Conroe and X264 (64 bit)

[ndkamal]

I would like to show you one of the first benchmark of x264 (64 bit) with the new Intel processor.

http://www.hardware.info/nl-NL/articles/am5nZWppZA/Intel_Core_2_Duo__Core_2_Extreme_Test/6

As we can see the Core2Duo is better than Athlon64.

[GodofaGap]

Does x264 VFW support multithreading? I guess that is what they used?

[n3r0]

De huidige versie van deze H.264 encoder is helaas nog niet geoptimaliseerd voor dual-core processorsi think this means that "this h.264 encoder isn't optimized to take advantage of dual-core processors yet"
and the framerates are almost the same with my e6600

[GodofaGap]

Well that is wrong, because x264 cli supports multithreading for more than a year now I think.

[Sirber]

wep, x264 can use SMP with no problem (--threads 2/4/6/8/n)

can't wait to get my 3800+ X2 :D

[squid_80]

From what I gather they used the 64-bit version of the x264 vfw codec, which is just plain ridiculous for benchmarking.

[Sirber]

seems so, haven't seen the second graph at first.
Is it me or VFW ain't SMP?

[squid_80]

The last time I built 64-bit x264 vfw was in october last year. I think at the time I did accidentally leave the multi-threading disabled. They might be using a build from someone else or maybe celtic_druid's 64-bit ffdshow, but x264 hasn't had working win64 assembly since some time around february. So it would be a) a very old build, b) an unoptimized build or c) both.

[Sharktooth]

or they're on linux... but... dont think so.

[dragongodz]

and where are the E6400 and E6300 results ? since AMD's price cuts it would be interesting to see results of closer priced cpus aswell.

[bkman]

If anyone is doing benchmarks, please compare to 32-bit performance also.

[mezzanine]

Well i don't know how well AMD X2 perform with X264 but my c2d@3.20GHz (E6400) does around 8 fps for 720x576 second pass using HQ Insane Profile (using meGUI).

[dragongodz]

my c2d@3.20GHz (E6400)
not everybody is interested in overclocked results. also for it to be really relevant you have to use the same settings and source and x264 build etc etc etc.

[mezzanine]

Sorry, i am excited

[dragongodz]

Sorry, i am excited
hahaha no problem. it just would have been nice, and more accurate, if that review in the first post had actually covered all core2 cpus. :)

[DeathTheSheep]

Hmm... this (http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=12) seems to be a pretty nifty comparison of the chip's performance, but x264 was *not* tested (instead, Craptime :D)...

Or you could go to the conclusion page, where it says frankly Compared to AMD's Athlon 64 X2 the situation gets a lot more competitive, but AMD still doesn't stand a chance. The Core 2 Extreme X6800, Core 2 Duo E6700 and E6600 were pretty consistently in the top 3 or 4 spots in each benchmark, with the E6600 offering better performance than AMD's FX-62 flagship in the vast majority of benchmarks. Another way of looking at it is that Intel's Core 2 Duo E6600 is effectively a $316 FX-62, which doesn't sound bad at all.empasis mine

... And Intel didn't lose a single benchmark (not by a long shot)...not in gaming, overclocking, etc...and the testing seems very, very complete.
So, the overall results seem to indicate...well...that AMD needs to do something quick :(

But who knows? Maybe x264 is the lone little program that acheives better benchmarks on an AMD! ... ...didn't think so.

PS: Speculation is always fun.

[woah!]

Well i don't know how well AMD X2 perform with X264 but my c2d@3.20GHz (E6400) does around 8 fps for 720x576 second pass using HQ Insane Profile (using meGUI).


if you upload a clip and a avs script i will run it with my amd x2 and let you know what i get. lets also remember that intel have been trailing amd for about 3yrs now without doing anything. now all of a sudden amd have to do something quick?

they will be releasing a quad-core K8L in the new year which is on 65nm silicon, which should be able to run with intels Kentfield quad-core. anyways if amd wasnt about you lot would be still running P3's and paying through the nose for them too.

