Socket 939 Sempron found........

[Gamingphreek]

Originally posted by: porkster

Originally posted by: OpteronSo Intel runs games faster.. OK ! ...Something smells fishy here.

At the start of a test with a clean score reset, the Intel lead on all the sub tests. When the Intel crashed, the AMD caught up. Also there was a strange situation where the AMD slowly distracted itself form the divx task and improved on the other three.

The overall observation is the Intel 840EE would have nailed the AMD X2 if the AMD stuck to the divx thread. Maybe THG needed to install a AMD CPU driver to get the cpu to equally dedicated itself.

The THG was a real life test on what a customer could expect to see with the system sitting in front of them, not these single/solo brenchmarks from other sites that don't reflect the true nature of the CPU other than to prove who's winning from the angle of the reviewer that has a vested interest in one CPU winning.

Here is an example of a AMD cpu driver that allows os user to change priority and other schemes. http://www.flexbeta.net/main/comments.php?catid=5&shownews=13382

.

Youre insolence and relentlessness are truly remarkable.

Please if you are SO much smarter than us please, give me a full run down of AMD's and Intel's microarchitecture. Then tell me the pros and cons of each. Do you even know what dual core or any of these terms are?

-Kevin

 

[porkster]

The other mistake posters are doing is stating sub-test wins by pointing out loop counts that were in error, due to the other machine being shutdown.

All the loop counts were off, including the Farcry one. So the Intel was doing a massive effort leveling with the AMD cpu on Farcry whilst it converted DVD's on the fly to a rough factor of 100:1(AMD).

.

 

[Gamingphreek]

I'm replying to posts that contest my opinion of the THG tests

1. You cannot turn concrete/quantitative data into opinion. There are numbers. I can say that 3>5 all day, doesn't make it true.

2. If this is, somehow, your opinion, then why in the hell do you care what we think. If you think that your Intel chip is 40000009090000093904309% better than why are you wasting your time trying to convince us, the people who supposedly "know nothing".

-Kevin

 

[Markfw]

Originally posted by: porkster
The other mistake posters are doing is stating sub-test wins by pointing out loop counts that were in error, due to the other machine being shutdown.

All the loop counts were off, including the Farcry one. So the Intel was doing a massive effort leveling with the AMD cpu on Farcry whilst it converted DVD's on the fly to a rough factor of 100:1(AMD).

.

The counts on the intel were over the same 95 hour period, with only an hour or two to change out hardware. Use the same divisor for both, 95 !
Intel : 494 CD's divided by 95 hours = 5.2 CD's per hour
AMD : 472 CD's divided by 95 hours = 4.968421053 CD's per hour 0.955465587 4% behind

WINRAR

Intel : 2695 archives divided by 95 hours = 28.36842105 rar's per hour
AMD : 3665 archives divided by 95 hours = 38.57894737 rar's per hour 1.359925788 36% ahead

FarCry

Intel : 1719 runs divided by 95 hours = 18.09473684 runs per hour
AMD : 2056 runs divided by 95 hours = 21.64210526 runs per hour 1.196044212 20% ahead

Divx

Intel : 2240 minutes divided by 95 hours = 23.57894737 minutes encoded every
AMD : 170 minutes divided by 95 hours = 1.789473684 minutes encoded every 0.075892857 93% behind

So where do you get your facts from ??? Imagineing them ???

Edit, not to mention the point above that none of the tasks are weighted, so it also doesn;t really show who is winning overall. And it supposed to be a stability tes, and we all know Intel failed there. And it won;t run SLI sable at all. So what does it win in ? It can suck more power, or generate more heat ?

 

[porkster]

Gamingphreek, it's not about being better than others, it just the small effort to read the results properly from the test conducted.

.

 

[AnonymouseUser]

Originally posted by: porkster
I'm replying to posts that contest my opinion of the THG tests. Alot of people saw the Intel winning all the tests for themselves. There is no lie about the statement.

Can you stop declaring situations that are off topic to the thread, please.

