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

[Duvie]

Originally posted by: MDme
what tests will it win?? certainly not winrar and farcry. the cd encoding is too close to call really. divx is obviously intel. the question would be what if they change the order they start the programs. maybe the x2 will be more competitive with intel in divx.

Order wont effect this one....This is clearly priority settings....

Intel vs AMD (in any 2 of the 4 test)

Winrar + gaming = AMD kills
Winrar + encoding = Amd wins

AMD wins all 2 test combos except potentially encoding divx and CD if both are multithreaded

Intel vs AMD ( in any 3 of the 4 test)

I think Intel wins most of these due to the 4 cores (2 real, 2 virtual) probably would not win a winrar + gaming + cd encoding combo....AMD may take on of the test and have the second almost a tie but the third would be an Intel advantage

INtel vs AMD ( using 4 test ) intel ultimately with a stable system would be up in 2 close in the third and lose 1....


This is no spin ppl...This is looking at data as I have seen it in the test during early stages of each round of testing prior to crashes....


Now like I have mentioned repeatedly different setup, different reviewers (wink), and different apps could give different results in the 3 test setup but I think the iNtel wins or at least ties in most 4 app test I think of...Now this is 4 cpu intensive apps, meaning an app that could use 100% of 1 core by itself...Not just 4 apps running cause things like watching a DVd only takes like 10% cpu usage on some PCs...Capturing DVD video as a pBVR only take 50% or less on most PCs....I am talking 100% cpu hogs like encoding, rendering, etc...Ones that will go as fast as you let them....

Just have to talk honest here...

honestly I would never run this number of cpu intensive apps...I can see 2 with some other apps that may make up 30-50% more but that would it and not for extended periods of time....

 

[Duvie]

Originally posted by: cscpianoman
I love the new update just posted today at THG. Talk about instability, they went through 5 motherboards for the Intel box before they could get one to work and that board is still in beta!! Over both tests AMD has not required any reboots, while Intel has required several. Is anyone getting the impression Intel rushed this out the door and has overheating issues?

Well one thing can be said...Intel EE models with stock cooler just are not going to cut it....I guess for those who spend 1000+ on a cpu a 40 dollar aftermarket cooler should be a no-brainer.....

I agree it shows the sckt 939 platform is mature....rock solid for X2s which likely were not the intended target..It appears the X2 and their relatively lower power needs are key here....Look at INtel...i see many a person with blown power regulators on their boards...

 

[Keysplayr]

Originally posted by: JackBurton

Originally posted by: Zebo
I've seen that intel box reboot 3x now toms still says once.

No sh!t! What is that about? The AMD machine has been up for 22hrs+ with 0 reboots.

Yes there were some very annoying mobo issues all in a row.

Reboot ONE
The first board, Gigabyte, was suffering from a very hot northbridge. The mobo manufacturer suggested a BIOS flash would help alleviate this problem. Tom's flashed the BIOS as per Gigabyte and killed the board completely. Never posted.

Reboot TWO
Next we have the Epox board. It would seem that the voltage regulators (mosfets?) overheated and burned out. Board 2 gone. This counts for one of the two reboots at least.

Reboot THREE
Then the ASUS board. Suffered from high heat. This I believe is the 955X board. The other 2 boards were Nforce 4 boards correct? This board was reporting extremely high temps much higher than possible. Fan speed was painfully slow on the stock Intel HSF that came with the Pentium D XE. 2100rpm at its fastest.

So thats 3 reboots. Finally we now have the ASUS 955X with BIOS update and the Intel HSF was replaced with another much faster spinning Intel HSF, about 4000 rpm.

Temps are now reporting more along the lines of where they are supposed to be at around 68 to 70 under full load. Still hot, but it has to be a helluva lot cooler than a HSF with a 2100 rpm fan. We shall see if it can maintain stability now. Who knows, maybe the plague of problems have been dealt with and maybe they have not. I hope we won't see any more reboots.

