding.

[Doom9]

I have my doubts about the cache size.. so I saved the money and got for the smaller cache size.. cache sometimes helps but often it has no effect. I haven't tried overclocking.. it may reach 2.6 GHz easily but I've had 3 crashes so far (blue screen moving files over, freezes when burning DVDs), and until I'm convinced that the system is 100% rock stable I wouldn't even dream changing anything to non standard settings.

As far as the barebone goes, I consider it an improvement over the SN95, but we're definitely not there yet noise-wise.. I hope we'll get to a Dell 8400 level eventually.

[tomos]

I'm going to buy a 4800+ when and if MSI finally releases a BIOS update for my board. I wonder if the 1mb of cache on the 4800+ makes a difference or not. I think it will when it comes to encoding.

there was one iirc - at least for the neo4 platinum. here (http://www.msi.com.tw/program/support/bios/bos/spt_bos_detail.php?UID=637&kind=1)

Update CPU ID for DualCore DT 2.2G & 2.4G

[Doom9]

bah, that PC will go back to the shop.. 3rd freeze today, and the machine was under no particular stress.. not unlike when I got the last barebone, but there the problems ocurred more frequently. Either it's those high speed RAM modules that I also used last time (and that after making me wait for 4 weeks they replaced along with the MB), or the MBs are really prone to error :(

[mrbass]

download ultimate boot cd (it's free)
run memtest of course
and a couple of CPU stress testers
cpu-burn in (http://mrbass.org/ubcd/01mainboardtools1of2/02cpuburnin.png)
lucifer (http://mrbass.org/ubcd/01mainboardtools1of2/01lucifer.png)
Mersenne Prime Test (http://mrbass.org/ubcd/01mainboardtools1of2/03mersenneprimetest.png)

[Doom9]

I'm afraid I'll have to reconnect audio and use my "old" PC again tomorrow to run some tests.. not that I'd really have to.. I paid 100 bucks to have my PC assembled just so that in the error case I can throw it back in their face.. but it's always good knowing what's going on. Though the last time I ran memtest for hours upon hours, and switched out every component but the mainboard and CPU.. it was those last two parts. Since they changed RAM and MB last time and I have the exact same effects.. it might just be the MB yet again.. and my confidence in barebones is dropping again :(

[Joe Fenton]

There is not even a 64 bit version of x264 as far as I can recall. I guess I could go hunting for drivers and if everything is available, give the 64 bit version a shot.

I don't know about Windows, but I'm running 64bit x264 in 64bit linux. That threads optimization was awesome! I've got a dual opteron 240 system, so running on one CPU isn't too terribly fast compared to newer AMD64 systems. Getting to use that second 240 really helps cut the encode time considerably... or I can use parameters that make the encoding slower but better looking without losing as much speed.

[sandman1687]

@doom9
i went onto shuttles website (i was looking to buy the SN25P), but nowhere did it mention support for dual cores, nor did it have any bios updates. either it already handles it, or poorly tries to. now that i've researched it and all, the lack of a PCI slot (wtf am i going to put into the PCIE 1x slot), compounded with your experiences with your system, which has very close specs to one i was going to put together, im holding off on the SN25P, or at least going to look into other SFF brands. thanks for your Dual Core stories.
-sandman1687

[Doom9]

(wtf am i going to put into the PCIE 1x slot),I liked it for that.. get rid of the old thing as fast as possible. I haven't used serial, parallel and floppy in ages (first thing I do when I get a new system is go into the BIOS and deactivate that crap). I'm looking forward to dumping IDE as well and move on.

As far as BIOS support, I think you missed this: http://global.shuttle.com/ADVNews/NewsDetail.asp?CTID={A69AC49F-B90C-445E-A412-560BA5F2A03D}

The problems I'm having have nothing to do with dual core.. they're exactly the same I've had with the last barebone, which used the same kind of RAM. Last time they asked about my RAM timing.. (I built it myself them), this time they should've taken are of that. If it's RAM again I'll go back to regular stuff and forget about the high speed Kingston modules.