[mezzanine]

if you upload a clip and a avs script i will run it with my amd x2 and let you know what i get. lets also remember that intel have been trailing amd for about 3yrs now without doing anything. now all of a sudden amd have to do something quick?

they will be releasing a quad-core K8L in the new year which is on 65nm silicon, which should be able to run with intels Kentfield quad-core. anyways if amd wasnt about you lot would be still running P3's and paying through the nose for them too.

My previous proc was an Athlon64 venice core, and i know that if it wasn't for amd we would still pay too much for P3's.
That has nothing to do with the fact that the conroe is an amazing proc and the X2's cannot keep up now.

I wanted a dual core system and at this time and for my budget this was the best solution, i don't care if it is intel or amd or whatever. Both are corporations, it's not a matter of being loyal or anything.

[Revgen]

Intel is using a new generation core while AMD is using an old generation core. AMD will catch up and possibly pass Intel sometime soon.

[Sharktooth]

K8L core has all the "upgrades" the "Core 2" had + some AMD re-touch (Level 3 shared cache, HT 3.0, DDR 3...) :)

[Kostarum Rex Persia]

Well, no, K8L will stick with DDR2. DDR3 for AMD will come in 2008.

[markrb]

The only new advance on the AMD road map is Quad Core, but based on the same X2 core.
Intel is also heading this way.

I don't see another leap like the Core 2 Duo for at least a few years.
I am a die hard AMD fan, but I am getting an E6600 as soon as the prices come down to earth. 20-30 percent over Suggested retial is a bit much for me.

Right now the AMD X2 can't even use all the bandwidth that DDR2 gives as it doesn't need it. The on chip memory controller is very good. At this point DDR3 would be useless until they ramp the speed up much more.

Intel gets a good boost from the DDR2 because it doesn't have an on chip memory controller, it's in the chipset instead.

Mark

[Sharktooth]

KRP: for the first time in history you're right :p well, DDR3 was moved forward in the roadmap.

However K8L core has dual SSE2 (per core) execution units, HyperTransport 3.0, a 2MB (maybe more) shared L3 cache, optional FBDRAM support and Dynamic Indipendent Core Engagement.
So it's a BIG step for AMD similar to what intel did from Core "1" to Core 2.

[Kostarum Rex Persia]

Yes, but K8L will be a server processor, not a desktop platform chip.

Socket AM3 will be desktop solution, probably without shared L3 cache .

[woah!]

Yes, but K8L will be a server processor, not a desktop platform chip.

Socket AM3 will be desktop solution, probably without shared L3 cache .


are you feeling ok?

2 posts i believe are correct!

i think amd will move am3 socket over to desktop pretty quick tho

[bkman]

Intel gets a good boost from the DDR2 because it doesn't have an on chip memory controller, it's in the chipset instead.

Core 2 Duo doesn't really (see some DDR2 benchmarks for it), probably in part due to the memory disambiguation technology.

All in all DDR2 was the most pointless and premature hardware advancement that has been forced upon is in recent history.

[Kostarum Rex Persia]

I think that AMD will support DDR2 1066 in their next generation core, because DDR2 have a lot potential. Don't be surprised if, around 2nd quarter 2007, DDR2 clocks on 1333 MHz.

[Sharktooth]

AM3 is a socket and K8L is a core tech.
AMD always used the same core tech for both opteron and athlons.

[Blue_MiSfit]

Using ffmpegX on a MacPro (dual 2.66 Woodcrest Core 2 Duo) and bvframes, 10mbit, 2pass etc I got ~7fps average for both passes to take a 1080p24 mjpeg video and go to x264....

not bad, and it lit up the cores about 60%

And yes, DDR2 and FBDIMM especially are bullshit that is totally unnecessary IMO...

~MiSfit

[jvrobert]

Just for kicks I just compiled X264 build 588 with ICL 9.1, all optimizations turned on for Conroe only. Quick testing shows it may be a frame or two per second faster on my simple test encodes. I know most of the real work is already in asm, just wanted to see what it could do.

I just grabbed first pthread library I could find, it's not static so just extract x264.exe and libpthreadVC2.dll (included) to same directory.

I dumped it at: http://www.sendmefile.com/00472114

Note this will only run on Core 2 Duo machines, it's 32-bit.