 

[Keysplayr]

AMD FarCry dropped to 15fps. Could be FarCry might crash soon. It crashed once before on the AMD rig. Might come back though.

 

[Keysplayr]

Next refresh

..................CD's .............RAR...............FC.............DIVX

AMD...........130.................1104............658..............30
Intel............117................629...............428............560
difference....-13..................-475.............-230...........+530

don't think INTEL chip is "creeping up", except for DIVX[/quote]

I think it is in everything except WinRAR. Since my last post, Intel has gained 2 CD encodes, 3 Far Cry runs and 30 additional minutes in Divx. AMD gained 10 archives in winrar. Intel is definately creeping andrewin, albeit slowly. If we, or anyone else can figure out why the intel chip has a heartbeat in the graphs, and rectify it, the creeping would become crawling.


[/quote]

..................CD's .............RAR...............FC.............DIVX

AMD...........143.................1188............702..............40
Intel............130................697...............478............620
difference....-13..................-491.............-224...........+580
.........gained 0...........lost 16..........gained 6.........gained 50

 

[bREwIn dOdgEr dUDe]

after a couple o' days of seeing the last gasps of a (very) proud few, and alot of reading, flaming, chest-thumping & such... will someone please put a fork In(tel) and call this ThG stress-test somewhat "well done"... or as they say in japan... c-ya?!?...

 

[Duvie]

Originally posted by: keysplayr2003
AMD FarCry dropped to 15fps. Could be FarCry might crash soon. It crashed once before on the AMD rig. Might come back though.

I thought that crash was a bug that caused it to quit at 1000....As far as everything i have seen this board and cpu have been ROCK SOLID, STABLE, and very CONSISTENT.....3 things I do like in a board.....

I think one has to be concerned about retail package at this point and possible all the way down to the 820.....I hope newbies are watcing this and learning..These still are prescott heater cores....

The longest the intel machine went was a reported 22 hours but it had actually carshed 2 times in that period...I think if it makes it to 12 hours that will be the longest it has offciially gone without a crash...

 

[AnnoyedGrunt]

I'd like to see a sepearate power measurement for each system.

-D'oh!

 

[Duvie]

Originally posted by: keysplayr2003
Next refresh

..................CD's .............RAR...............FC.............DIVX

AMD...........130.................1104............658..............30
Intel............117................629...............428............560
difference....-13..................-475.............-230...........+530

don't think INTEL chip is "creeping up", except for DIVX

I think it is in everything except WinRAR. Since my last post, Intel has gained 2 CD encodes, 3 Far Cry runs and 30 additional minutes in Divx. AMD gained 10 archives in winrar. Intel is definately creeping andrewin, albeit slowly. If we, or anyone else can figure out why the intel chip has a heartbeat in the graphs, and rectify it, the creeping would become crawling.


[/quote]

..................CD's .............RAR...............FC.............DIVX

AMD...........143.................1188............702..............40
Intel............130................697...............478............620
difference....-13..................-491.............-224...........+580
.........gained 0...........lost 16..........gained 6.........gained 50

[/quote]

same pattern emerging...Farcry has dropped down and held at 27fps for about a half dozen refreshes....CD encoding may have taken the boost...So evidently thread priority may change around a bit....

..................CD's .............RAR...............FC.............DIVX

AMD...........154.................1272............756..............40
Intel............143................769...............523............680
difference....-11..................-503.............-233...........+640
.........gained 2..........lost 28..........lost 3.........gained 110
from last time
.........gained 2..........lost 12..........lost 9.........gained 60
now down to 24fps....

 

[Keysplayr]

Originally posted by: bREwIn dOdgEr dUDe
after a couple o' days of seeing the last gasps of a (very) proud few, and alot of reading, flaming, chest-thumping & such... will someone please put a fork In(tel) and call this ThG stress-test somewhat "well done"... or as they say in japan... c-ya?!?...

I feel your pain, but nobody says you have to hang around if you don't want to.
Besides, I don't think it's anywhere near done.