.

Porkster, you are correct. Intel WINS! Bask in the glory! I can imagine you will be the talk of the Anandtech forums for some time, as well as a few other forums, for your amazing knowledge and stupendous ability to declare a clear winner despite the negativity trolls that constantly berate and attempt to contradict you. Your mommy will be so proud!

Now go take a (long) vacation or something. You deserve it.

 

[Gamingphreek]

Originally posted by: porkster
Gamingphreek, it's not about being better than others, it just the small effort to read the results properly from the test conducted.

.

Like i said. If we are too stupid to read the tests right why are you trying to convince us. Obviously your knowledge far supercedes our own :roll:

-Kevin

 

[porkster]

The frames-per-second are the only valid value when quoting the Farcry sub-test results. Also the other counts like I mentioned were also off due to the Intel system having a board change, etc. In fact there were 3 interruption on the Intel system for the AMD to catch up. Three interruption is = the 3 reboots on the counter.

You are mistaken to think that both systems ran for the same test duration. It's just totally wrong.

.

 

[AkumaX]

porkster, youre a god

 

[boran]

really, there should be some forum rule against trolling. I was having a meaningfull discussion about the finer points of the windows XP thread scheduler and then this guy here degenerates the whole thread into a trollfest., an ignore button would also be handy.

Ppls, just ignore im, if everybody acts like he aint here it will just go away. he basks and laughs his ass off by making these bold statements just to provoke us. and if you decide to trollbash, at least add something meaningfull after the bashing (see this post ) that way the thread goes at least forward instead of standing still in a mudsling contest.


Anyways, I'm still wondering if anyone knows how the OS deducts the CPU usage of a CPU, wether it's a virtual one or a real one.

Another main question is, do these processors work with windows XP home, I had red somewhere amd fakes HT to get it working with home, anyone can confirm this ?

and the main question still stands, does winXP recognize 4 cores as being 2 in hardware ? There's still not 100% confirmation the DivX thread was set to low (if anyone can confirm please do). but as I said before, if the DivX thread was set to lower priority than the rest the answer on this question is: no windows XP does not see the 4 cores as 2 times 2 virtual ones.

 

[Duvie]

Originally posted by: Opteron

Originally posted by: Markfw900

Originally posted by: Opteron

Originally posted by: keysplayr2003

FarCry

Intel : 1719 runs divided by 73 hours = 23.54 runs per hour
AMD : 2056 runs divided by 95 hours = 21.64 runs per hour

So Intel runs games faster.. OK !

Something smells fishy here.

Check my updated numbers on the previous page.... The AMD wins two, almost ties one, and looses one big time.

You are calculating downtime as work time, and keysplayr2003 do not calculate time before last reboot.
He presents those number like it only took 73 hours for Intel to make all the work..

NO....Keys is really fvcking this one up...Tghe INtel machine was not down for 22 hours and they did not reset the score after the Intel system crashed twice and was shut down for few hours at the most to place new HSF on it....You can't just count the latest uptime...This is fvcking stupid as anything said in this thread....

Get it right ppl...The INtel cpu was calculating at similar rates for many many hours before it even rebooted then was up for at leat 4 more hours before it booted again....I would say maybe take the numbers and divide them by 90 as the true uptime for the INtel cpu...

 

[Duvie]

Originally posted by: porkster
The frames-per-second are the only valid value when quoting the Farcry sub-test results. Also the other counts like I mentioned were also off due to the Intel system having a board change, etc. In fact there were 3 interruption on the Intel system for the AMD to catch up. Three interruption is = the 3 reboots on the counter.

You are mistaken to think that both systems ran for the same test duration. It's just totally wrong.

.

All the downtimes (especiually restarts) were minimal and not enough to change results of Far Cry or Winrar....CD encoding would be further ikn INtel's favor though.....Quit trolling or I will start to see that you are removed...

I will chalk up your latest comments to continual stupidity....