[Doom9]

well, I guess normally red is a bad sign, but memtest just came up with a load of red lines, so that's it.. faulty memory (again). And I figured the 100 bucks for assembly included a stress test which would've certainly prevented me from getting faulty components yet again.

@markrb: thanks for the tip on that CD.. it is extremely useful.

[Revgen]

there was one iirc - at least for the neo4 platinum. here (http://www.msi.com.tw/program/support/bios/bos/spt_bos_detail.php?UID=637&kind=1)

I have the K8N Neo2 Platinum.

I still appreciate the help. Thanks. :)

[tomos]

ah sorry about that.

how do you find MSI boards anyway? never tried them before

[Revgen]

ah sorry about that.

how do you find MSI boards anyway? never tried them before

If you live in the USA you can get them here.

http://www.newegg.com/Product/ProductList.asp?Submit=Browse&Manufactory=1312&Subcategory=22

[xyretp]

I recently had the opportunity to benchmark a mathematical analysis using both the IBM(dual-p4 w/4G) & Sun (dual-opteron w4G) memory. Both workstations were running XP/SP2. The app

The SUN XP (AMD opteron version) ran the benchmark 110% faster. This seems to be in line with other floating-point intensive results (16 v 34 hrs)

[Joe Fenton]

I got an MSI Master2-FAR dual-opteron mobo from NewEgg last year. That's my main system. I got mine with really cheap memory. I wound up having to raise the memory voltage a little (2.65V) to get it to work right. You might give that a try, Doom9. It's better than backing off the timing.

Other than the memory issue, I've had no complaints about the MSI board. It uses the VIA K8T800 chipset.

[Rash]

Doom9, what's your chipset?

[Doom9]

nForce4. Don't bother with suggestions though.. the PC is back at the shop.. I paid them to assemble and test it so it's their problem. If they haven't set up the RAM properly (it's not cheap, it's high end, Kingston HyperX) it's up to them to fix it. My trusty old 3500 works just fine until I get the other one back.

[puffpio]

In the Anandtech tests..the kernel compilation was the fastest while doing 3 jobs even though it's a dual processor..I was wondering if you could try your encoding test again but w/ 3 threads going on and see what kind of speed you get

[Doom9]

but w/ 3 threads going on and see what kind of speed you getNot while the PC is at the shop obviously ;) Hopefully I'll have it back by the week-end.

[ReTox]

There are going to be tonnes of wares coming out for dual-core. I'm heavy in the IT sector and most server apps already have a dual-core version either available or on the drawing board. As always it won't take long for the desktop user to see X2 apps.

Can't wait to see Dual-core 64 bit apps on a 64 bit OS, should smoke pretty good!

[luders]

You all must think I am crazy but I love my Pentium D. Good price in comparison.

[dvd_maniac]

You all must think I am crazy but I love my Pentium D. Good price in comparison.

I second that!

[Doom9]

everybody picked his cpu for a reason and naturally all those reasons are valid.. it's all a matter of choice and preference ;) Feel free to open another thread where you compare the performance of your Intel CPU with existing single core P4s and the effect of using more/less threads for x264 or another encoder that supports smp (libavcodec also offers threads, so you could use that, or the divx helium beta also supports smp).

[ELNINO]

If you get it back and you still have errors. post your problem over at overclockers.com and you should be able to solve your problem or get some advice on what timings will get rid of your memtest errors.

which test gave the errors?

which mobo/ram/powersupply is very important to newer chips?

[Rash]

nForce4. Don't bother with suggestions though..
I was going to say you are one more facing the driver problem of the NForce 4. See this (http://forums.nvidia.com/index.php?showforum=34).

I also have Nforce 4 and I just found out there is nothing wrong with the hardware. The problem is in the drivers. So your computer may not come completely fixed from the shop unless you change the chipset. ;)

Install NVIDIA NForce4 Standalone drivers, enable the hardware firewall and it is BSODs everywhere. :( Now I am using WinXP SP2 firewall and Microsoft IDE Drivers and it is rock solid.

