TR Forums

Ma10n3! wrote:

How many folks out there actually try to DIVX encode, compress MP3s, and play a DirectX 9.0+ game all at the same time on the same machine?

Umm, actually, I have done that...well, sorta...the game was UT2K4 - which is certainly NOT Far Cry (and this on a 2.4GHz P4C)...but on a dual-core, that really shouldn't be a problem; isn't that the whole POINT of dual-core?

If a single-core cpu can (and it should be able to!) encode MP3s while playing game (MP3 encoder set to low priority, should be fine, "normal" for HT machines, I would think), then a dually should be able to handle DivX along with it...

As for the CPU prefetching instuctions: that SHOULD NOT BE A FACTOR. Windows sets the task priorities, Windows dispatches tasks to the CPU. Windows is simple not giving the encoding task ample prioritiy. We see this alot on single-core, non-HT machines under heavy load...the UI is sluggish (at best). With HT (or a dually), Windows is throwing more work at the cpu in the same time frame. Same thing is happening here; the 840XE is HT enabled; it shows to Windows as 4 cpus; the X2 shows as two cpus. Windows is simply able to dispatch more work to the CPU.

If they took off a single app from that test, the results would shift to favor the X2 big time.

Porkster wrote:

Would you trust a brenchtest result when you own a ATI graphcis card but the game has Nvidia plasted all over the product?

It is a fact, some companies allow for pro-advertising. Pro-Advertising is where another product is used to bolster another. Much like TV station do when one tv show will promote another within its content. The game develeopers do the same and will tailor a game to suit a paying client's promotion, be it Nvidia or AMD, or Intel, etc.

Doom3 is a Nvidia and AMD paid game, so I don't take accoutn of it's results in speed tests.

Wow. I'm done here.

It would be a lot clearer if they just posted frames per second for the divx test. We also have to acknowledge the possibility that one of the systems has gained sentience and just doesn't want to watch "die another day" again. Awful movie.

danazar wrote:

It seems that they're counting the amount of time it takes to compress the entire movie "Die Another Day" into the Divx format. It takes the AMD processor 30 minutes to do this, and the Intel processor 470 minutes. It's illogical to assume it's the other way around on the sheer basis that there's not 470 minutes of stuff on a single DVD to encode, so this is likely the correct interpretation, which means that it is Intel that is having its butt handed to it on a plate.

Not sure if there is a slight chance you're correct, but if so then the 30 mintures would be the final result. Like once done then it dosn't need to re-encode, the value listed would then be the best able encoding. If you have been monitoring the test, the score has slowing increased upto 30 from 0.

The logic is still pointing to the minutes value being the total data encoded, referenced in minutes of movie time/data.

.

They're not very clear on what is being measured. The chart says encoded minutes, the article says encoding time.

Another reason to hate Toms, I guess...

HT doesn't just work for HT aware apps. HT helps in multitasking, especially on single core cpus. On a dual-core, it can hinder some things, but still, 4 vs. 2 will help multiasking when there's enough junk running.

One "app" that REALLY needs to be HT aware is Windows...then maybe it could dispatch tasks a bit better and HT would always help instead of hinder...

Porkster wrote:

Not sure if there is a slight chance you're correct, but if so then the 30 mintures would be the final result..

Yes, the "30 minutes" would be the amount of time taken to encode a DVD movie, whereas the "470 minutes" would be the same amount of time taken for the Intel processor to encode the same DVD movie, supposedly with the same settings. (Review sites like this tend to use the same Divx quality settings in each test, so the resulting Divx file should be roughly the same quality and file size.)

This suggests Intel is getting smeared due to a lack of core prioritization in the OS or the app.

TheDVDMan wrote:

HT doesn't just work for HT aware apps. HT helps in multitasking, especially on single core cpus. On a dual-core, it can hinder some things, but still, 4 vs. 2 will help multiasking when there's enough junk running.

Maybe someone could correct this if wrong, but wasn't many saying the HT abilities are more or less useless on mutlitcore cpu's? The percentage increase was like at best 7% and at some times gave worse results compared to having HT turned off.

In the test the Intel system is doing a fantastic job of keeping the frame rates in Farcry on par with the X2. We all have seen that Farcry suits AMD cpu's so it's a good result for the Intel system.

.

danazar wrote:

This suggests Intel is getting smeared due to a lack of core prioritization in the OS or the app.

It's unlikely though, as like I said the results are increasing. The logic isn't backing your argument.

.

Uhh...some of you aren't understanding what this test is actually doing (not that I blame you, that article doesn't make things very clear, and finding anything in the mess of ads is another story). It's a stress test, so they've started loops on all of these applications and have been continuously running them. So since the start of the test, the Intel machine has encoded 500 minutes of DVD data, while the AMD machine has only encoded 30.