 

[Markfw]

Yea Duvie, I didn;t follow close enpough to know real downtime, I figured it was like an hour max...

 

[Gamingphreek]

Originally posted by: boran
really, there should be some forum rule against trolling. I was having a meaningfull discussion about the finer points of the windows XP thread scheduler and then this guy here degenerates the whole thread into a trollfest., an ignore button would also be handy.

Ppls, just ignore im, if everybody acts like he aint here it will just go away. he basks and laughs his ass off by making these bold statements just to provoke us. and if you decide to trollbash, at least add something meaningfull after the bashing (see this post ) that way the thread goes at least forward instead of standing still in a mudsling contest.


Anyways, I'm still wondering if anyone knows how the OS deducts the CPU usage of a CPU, wether it's a virtual one or a real one.

Another main question is, do these processors work with windows XP home, I had red somewhere amd fakes HT to get it working with home, anyone can confirm this ?

and the main question still stands, does winXP recognize 4 cores as being 2 in hardware ? There's still not 100% confirmation the DivX thread was set to low (if anyone can confirm please do). but as I said before, if the DivX thread was set to lower priority than the rest the answer on this question is: no windows XP does not see the 4 cores as 2 times 2 virtual ones.

IIRC XP Home does not support SMP/ Dual Core. Additionally AMD in no way uses Hyperthreading technology.

THe only reason the intel uses it is because of the extremely long/deep pipeline. A cache miss could cost Intel ENORMOUS losses. Additionally it looks something like this (in a nutshell)

|= Data

>>>>>|>>>>>>| = Intel with HT. Notice the two groups of data being sent. That way the pipeline is always busy, and not just sitting there waiting to fetch and execute more data.

>>|>> = AMD. Notice it is so short that HT would do nothing for it. In fact it may even hurt performance. A cache miss wont cost it nearly as many clock cycles, additionally the pipeline is nearly always doing something. It would be pointless to send 2 data streams through.

-Kevin

 

[Duvie]

Anyone find the intel pattern for the last hour to be interesting??? especially since it rate of change ( I am starting to track numbers on an hourly basis) in apps are the same...Seems odd considering one to 2 cores each time are running 66-70% range.....

I have never seen more oddities in a test then this one ran by THG....

 

[porkster]

Yep, boran, lets talk about the poor dediction of the divx thread.

I'm currently 50/50% on the OS and the CPU being the problem of divx thread. Why did the AMD X2 slowly move it's priority off that application or thread if it was a seperate thread?

What would have been the scores if the AMD X2 was forced to do the divx thread?

.

 

[porkster]

Originally posted by: DuvieI have never seen more oddities in a test then this one ran by THG....

Totally agree with you , Duvie.

.

 

[Gamingphreek]

Originally posted by: porkster
Yep, boran, lets talk about the poor dediction of the divx thread.

I'm currently 50/50% on the OS and the CPU being the problem of divx thread. Why did the AMD X2 slowly move it's priority off that application or thread if it was a seperate thread?

What would have been the scores if the AMD X2 was forced to do the divx thread?

.

That would be your error. The processor has, IN NO WAY, control over application priorities. That is all controlled by Windows.

-Kevin

 

[boran]

and as I said before, the numbers mean exactly nada nothing nill, zero, de botten, niets.
they cannot be compared amongst eachother nor between eachother, because on the AMD proc there are at least 2 threads running one one proc (it has but 2 it runs 3 benchmarks very decently) and on the intel 4 threads are running heavily on two cores. (four virtual) so the numbers cannot be compared to eachother and as I said before how much worth is one CD encoded in divX minutes ?

The only thing you can make out of the numbers is one of two things:
either all threads are equal priority, which means windowsXP is faulty behaving on the AMD and not handing enough processor time over to the encoding thread (you'd then see one thread alternating probably, or cycling thru)

or the DivX thread is at lower priority which means the windowsXP sheduler is faulty in assigning so much processor time to the encoding thread.

either way the issue lies with the process scheduler of winXP.

and for the rest nothing can be concluded out of the numbers from a performance point of view.

edit:

That would be your error. The processor has, IN NO WAY, control over application priorities. That is all controlled by Windows. -Kevin

Gamingphreek ignore Porkster, it has no use, and if we continue the bickering and overreacting this thread will get as locked as the other.