[Thunderbolt8]

atm the current intel processors are indeed faster than the amd ones. my budget advice would be to buy a core 2 duo 6300 for like ~170 € atm and then overlock it to like 2.8-3.0 ghz. youll get performance between a E6700 and X6800 which cost ~700-1000 € and you dont even need water cooling.

[rezard]

but intel's 975,965(for overclocking as you say) chipset mainboards are expensive.
and another cooling for board chipsets may be needed as well as cpu. (depending on the room temperature)

btw, any more benchmarks for x264?

[woah!]

i tried that version above on my E6600 and am seeing no real difference on my encodes. but thats not to say others wont with theirs tho.

[Sharktooth]

i would suggest core2 only if you have plenty of $$$ to spend.. coz only the 6700 and 6800 are really faster than AMD processors.

[bkman]

i would suggest core2 only if you have plenty of $$$ to spend.. coz only the 6700 and 6800 are really faster than AMD processors.

You mean faster than all AMD processors? Because E6600 is faster than all AMD processors, and E6300 and E6400 are certainly faster than some AMD processors (4600+ and 5000+ respectively), and can probably overclock to be faster than all (even overclocked) AMD processors.

[Sharktooth]

E6600 is not faster than ALL AMD processors (FX-62 is on par and there is always the x2 5200+...). however overclocking is always a risk.

[Romario]

E6600 IS faster then majority of AMD processors, perhaps only FX-62 is only on par.

E6700 and E6800 is way too fast for ALL AMD processors.

[Sharktooth]

5600+, 6000+, FX-70, FX-72 and FX-74 are scheduled for the next 2 months...
So dont think it will last forever :p
Also 6700 and 6800 are way to expensive... if i had to chose FX-62 and e6600 i would go for the first, so i can safely updgrade the CPU in the next 3 years without changing motherboard and RAM.

[bkman]

FX-62 is still slower than E6600 in most tasks. I would not call it "on par".

5600+, 6000+, FX-70, FX-72 and FX-74 are scheduled for the next 2 months...
So dont think it will last forever :P

E6800 at least will probably still beat all of of those.

Not to mention Kentsfield is coming out soon too, which will kill all of the above in tasks that can make even use of 4 cores.

if i had to chose FX-62 and e6600 i would go for the first, so i can safely updgrade the CPU in the next 3 years without changing motherboard and RAM.

Afaik no Intel socket changes are planned for a while. Meanwhile AMD is moving to AM3.

[Sharktooth]

FX-62 is still slower than E6600 in most tasks. I would not call it "on par".



E6800 at least will probably still beat all of of those.
it's your opinion... benchmarks on COMPETENT sites show different results. however you cant say E6800 will beat all of those until there are samples to benchmark...

Not to mention Kentsfield is coming out soon too, which will kill all of the above in tasks that can make even use of 4 cores.
kentsfield is not a quad core.. is a dual dual core. AMD will have 4x4 ready for that time... and let's see who's faster.


Afaik no Intel socket changes are planned for a while. Meanwhile AMD is moving to AM3.
intel will change socket the next year when it will be slaughtered by K8L performance.
FYI AM3 CPUs can run on AM2 and AM2+ sockets...

[bkman]

it's your opinion... benchmarks on COMPETENT sites show different results.

No, it's fact supported by many site's benchmarks. You don't get to pick which site is competent based on which shows the least difference between AMD and Intel.

kentsfield is not a quad core.. is a dual dual core. AMD will have 4x4 ready for that time... and let's see who's faster.

4-cores in a processor = quad-core. That it is two dual-cores in a single processor package is just a distinction that AMD fanboys like to make which may not mean much in terms of real-world performance. And FYI AMD 4x4 is a dual-processor platform initially for dual-cores, so it will naturally be slower than a 4-cores on one package.

intel will change socket the next year when it will be slaughtered by K8L performance.

I believe you Nostradamus :)

[Sharktooth]

No, it's fact supported by many site's benchmarks. You don't get to pick which site is competent based on which shows the least difference between AMD and Intel.No i pick the competent sites by looking at the system setups and benchmarking methods.