But to what that actually means, is entirely up in the air. There are so many variables that aren't being controlled with this test, that it's unreasonable to only compare the CPUs in the test.

Porkster: it looks like DivX is important to you. Is your world in upheaval? Do you still cling to Intel=good, AMD=bad dogma? Encoding a DVD to DivX is, if not Verboten or illegal , is the gateway to piracy! To lunacy! Next thing you know, you'll be sharing out naked pictures of your housepets! Or worse: naked pictures of your house lamps!

It may be hard to believe, but the X2 is kicking some ass. We'll have to watch the stability, as both machines have required a reboot.

I propose, however, since the explaination of terms for the Tom's DivX graph is not particularly good, why not fire them an email and have them further explain it?

Thanks for the heads up for the article.

drsauced wrote:

...It may be hard to believe, but the X2 is kicking some ass. We'll have to watch the stability, as both machines have required a reboot.

The AMD X2 is failing big time, sorry. If you haven't notice, (which is the topic of the thread) the AMD X2 is on the edge of ignoring the divx encoding thread.

If the AMD X2 was doing its equal share of time for the Divx encoding process, then it would be losing all the other test values. Some say the advantage is Hyper Threading on the Intel CPU's, but this is not argument.

Intel 840EE can do 8 1/3 hours of dvd to divx encoding whilst the AMD X2 is near to ignored the thread and does a mere 30 minutes. Something is major wrong, even if the OS thread is on low.

.

umm dan, I dont think your correct about that one, I am going to email tom and ask, what exactly its measuring.

Two more things, these tests are actually more geared towards Intel platforms with the two media encoding (LAME and DivX) benches. I haven't ever seen benchmarks about RAR archiving before, at least not remembered, and FarCry is probably the least geared toward any particular CPU.

This thread seems to have ventured into OS vs. CPU territory, which deserves its own lecture hall. The P4 needs to use hyperthreading because it normally has a low IPC; Hyperthreading enables the P4 to use its long pipeline better. AMD CPU's are better in this regard as they have better use of its pipeline, and thusly AMD uses the PR rating to make it somewhat clear about the fact.

Good reading about context switching, from the masters, IBM:

http://www-106.ibm.com/developerworks/l ... ary/l-rt9/

This really sounds more like a software problem, i.e. the Windows scheduler is getting confused by something in AMD's dual-core implementation.

probably priority. If they switched the priority then we might see some differences. Obviously though the amd system will go somewhat slower, how much though, is unknown. (especially since I don't have amd x2 or a p d ee, or ever will).

However it is an interesting test. Since both have have been reset already. The amd one went first then the intel one. (I think the intel one was down longer which would explain the time differences between the two).

You have to understand that the load on the systems is utterly unrealistic. First of all, I'm pretty sure that the divx encoder is running on "idle" task priority (since divx is multithreaded and would monopolize the cpu time otherwise) so the XE system has the advantage in that it can dedicate a thread to the process, whereas the X2 system has to be switching threads between the two cores. If you want to compare performance between the two cpus you'd be better off just looking at a review with a more realistic load.

Personally, I think running 4 different programs and getting four different performance metrics doesnt allow you to compare the results properly. I understand that they are running 4 programs to fully load the intel cpu but it would be easier to say which cpu performs better if, for instance, they were running 4 instances of winrar (or of lame, for that matter).

Anyway, I think that the main point of this test is to measure the stability of the systems--not to measure the performance of the cpus, since that has already been done all over the net--and they just provide the performance information just for the hell of it, just as they allow you to monitor the memory usage, etc.

Now, slightly out of topic, I really hope AMD doesnt bundle that crappy heatsink with the retail X2s. Considering the price they should at least bundle the heatpiped one that comes with the FX.

i know the load is unrealistic but that is the point.

just brew it! wrote:

This really sounds more like a software problem, i.e. the Windows scheduler is getting confused by something in AMD's dual-core implementation.

It wouldn't surprise me if were a BIOS or NF4 chipset thing, either.

I believe the DivX times are encoding the video in a loop, which is how they could get 470 minutes. Same thing with the LAME and RAR tests. They compress the same CD and software each time and count how many total cycles could be performed within the alloted time.

For the Bond DVD, assuming 90 minutes, the Intel could do 5.2 cycles while the AMD could only do 1/3 of a cycle.

In the great words of Yahoolian, Don't feed the piggy...