 

[porkster]

Originally posted by: GamingphreekThat would be your error. The processor has, IN NO WAY, control over application priorities. That is all controlled by Windows.

-Kevin

Unless it had opcode/instruction congestion within the CPU. This was my first statement about the AMD being faulty, if you recall.

The other factor could be the Intel is over doing the divx thread due to having the extra thead ability, but that doesn't explain why the AMD slowly moved off the divx thread. The AMD always start nice on the dvd conversion then it slowly died after about 20 loops and then started pushing the other sub-tests better.

.

 

[Gamingphreek]

Unless it had opcode/instruction congestion within the CPU. This was my first statement about the AMD being faulty, if you recall.

Wtf are you talking about!?! Are you just spouting out random microarchitecture terms?

-Kevin

 

[Duvie]

Originally posted by: porkster
Yep, boran, lets talk about the poor dediction of the divx thread.

I'm currently 50/50% on the OS and the CPU being the problem of divx thread. Why did the AMD X2 slowly move it's priority off that application or thread if it was a seperate thread?

What would have been the scores if the AMD X2 was forced to do the divx thread?

.

It didnt...this is how things react when you have more threads being assigned then cores to do them and 1 of the apps is low priority...please go read up on prioirty process and you will see that since it is low priority it is like a guy standing at a front of a restaurant holdingthe door open for other ppl behind him....except the line is endless....

This is no different then a single core A64 now acts when you run Folding at Home (low prioirty default) and a cpu intensive encoding app.....This is no shock ppl...Really!!!

Not a true analogy cause we can see it does get in a few cycle now in then...Like I mentioned above this may have more to do with in between times of the other apps...

 

[imported_Opteron]

INtel machine was not down for 22 hours and they did not reset the score after the Intel system crashed twice and was shut down for few hours at the most to place new HSF on it....You can just count the latest uptime...This is fvcking stupid as anything said in this thread....

I didn't say anything stupid, you just interpreted it that way.

Thing is that, you both should take down time in to account if you are evaluating performance.

Reliability has already been solved, and Intel won't even deliver XE840 CPU's this summer becouse it's only a PR release.

 

[porkster]

Originally posted by: boran
and as I said before, the numbers mean exactly nada nothing nill, zero, de botten, niets.
they cannot be compared amongst eachother nor between eachother, because on the AMD proc there are at least 2 threads running one one proc (it has but 2 it runs 3 benchmarks very decently) and on the intel 4 threads are running heavily on two cores. (four virtual) so the numbers cannot be compared to eachother and as I said before how much worth is one CD encoded in divX minutes ?

An analogy could be, the Intel has four horses pulling the wagon, whilst the AMD has two race horse pulling theirs. Both wagons get pulled along.

Just becuase the AMD only has two real threads, that doesn't mean any higher thread requests are ignored, it has a duty to share the requirements out fairly, based on the OS priority levels. Window's XP's kernel is programmed to share fairly those requirements to multitask.

.

 

[Duvie]

Originally posted by: Opteron

INtel machine was not down for 22 hours and they did not reset the score after the Intel system crashed twice and was shut down for few hours at the most to place new HSF on it....You can just count the latest uptime...This is fvcking stupid as anything said in this thread....

I didn't say anything stupid, you just interpreted it that way.

Thing is that, you both should take down time in to account if you are evaluating performance.

Reliability has already been solved, and Intel won't even deliver XE840 CPU's this summer becouse it's only a PR release.

I agree...I wasn't trying to calculate anything over the duration of the test (2nd round)....

That is why when I show performance numbers I do them in rate of change from previous scores hours before...i have long said the numbers are borked to just look at tallied total right now...