[Doom9]

Guess what.. they called me an hour ago: faulty RAM. Not that I am surprised.

@Rash: drivers has zero influence on a test executed in freedos ;) If any of the UBCD tests show problems it's most likely a hardware issue.

[Rash]

Not to all Nforce 4 users. :) Can't believe that by just uninstalling the drivers the BSODs go away and it is still hardware fault. That's not just one person's case, but everybody.

Seriously, I don't know the problems you were experiencing, so maybe they weren't the same as the people on that board. ;)

[Doom9]

drivers can cause system crashes, they have access to a lot of potentially critical resources.. if they're not installed, they cannot. If you don't boot an OS that doesn't have them, they cannot cause crashes either.

but everybody.Definitely not everybody. There never was anything like that on nforce 4 board reviews, reviews of my barebone, or even on my machine.

[Doom9]

almost 5 passes done in memtest and not a single error. This week-end might just prove to be crashfree :)

[Doom9]

6 passes without error. I'm back on the new PC now. Noisewise, the upgrade was also a great choice.

[netterpaladin]

Hi,
I am also interested in the Shuttle SN25P. My question is which graphic card would you put in. I am sure most of you are also interested in having a TV-card in there. Since there are so few slots what do we do?
1) PCIe Card e.g. gforce and an additional TV card
2) ATI all in wonder card. e.g. Ati allin wonder X800.

Problems to consider are:
A) noise of the card, prefereably passivly cooled
B) has to fit into barebone
C) price

regards,
netterpaladin

[Doom9]

:( the pc is going back to the shop again. I cannot believe I paid them over 100 bucks to assemble and test that pos.. it froze downloading patches from windowsupdate and the first benchmark tool I install on a freshly installed Windows with nothing but all patches and drivers consistently crashes the system.

I'm not sure there is a PCX TV card at this point.. I haven't heard of any. They are certainly coming in the future though. But this kinda limits your choices.

[puffpio]

What is the power supply rated for? I could see how memtest would run flawlessly because you aren't stressing the power supply by not running Windows and downloading stuff thus using the network card etc etc.....

[Doom9]

puffpio: I have assembled PCs for some 13 years know.. I wouldn't miss something as basic as that. Any not anywhere in my error description would you find any indication that the Power Supply is at fault. In fact I have 7 hours of Burn in Test with all the options enabled to show for.. that does rule out a lot. And since nobody from the shop is reading this.. it's kinda pointless wouldn't you agree? I'll ask for help when I think I need it ;)

[netterpaladin]

In the C´t (german computer magazine) from last week (edition 14), there is a review of 10 mini-barebone incl. 3 from shuttle. Maybe you should take a look at those. From the 3 shuttle barebones, 2 are for P4, 1 for AMD XPC ST20G5. I am also interested in the combination of a barebone with fast CPU, otherwise I would take the Imac mini!!!! The nicest SMALL PC.

I wonder where your System instabilities come from. Whether it is a general problem, or just because of all the high performance parts.

Please keep us informed. I am interested if one can get this system to work

[Doom9]

Please keep us informed.I will.. too bad I can't just go to the shop one evening, pick the components I want and throw all my tests at it.. I'm sure I'd find the culprit in no time. I suggested to them that the first thing they try it throw out the Kingston RAM and use regular RAM, then proceed to GFX card (BIT reported video memory corruption), and finally the barebone itself (I'm afraid barebones have a higher tendency to faulty MBs.. this is the third barebone I have, and the last two required at least one MB change). I suppose by the end of the week I'll know more.

[Doom9]

hmm.. I called them today. They've tried another barebone, different RAM (even different brand), different CPU (single core) and that didn't yield anything yet (the Video Memory Corruption errors seem to have gone away somewhere during the process though). Now it's down to harddisk, GFX card and DVD burner (though the latter really shouldn't have any influence). It is a rather puzzling thing.. definitely not your 08/15 issue.