4-cores in a processor = quad-core. That it is two dual-cores in a single processor package is just a distinction that AMD fanboys like to make which may not mean much in terms of real-world performance. And FYI AMD 4x4 is a dual-processor platform initially for dual-cores, so it will naturally be slower than a 4-cores on one package.
No, dual dual core is 2 cores interconnected with other 2 cores by FSB without the help of the cache (same as core duo BS)... thats like having 2 separate dual core processors. AMD multi processor solutions are much better, more elegant , technologically advanced and alwyas faster than Intel solutions.
So, the more processors in a system, the more the advantage for AMD.
I believe you Nostradamus :)you will believe when the first engineering samples will be taped out... AMD added all the stuff that grants core 2 to be the performance king and even more performance stuff. You can read more about that on this thread: http://forum.doom9.org/showthread.php?t=116127
However we'll see the first eng. samples before the end of the year... so...

[bkman]

No i pick the competent sites by looking at the system setups and benchmarking methods.

And the ones that show AMD more favourably of course... :p

AMD multi processor solutions are much better, more elegant , technologically advanced and alwyas faster than Intel solutions.

Even if that is the case, there will only be a difference in apps that work on the same dataset with 4-threads. There is not much efficiency to gain in apps like POV-ray for instance, where Kentsfield already hits 98.5%. And 4x4 is still stuck with K8 for a while, which tips the scales pretty heavily to Kentsfield's side until Barcelona comes out.

AMD added all the stuff that grants core 2 to be the performance king and even more performance stuff.

They implemented SOME of it, so the performance crown is not clear yet. There is a more detailed writeup over here: http://www.xbitlabs.com/articles/cpu/display/amd-k8l.html

[foxyshadis]

No, dual dual core is 2 cores interconnected with other 2 cores by FSB without the help of the cache (same as core duo BS)...

The original core duo does not run cache coherency over the bus, nor does core 2. In fact the L2 is shared between cores in both designs. You're thinking of the pentium D.

[Sharktooth]

yeah sorry. however lot of speculation on xbitlabs...
AMD already unveiled the K8L architecture... i've already posted the article in that thread.

[bkman]

Xbitlabs analysis is older but based on official AMD slides, same as [H] but much less simplistic. The speculation is based on details that still haven't fully been revealed AFAIK.

[Romario]

x264 and K8L, I just can't wait the results.

What do you think, Sharktooth, can x264 obtain 70-85 fps on K8L?

[Sharktooth]

K8L has been "REVEALED"... there are not undisclosed details..

@Romario: Dunno, we have to wait to see.

[hajj_3]

the e6600 2.4ghz overclocks to around 3.3ghz on air upto 4ghz on water. the fx62 wont overclock too much.

quad core intel is out next month i think, will cost £220-350 for the 2.16ghz, the 2.66ghz will be about same price as the x6800 when it was first released, about £700.

[jvrobert]

Crikey, when did this turn into a misinformation fest and fanboy haven?

To clear some things up:

1. Overclocking boards for Conroe aren't that expensive, contrary to popular myth.

2. Overclocking modestly is not risky or dangerous. Taking an E6600 from 2.4 to 2.8-3.0GHz, for example, is a mild overclock and poses very very little risk even with stock cooling.

3. Kentsfield is quad core. End of statement. There is one physical CPU, and 4 cores, ergo "quad core". In Windows, 4 CPU's will show up. Fanboi distinctions about the number of dies etc... are irrelevent and silly.

4. Conroe is the better choice, performance-wise right now. AMD will have a much better product in K8L and it will be a close call at that point, K8L may be better, it may not be. To assert either that K8L will "destroy" Conroe, or that AMD is out for the count because Conroe beats it right now is ridiculous.

5. 4x4 is a riduclous idea. Seriously, you need to be pretty crazy to drink that kool-aid. Wait for real quad core from either AMD or Intel.

6. Your 975X or 965 motherboard will be able to use any new Intel CPU, more than likely, through most of 2007. No upgrades required. Sometimes AMD fanboys make it "reverse day" and claim Intel needs frequent MB changes. This is nonsensical and the reverse of true. 754/939/AM2 anyone? Now we have AM2+ (K8L), AM3, etc.. in the pipeline too.