Furen wrote:

You have to understand that the load on the systems is utterly unrealistic. First of all, I'm pretty sure that the divx encoder is running on "idle" task priority (since divx is multithreaded and would monopolize the cpu time otherwise) so the XE system has the advantage in that it can dedicate a thread to the process, whereas the X2 system has to be switching threads between the two cores. If you want to compare performance between the two cpus you'd be better off just looking at a review with a more realistic load.

Personally, I think running 4 different programs and getting four different performance metrics doesnt allow you to compare the results properly. I understand that they are running 4 programs to fully load the intel cpu but it would be easier to say which cpu performs better if, for instance, they were running 4 instances of winrar (or of lame, for that matter).

Anyway, I think that the main point of this test is to measure the stability of the systems--not to measure the performance of the cpus, since that has already been done all over the net--and they just provide the performance information just for the hell of it, just as they allow you to monitor the memory usage, etc.

Now, slightly out of topic, I really hope AMD doesnt bundle that crappy heatsink with the retail X2s. Considering the price they should at least bundle the heatpiped one that comes with the FX.

Welcome to the forums!

I agree with ya here. The load is unrealistic. Even if I had dual cores yadadada I'd prolly never try to encode something massive while i was gaming or otherwise. But on OTOH im pretty ignorant aside from what i've dealt with...

Am I correct in assuming that with two cpus you can completely and totally give two of them a different task and it will fair just as well as well as two seperate computers doing those tasks, or is there a hit somewhere along the line. My system is aging fast, and by income tax time i'd like a nice new system, something to last another 3-5 wiht minor upgrades.

pete_roth wrote:

Am I correct in assuming that with two cpus you can completely and totally give two of them a different task and it will fair just as well as well as two seperate computers doing those tasks, or is there a hit somewhere along the line.

There will be a performance hit because of shared resources, etc. The performance hit going from single to dual core with AMD cpus is much less than the hit on the Intel side, which is why basically everyone thinks AMD dual-core cpus much better executed.

Edit: I'll give you some examples

If you have two tasks that are very cpu intensive but do not access the memory/bus then you'd probably get close to no performance hit.

If you have two tasks that are writing to/reading from disk, then you'd probably end up being limited by your harddrive

So basically as long as the two processes dont interfere with one another too much, you'd get close to no performance hit. The real boost from dual core comes with programs that use both cores at the same time. For example, when encoding with a multi-threaded program (like divx5), dual core AMD cpus encode very close to twice as many frames per second (not sure about intel since I really dont care about pentium Ds ).

In response to the original topic:

Yeah, it really looks like a hardware problem.

I actually dont think it's a hardware problem.

If you think about it it makes sense:

you have 3 processes at normal priority, 1 at idle (I'm guessing) and you have 2 execution cores, of course the process at idle gets neglected; try encoding with single core cpu (with no HT) while doing something else that is VERY intensive and has higher priority, you'll probably get very little progress on the encoding.

The same thing would happen if the intel cpu was running more processes than it has the ability to execute, it's just that 4 threads happens to be the sweetspot for intel.

sarcasm? anyone? do i need more than three rolling eyes smileys in there?

haha, sorry, I ignore smileys while reading, hehe. And since everyone else has been talking about hardware problems I just assumed you agreed.

Now are they trying to run more than two intensive processes at one, on that AMD X2? Because that isn't a good idea, since it does not do HT (which allows the Intel chip to work on 4 things at once)...

Why not turn off the HT on the intel chip and then do the benching....?

We are hearing alot of excuses here.

* consider both systems have the same starting base. That means exactly the same files and operating system settings. So if "IDLE" is set on one unit then it's also on the other. The only difference is the drivers that are loaded to suit the different hardware.

* consider the point, is it reasonsilbe to have a CPU ignor a process if it finds that task difficult to do? Great for winning benchtests for those that don't look deeper into results, but a poor performer when purchased and expected to do a work load.

.

Porkster wrote:

* consider the point, is it reasonsilbe to have a CPU ignor a process if it finds it difficult to do?

The CPU can't "ignore" a process. Process scheduling/prioritization is done by the OS, not the CPU.

This is starting to verge on tinfoil hat conspiracy theory type stuff...

just brew it! wrote:

Porkster wrote:

* consider the point, is it reasonsilbe to have a CPU ignor a process if it finds it difficult to do?

The CPU can't "ignore" a process. Process scheduling/prioritization is done by the OS, not the CPU.

This is starting to verge on tinfoil hat conspiracy theory type stuff...

Hardly its just someone that doesn't know hardly anything about OSes and processor talking out of their ass.

Porkster, as much as I like intel, you don't have enough knowledge about hardware or software to make you even close to dangerous except to the side you are backing. I would suggest a little research before making comments...