[puffpio]

hey doom9 I found the source of your problems
while browsing the sudhian sff shuttle xpc forum, there is a known issue w/ the SN25, dual core athlon's and usb2

if you set the bios to use usb 1.1 then the lockups go away...alternatively if you keep usb 2.0 settings and force windows to use single core..but i like the other altenative better

the shuttle guys are aware of the problem and 2 BIOS engineers are working on it

[Doom9]

@puffpio: could you give me a link please?

though they already tried using a single core chip (athlon 3500+) and got the same freezes in PC Mark.

[Ice =A=]

Well, I'm somewhat lost, could you clear this a little up for me?!
Some wrote that only the codec can be optimized for multi core systems, however for example the links on the first page point to some tests where there was a lot of gain with standard DivX5 and Xvid, both of which don't support multithreading as far as I know. Can anyone give an overview of different ripping/encoding tools and their gains from dualcore cpus?
Like Xmpeg, AutoGK, Vidomi (about 60% I think) and so on?
That would be great! :)

[Doom9]

The X2s are faster than their singlecore counterparts, SMP optimized apps or not. This is because Windows doesn't put all the load on the same CPU, it tries to balance.. so if you have a very CPU intensive task, it may run on core A, and the rest may run on core B. The more apps you have running concurrently the more you'll feel that. Now let's take Xvid encoding as an example.. there's different components involved here.. VirtualDub with the codec may run on one core, and the MPEG-2 decoding (which takes, I guess about 20% of the whole computing power used in the process) may run on another core. So that's why you'll generally see an improvement.

Of course, an optional distribution is only possible if all those components have multiple threads that can be balanced.. that is what SMP optimization is basically about.

Xmpeg, AutoGK, Vidomi, VirtualDub, Gordian Knot, and whatnot are all NOT SMP optimized. In case of Xmpeg (who uses that... will get a free beating from me ;), AutoGK, VirtualDub, Gordian Knot all rely on VfW codecs.. that's where the heavy lifting is done. XviD isn't multithreaded.. so you get whatever speed gain you get from because of what I explained above. DivX, and x264 are SMP optimized, so they offer more speed gain.

Mind you, none of the applications you listed does the encoding.. a codec does it. And the decoders they use (you first have to decode a DVD so that you can re-encode it to another format) is also done by mechanisms that are single threaded. The only decoder I'm aware off that seems to contain some SMP optimization is DVDx. But since it's the encoding that takes most of the cpu cycles, SMP optimized codecs are much more important than SMP optimized decoders.

[puffpio]

http://forums.sudhian.com/messageview.aspx?catid=43&threadid=80305&enterthread=y

I've also read about problems with the nforce 4 IDE driver and the latest driver is supposed to stop data corruption

[Ice =A=]

@Doom9:
Well, just recenty someone gave me a little insight into such issues and it seems that it's indeed not so easy with multithreading. :)
On one hand, there are encoding programs which do have a boost in encoding speed (about 15%, which isn't too bad actually) with dualcores, even when using non optimized codecs. (Like Vidomi, which I will however have to test a little further to be sure).
On the other hand it seems to be more difficult than I thought to split a program in threads so the syncronizing doesn't eat up the benefit...
And nearly all the (re-)encoding programs I tested didn't improve at all using two cores!

[Devinator]

TMPGEnc and CCE are supposed to be multithreaded... Anyone using them with an X2? I am going to upgrade to a socket 939 system soon. But I'm uncertain as to whether I want to get a cheaper single core processor and wait a while before going dual core. Soon, AMD will ONLY be making dual core processors(except for the sempron) so you know that it will not be long before all apps/games are multithreaded/optimized for dual core processors. Clearly it will be standard within a few years.

[luders]

I havent used CCE since I got my dual core but TMPGEnc is sure multi threaded and is so much faster.

[Ice =A=]

How much faster?! :)

[luders]

Double speed pretty much on the nose.

[Revgen]

http://forums.sudhian.com/messageview.aspx?catid=43&threadid=80305&enterthread=y

I've also read about problems with the nforce 4 IDE driver and the latest driver is supposed to stop data corruption

In my experience, Nvidia's optimized IDE drivers can cause problems if you are burning CD's or copying files from drives.