 

[Keysplayr]

Originally posted by: Duvie

Originally posted by: keysplayr2003
Next refresh

..................CD's .............RAR...............FC.............DIVX

AMD...........130.................1104............658..............30
Intel............117................629...............428............560
difference....-13..................-475.............-230...........+530

don't think INTEL chip is "creeping up", except for DIVX

I think it is in everything except WinRAR. Since my last post, Intel has gained 2 CD encodes, 3 Far Cry runs and 30 additional minutes in Divx. AMD gained 10 archives in winrar. Intel is definately creeping andrewin, albeit slowly. If we, or anyone else can figure out why the intel chip has a heartbeat in the graphs, and rectify it, the creeping would become crawling.

..................CD's .............RAR...............FC.............DIVX

AMD...........143.................1188............702..............40
Intel............130................697...............478............620
difference....-13..................-491.............-224...........+580
.........gained 0...........lost 16..........gained 6.........gained 50

[/quote]

same pattern emerging...Farcry has dropped down and held at 27fps for about a half dozen refreshes....CD encoding may have taken the boost...So evidently thread priority may change around a bit....

..................CD's .............RAR...............FC.............DIVX

AMD...........154.................1272............756..............40
Intel............143................769...............523............680
difference....-11..................-503.............-233...........+640
.........gained 2..........lost 28..........lost 3.........gained 110
from last time
.........gained 2..........lost 12..........lost 9.........gained 60
now down to 24fps....[/quote]

Do you think this means that although four full blown apps is impractical on a desktop, would these XE's be good for application servers and database servers? They get hammered seven ways from Sunday. And, would each app/data fetch request a separate thread? Or is that not how it works. Just curious.

 

[Duvie]

I dont know about that stuff...


QUESTION:

NOtice how the AMD system is getting network hits and the P4 is not.....Why is that...And if so does this not take cpu cycles from the AMD system???

Edit: intel system is not connected...the network on the AMD system is uploasding and download and I am certain that is stealing some cycle...How much I don tknow but all of it counts even if it is 3-5%.....

 

[SlitheryDee]

Originally posted by: Duvie
I dont know about that stuff...


QUESTION:

NOtice how the AMD system is getting network hits and the P4 is not.....Why is that...And if so does this not take cpu cycles from the AMD system???

Edit: intel system is not connected...the network on the AMD system is uploasding and download and I am certain that is stealing some cycle...How much I don tknow but all of it counts even if it is 3-5%.....

I noticed that a while back.

This is one more thing that is different about the two systems. Different PSU, MBs based on wildly different chipsets (I can sorta forgive that one, given the nf4 SLI problems), and now apparently one is connected to a network while the other is not. Hasn't anyone at THG ever heard of controls?

I mean the purpose of this test is to ascertain the relative stability of the processors right? That's kinda hard to do when there are so many other different components in the system that could now be the culprits.

THG has deviated so far from standard methodology in this test that it's laughable.

 

[AnonymouseUser]

Originally posted by: Duvie
I dont know about that stuff...


QUESTION:

NOtice how the AMD system is getting network hits and the P4 is not.....Why is that...And if so does this not take cpu cycles from the AMD system???

Edit: intel system is not connected...the network on the AMD system is uploasding and download and I am certain that is stealing some cycle...How much I don tknow but all of it counts even if it is 3-5%.....

Well, they have to send their data somehow, so there has to be network activity, but that doesn't explain the lack of any on the Intel system.

 

[Keysplayr]

Originally posted by: AnonymouseUser

Originally posted by: Duvie
I dont know about that stuff...


QUESTION:

NOtice how the AMD system is getting network hits and the P4 is not.....Why is that...And if so does this not take cpu cycles from the AMD system???

Edit: intel system is not connected...the network on the AMD system is uploasding and download and I am certain that is stealing some cycle...How much I don tknow but all of it counts even if it is 3-5%.....

