AMD Chip Rapes Intel C2Duo discussion

[ikonia]

as I said eariler mencoder

[Clunk]

Well, to give you an idea of how much faster a C2D is over an AMD64....

I had an opteron 170, overclocked to 2.7Ghz with 2x1GB of G.Skill DDR500, thats did me a CCE 4 pass encode in around 1 hour 50mins.

The E4200 that I had, overclocked to 3.4Ghz with 2x1GB, did the same encode in 1 hour 15 mins.

The quad core and the E6600 both do it in just under an hour at 3.4Ghz with 2x1GB.

As for the 939 feeling faster, I honestly cant see that at all on any of the rigs Ive built in the last year or so, unless you compare an overclocked AMD to a stock, lower end C2D, and then it would start to even up a bit.

[chaotic_russ]

chip is 700mhz faster so it will beat the slower cpu most of the time

[Steve B]

pandoras box = open

[KowShak]

The C2D has a bigger cache than the Athlon 64, when your program is small and mostly fits into the Intel cache but not the AMD, the Intel wins the benchmark.

The C2Ds latency to main memory is a lot worse than the AMD because the AMD memory controller is on die and the C2D controller is on the mobo chipset. Once you get an data set that doesn't fit into the processor cache (i.e. encoding 40minute videos) the AMD on die memory controller is going to kill an Intel northbridge based memory controller.

[ikonia]

just done some research while doing some more tests.

The Intel boxes appear to max out one core %99 and every now and then touch the second core - which I can only assume is general OS operations - not the encoding.

The AMD box seems to thrash one core - but not as hard say %80 utilistion, then touch the other core for both the ripping and general OS operations of say %35 -

CHIP CORE 1-ripping CORE 2-ripping CORE 2-OS operations
Intel %99 %0 %10-15
AMD %85 %20 %10-15

I'm using these figures as a rough example as this is not prolonged or exact

But there is a %5 cpu increase on the AMD - and I'm not sure how splitting the load over the multipe cores wiill effect the performance.

Just for interest, and yes, a great can of worms.

[ikonia]

The C2D has a bigger cache than the Athlon 64, when your program is small and mostly fits into the Intel cache but not the AMD, the Intel wins the benchmark.

The C2Ds latency to main memory is a lot worse than the AMD because the AMD memory controller is on die and the C2D controller is on the mobo chipset. Once you get an data set that doesn't fit into the processor cache (i.e. encoding 40minute videos) the AMD on die memory controller is going to kill an Intel northbridge based memory controller. On

Surly most programs will have the ability to use up greater than say 8 meg cache on the intels ?

That would make the Intel performe worse than AMD on almost anything, such as encoding, ripping, gaming, etc etc etc.

I take your point and the on chip memory controller is one of the reasons I've always found the AMD chip a killer for my workstation, however clunk (you better be right !! ) twisted my arm to make use of a intel quad.

I'll study this a bit more as I have lots of ripping to experiment with

[Clunk]

Ummm, the quad thing was your idea You were going to get an extreme edition remember?

Either way, as long as whatever software you are using, can make use of all the cores properly, then you are going to see a huge increase in speed.

Try comparing an AMD and an Intel at the same clock speed, with the same speed/amount of RAM.

I forgot to mention, the thing I use for DVD encoding only uses 70% on the first 3 cores, and up to around 40% on the last core. It still leaves the AMD64 (at any speed) for dead.

[ikonia]

I'm only kidding you clunk, I appricated your help.

The distribution over the cores is how the AMD is behaving, but not how the c2d's are behaving.

I'll have to do further investigation on this.

[Clunk]

I know, I'm joking as well

There are always going to be people that say that this chip does x faster than that chip, the thing is, how can you conclusively prove it? Hard drives make a fair old difference in the encoding process that I use, so maybe a few minutes can be added/subtracted for that. Memory latencies can add/subtract a few mins, as can overall memory bandwidth, as can a drive that is in need of a defrag, or a rogue windows process/service.

Admittedly, I'm talking from my own experiences using XP, and I have no idea of the in's and out's of linux machines.

Its a shame you arent into overclocking, I would love to see a direct head to head at stock, and then at a reasonable overclock, and then a full overclock, and then with the RAM tweaked as well

[directhex]

I'm only kidding you clunk, I appricated your help.

The distribution over the cores is how the AMD is behaving, but not how the c2d's are behaving.

I'll have to do further investigation on this.

what mencoder options are you using? most codecs will allow multithreaded operation

[GSte]

ALthough I have a raptor, my brother's Opteron 170 rig which is running at 2.6GHz seem so much slower than my e6600 at everything that it feels frustrating to use at times, but as Clunk says there are many other factors that will affect the performance of both machines. I am inclined to trust the tech sites' benchmarks on this one though as they are all in agreement about which chips are faster. I'm sure if AMD chips were faster in any situation they'd be shouting it from the rooftops considering what a financial kicking they are taking

[kalniel]

I'm only kidding you clunk, I appricated your help.

The distribution over the cores is how the AMD is behaving, but not how the c2d's are behaving.

I'll have to do further investigation on this.

I suggest having a look at supreme commander as well - it exhibits the same behaviour with regard to AMD and intel multi-core chips. If you have a look at their forums some users have done a very details breakdown of what's happening, and come up with a tool that forces thread redistribution across cores which rectifies the situation. Something similar could probably work in this case.

[Oobie-]

I too would have thought that Hard drive speed/size/fragmentation have a part in this

[ash_rm]

I compared my Opteron 165 @ 2.7GHz to my E4300 @ 2.7GHz for encoding Mpeg2 to Xvid and the E4300 was a good 10-15% faster. (using gordian knot btw)

[Defenestration]

I've read about a problem on laptops where work is not always distributed evenly between the cores, and this can be fixed by a hotfix. Not sure if this also applies to desktops.

Give me a moment and I'll try to find out where I saw it.