I've never had issues when they would affect encoding or decoding.

But then again, I use an Nforce 3 chipset instead of an Nforce4. They might be different.

[Devinator]

Double speed pretty much on the nose.


Nice... Expensive but ever so tempting.

[Ice =A=]

Double speed?!? Well, how does it do that?! Is the codec itself optimized or does TMPGEnc run two instances in parallel?

[luders]

Well if I set the affinity of TMPGEnc.exe to one core then it uses 100 percent of that core. If I set leave the affinity on both it uses 100 percent of both cpus and the time gets cut in half.

[luders]

Just did a test with CCE 2.5 and it uses 70 percent of both CPU's. I dont know why it doesnt max them out but at 3 times real time Im not complainin. It is nice in Windows XP x64 because you dont have to run any compatibility modes for CCE 2.5 to load avs scripts.

[Ice =A=]

Unfortunately using more then 50% average on both CPUs dosn't mean it has to be faster at all (as I had to see recently)! All the additional power at least can be eaten up by syncronizing and God knows what...
You really would have to compare the encoding times!
For this, there is a switch /onecpu in the boot.ini (for example).

[luders]

Yeah I compared encoding times. You dont need to switch the boot.ini you can just change affinity via taskmanager to test times.

[Devinator]

Just did a test with CCE 2.5 and it uses 70 percent of both CPU's. I dont know why it doesnt max them out but at 3 times real time Im not complainin. It is nice in Windows XP x64 because you dont have to run any compatibility modes for CCE 2.5 to load avs scripts.


not bad... 3x realtime is that with any frameserving/resizing etc?

[luders]

Here is the script that is running.

v=mpeg2source("becool.d2v")
a=wavsource("becool T01 3_2ch 448Kbps 48KHz.wav")
audiodub(v,a)
LanczosResize(704,480)
ConvertToYUY2()

[Ice =A=]

Yeah I compared encoding times. You dont need to switch the boot.ini you can just change affinity via taskmanager to test times
Well, that should almost do it, but not quite. If you have some programs running in the background (firewall etc.), they will be executed by the second core allone (and the encoder has the other one for itself), whereas if you switch off the second core completely... Well, you get the idea! :)

[luders]

Yeah I do get the idea. I just dont know what your point is. There is a reason that Intel and AMD's roadmaps are going toward multiple cores. If an encoder is SMP capable, it is going to be faster. Windows doesnt load balance and intensive SMP process to one processor because a few low load processes like firewalls are running.

[Ice =A=]

My point is, that if you force a program on a dualcore system to just one core it at least could still be faster than on a singlecore system, since other small background programs won't run on the same processor.
So I'm actually arguing pro dualcore here! :)

So if you want to compare a program on a system with one or two cores, you still might want to really switch the second core off in the boot.ini (to simulate a REAL singlecore system).

[luders]

Yeah thats true. Thats the beauty of multi processor in general for sure. To me it doesnt really matter if a program is multi-threaded since windows just has more to work with.

Guess I feel I have the comparison down since my last computer was a 2.8 P4 :p

[tsp]

luders have you tried my version of avisynth here (http://forum.doom9.org/showthread.php?p=682790#post682790) (just remember to backup the old avisynth.dll). It should enable you to use the processors more allthough it is not completely stable yet.

[easyfab]

I have a question about the AMD X2.
Ok it can give better result in encoding but what's about decoding ??
Is a dual-core better in xvid,wmv9,AVC decoding?? Are the decoder multithreads??
For example my actual XP 3200+ (2200Mhz) can't render AVC HD (1080p) files. and i think also a A64 4000+ (2600mhz) couldn't.

But what's about X2 3800 (2 * 2000 = 4000 MHZ) or X2 4200 (2 * 2200 = 4400 MHZ)?

Can a x2 owners answer my questions, thanks

[Ice =A=]

Athlon X2s often are NOT faster with ENcoding (depending on the codec or encoding program you use).