Well, they have to send their data somehow, so there has to be network activity, but that doesn't explain the lack of any on the Intel system.

Well, is it possible that the ASUS mobo has dual nics and the graph is pointing to the wrong one? Whats the model number of that board? Arghh now I have to search. BRB

 

[Keysplayr]

Yup. It has Dual Gigabit LAN controllers.
ASUS P5WD2 Premium

So the graph may have the inactive NIC selected. An oversight on THG's part I guess or the NIC is not selectable. In which case, all they have to do is plug the Cat5 cable into the other NIC port to get readings.

 

[Keysplayr]

Nite Folks. Looking forward to seeing tomorrow's events.

:moon:

 

[SlitheryDee]

The intel system has been up for more than 12 hours now. The EKG er... cpu load graph seems uh... regular. Temps are stable.

Maybe the new psu/hsf is making a difference.

 

[Keysplayr]

Yup, looks good. The Intel system has switched its priority from FC to CD encoding. It's almost even with the AMD rig in Lame Encoder. Falling behind in WinRAR and FC though.

Yeah, I'd love to know what that is about. The EKG thing I mean. Tons of cycles are lost in those dips. Add them all together and the time lost is huge.

I wonder how Intel can justify asking for over 1000.00 for their 3.2 XE when all it has over the regular Pentium D is HT. And I guaranty that HT is only disabled on the Pentium D's. Kind of insane premiums. I am actually ordering 2 Pentium D 830's and 2 ASUS 955X boards this morning from the EGG. I need to actually run a stress test on these 2 systems for me to ok the purchase for our companies products. Pretty kewl. Too bad the X2's are not available yet otherwise I would get 2 sets of those as well, but the gouge prices may be too steep.

 

[Duvie]

..................CD's .............RAR...............FC.............DIVX

AMD...........154.................1272............756..............40
Intel............143................769...............523............680
difference....-11..................-503.............-233...........+640

..................CD's .............RAR...............FC.............DIVX

AMD...........194.................1593............955..............40
Intel............192................1037...............672............900
difference....-2..................-556.............-283...........+860
.........gained 9..........lost 53..........lost 50.........gained 220

As of 19:36 hours of uptime......


..................CD's .............RAR...............FC.............DIVX

AMD...........200.................1648...........988..............40
Intel............200................1082...............697............930
difference....0..................-566.............-291...........+890
.........gained 2..........lost 13..........lost 8.........gained 30

As of 21:00 hours of uptime....

so Intel is gaining 2 cds per 1-1/2 hrs.....In 24 hours from now it will have 16 more cds done and a calculated 5.4% lead...

IN winrar it will have ~208 less packed archives then now for about a 40.3% deficit

IN Farcry it will have ~128 less runs then now for about a 38.1% deficit...fps are a 33% deficit, so the rest is likely do to the Intel downtime

the Divx will just contiunue to run at a 250:10 pace






ead




I think it was heat and not power, personally...good luck anybody ocing these things short of some really high end air or water.....getting screenies is one thing, actually working is another....



Notice how the Farcry FPS have never come back...interesting it just changes midstream...Intel should take the CD test in about 1 hour or less....

Also notice how the fan speed has gone up to 4400 when in the fist 12 hours it never hit higher then 4100rpm...Temps have droped about 1.5-2c...cmon ppl.....I think someone is adjusting the speed!!! (wink) Clear sign this was about heat...Power on the mobos may have been an issue with a few but with this asus mobo it was likely heat...

 

[Duvie]

I think are related to when apps finish and likely the scheduler to HT has to work a bit more to reearrange things.....

It may be more of when apps coincide finishing.....since each of the physical core can likely switch on their own I think it is the fact that the virtual cores are not real physical cores that causes this.....I can t explain it in technical terms but the rate or period of this "hiccups" is to patterned to be anything but a normal reaction....

It would be interesting to have looked at a single core P4 with HT running 2 apps and see if same thing happens.....