Concerning DEcoding, at least WindowsMediaPlayer uses more than one core! I have an Athlon X2 4200+ and it can decode even the most demanding HDTV videos smoothly. For example StepIntoLiquid needs about 60% processor capacity and if I switch off one core it is visibly slower compared both cores (some judder all few seconds).
However I have other video players which only seem to use one core.

Concerning especially Apples Quicktime7: Also that player seems to use both cores, it is however far from being optimized for PCs (it's still a Beta version), and while it plays almost smoothly there is still some room for improvement. Indeed, those QT7 1080p videos are the only ones which do not play perfectly on my rig...
Also, there are other video players who allready play those QT7 cllips much faster then QT7 itself

[Doom9]

the latest DivX beta is SMP optimized, so is the new nero avc encoder (currently in private beta test), x264, libavcodec. Check a few hardware sites for results with non optimized codecs (anything up to official divx6 release, xvid).

[Revgen]

Update to my previous post: (http://forum.doom9.org/showthread.php?p=681349#post681349)

MSI finally released a BIOS supporting Dual-Core. I should be able to do testing in the next week or so.

I've since decided to get a 4600+ instead of a 4800+ per Doom9's suggestion and this very informative review. (http://www.gamepc.com/labs/view_content.asp?id=a64x2&page=1)

I also remember that somebody out there is/was developing a multi-proc program or plugin and was looking for beta testers to try it out. I'm game once I get everything working.

[Revgen]

Okay, I just installed my new Athlon 64 X2 4600+ and decided to run a test with X264(VFW).

I took a 40 sec. scene out of a 3 Stooges short that I had with a lot of moving foliage and some action.

My settings:

2000kbps
High Profile
Partition decision quality=6-RDO
ME Method=Exhaustive Search
ME Range=24

All Other Settings=default

Results:

One thread= 0.91FPS

Two threads= 1.41FPS

Overall it gave about a 56% improvement in speed.

Not bad at all. :)

[Doom9]

ME Method=Exhaustive Search
ME Range=24wasn't it aku that showed that exhaustive is a waste of time and can even degrade the results? and iirc an me range of 16 is more than sufficient. going to extremes mostly gives people a wrong impression about the performance of a codec while not really delivering on the quality end. RDO and the 8x8i transform are worth activating, but stick to the me defaults unless you want to encode at low speeds.

[Revgen]

wasn't it aku that showed that exhaustive is a waste of time and can even degrade the results? and iirc an me range of 16 is more than sufficient. going to extremes mostly gives people a wrong impression about the performance of a codec while not really delivering on the quality end. RDO and the 8x8i transform are worth activating, but stick to the me defaults unless you want to encode at low speeds.

I'm just trying to push the processor by putting it through a ridiculous workout. I wouldn't use these settings for a real project. :D

[Doom9]

it'll get the same workout with reasonable settings.. but you get more frames done in the same time ;) If you want ridiculously low fps at reasonable settings, you can always try dirac.. you'll wish for a whole farm of X2 CPUs to bring that up to reasonable speeds.

[popper]

it was interesting reading this old thread doom.

on the subject of farming out the encoding (AVC in this case)

why is it so hard to find a simple windows farming app?.

given that many readers here have more than one machine on a network, it suprises me that noone seems to have yet made a simple self contained windows app, that once installed, auto announces on the network 'im here and waiting for an encode job', that for instance x264 or a GUI front end can then see and use these networked windows boxs sitting almost idel?.

have i been looking in the wrong place and these community apps do exist somewere?, please enlighten me!.

from a hardware POV this low power KiloCORE looks interesting, i wonder if someone will try and put this in a USB2 video encoder box and port x264 or any no. of encoders in to its re-programable FPGA?
http://www.rapportincorporated.com/kilocore/kilocore.html
http://www.rapportincorporated.com/kilocore/kc256.html
http://www.power.org/forums/viewtopic.php?t=90&highlight=kilocore

[foxyshadis]

Look up ELDER. Not quite complete, but entirely functional (a little more manual work is all), and the perl source is available to finish if you wish.