 

[Duvie]

DUVIES ROUNDUP (6 days completed of the test)


so overall roundup is.....

2 round of test since they threw out AMDs uptime (no reboots form Friday morning to like Tuesday morning).....At that time Intel system had 4-5 reboots.....

now since the start of the second round they have had 3 more reboots and a complete shutdown to switch out parts...

they themselves have gone through 5 mobos....The used Nvidia chipsets for the first 3 but INtel chipsets have been the last 2...one of the Aus mobos would not post anymore and the second one the supposedly stable i955 is responsible for the 3 crashes of the latest round....

The resolution so far has been adding a different (non stock non retail cooler)....It was running at 4400 rpm but in the last 3 hours has mysteriously sped up consistently (and holding) another 200-300rpm and temps are now in the 67 range.....It was flirting with the 70s the first round that had intel machines crashing every 12 hours. When it was swapped to asus intel board in the second round temps were in the high 80's and it was crashing evry 3-4 hours....Now new hsf, lower temps and we see staility...had to be heat in my estimation.....power cap on the epox board for sure though....


INtel will win 2 out of the 4 test if they can stay stable, but I question Toms results since they differ from his first round when before the first rebooting spell of the INtel AMD was solidly leading and pulling away in 3 apps....now since the second round when the INtel chip is stable it will always be gaining in 2...now platform changes can account for this but not to the precentages of change...i smell a rat. I have said 4-5 days ago problem with Toms reviews is they differ from other repuable sites and they can never seeme to be reproduced...apparently even by him.....


I would like to see someone known to do QUALITY reviews, and seen as a bit un-bias in the past to rerun a review like this to see if numbers conincide.....Maybe INtel still win 2 out of the 4 but I wonder more about the percentage of each...


BOTTOM LINE:

AMD system has been up 140+ hours (no reboots)

INtel system has been up 130+ hours (8 reboots at least) ( 5 mobo changes)


I think that sums up the stress test portion of it....


UPDATE:

Also now that the iNtel is running an upgraded fan (non retail, non stock) in the same rpm range of the AMD and noticeably larger in diameter so meaning it is likely pushing more cfm, I wonder what an AMD aftermarket 80mm fan like even my Venus 12 would do...I bet I dont even hit 50c under load. A nice copper core hsf with an 80mm-90mm fan like many use and this thing is mid to upper 40's at stock vcore and stock speed....That would meean it is almost 20c better then the Intel system....

Currently AMD system on average is 12c cooler but using a stock cooler.....

 

[Duvie]

Keys watch it but I think the blip correspomds with the finishing of a CD for sure....It uaslo corresponds to the min blip in between so I think the larger blip is 2 program finishing simultaneously....This is the thread scheduler likely doing things here once apps end and start up again....kinda like a bunch of planets all aligning and then amplifying the effect.....It rythmic to say the least and since performance numbers semm to be holding pretty close the programs should also be rythmic...

 

[GuitarDaddy]

Even though the testing methods are totally flawed as far as performance is concern. THG stated the purpose of this test as a stability test, not a performance test. But they obviously are trying to show performance characteristics also. But unfortunately because of the flawed methodology the performance numbers mean nothing.

As a stability test it is a monumental failure for intel and a solid victory for AMD. It will be very interesting to see how THG summarizes this test. Because if they stick to their stated objective of stability measurement it will be very difficult to put any kind of positive spin on Intels failure. My guess is they will just end the test and say nothing, or make some statement like "We will attempt this again, once both platforms are more stable"

If they do what seems unthinkable for THG and publish their conclusions slamming Intel for the total instability of the platform, I might have to frame it and hang it on the wall

 

[rgreen83]

Thanks for the summary duvie, wonder why they didnt use the DFI in the AMD?

And werent they supposed to both use PP&C 850w PSU? The one in the intel system is not a PP&C unit, it has a fan on the bottom.

Also, the fan I see now in the intel system is one that I have seen in retail packages recently, the round "bifurcated" design.