AMD Processors
Decrease font size
Increase font size
Topic Title: Really bad quality of assembly
Topic Summary:
Created On: 06/03/2008 01:26 PM
Status: Read Only
Linear : Threading : Single : Branch
1 2 Next Last unread
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 06/03/2008 01:26 PM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

Sorry for the russian, look on the processor pictures after dismantling.
http://www.riddik.net/publ/1-2-1-25

thats really bad for processor cooling. Why thats possible?
 06/03/2008 11:02 PM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

Why?
Can you explain what you're saying please. It's not clear.
 06/04/2008 04:00 AM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

Look on the picture:


this is hardware problem (assemling - more correct)
Processor was demastling manually to determine why cooling work so bad.
Cover of the processor was soldered to the core by special substance (tin based solder may be)
Look on te BUBLES. That is poor and not effective for heat distribution.
I guess that is prolem for all AMD Phenom processors, and not Phenom may be.

Thanks.
 06/04/2008 04:18 AM
User is offline View Users Profile Print this message

Author Icon
ChristianJ
Member

Posts: 44
Joined: 01/10/2008

Hmm..not quite sure what you mean. That's what I would expect to see when ripping apart a processor and it's heatspreader. I'd guess your standard heatsink and CPU don't look very different if you separate them after some use.

Also, the cooling seems to be the only thing that's working quite fine and stable in my computer..so I'm not sure that's the problem with Phenoms...

-------------------------
ASUS M3A32-MVP Deluxe WiFi (BIOS 1002) w. AMD Phenom 9750
XFX 8800 GTS
4x1 GB Kingston HyperX 1066 RAM (2.2v)
1x Seagate 7200.10 160GB SATA2
1x Samsung 250GB SATA2
Gigabyte Odin 550W PSU
Lian Li PC-7 SE-B III
Windows Vista Home Premium 64bit
 06/04/2008 05:35 AM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

"I'd guess your standard heatsink and CPU don't look very different if you separate them after some use. "
No, i am not agree. Thermal grease and tin solder is not the same - different melting point.
This processor never warming above 80 degrees Celsius.
 06/04/2008 12:07 PM
User is offline View Users Profile Print this message

Author Icon
MD - Moderator
Deployer of Mjölnir - House Keeping

Posts: 11102
Joined: 11/05/2003

I would assume they used a lead free solder paste to do that. I see what you are describing, the voids or "bubbles".

I am not sure what exactly would cause that, I could speculate if there was some sort of flux that did not fully evaporate or a contamination present for the paste to not adhere.

Do not take that as the final explanation, I do work in a similar industry and we do similar work, but we have not run into anything like that in our processes.

We X-ray our final assemblies to make sure proper soldering takes place.

-------------------------
The opinions expressed above do not represent those of Advanced Micro Devices or any of their affiliates.

Physics? Ha! This is clearly magic and devilry at work. Prepare firewood! We have witches to burn!


MODERATOR
 06/04/2008 04:00 PM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

I don't understand "bad cooling" at all. They don't heat up more than they should, all around I've seen don't. So I don't understand the premise to begin with.
Secondly, I've seen four others taken apart and they were not like this, no bubbles were there.

Are you sure you never fried your processor before taking it apart? It looks like you overheated it.
 06/04/2008 04:13 PM
User is offline View Users Profile Print this message

Author Icon
magdary
Case Modder

Posts: 1228
Joined: 09/20/2007

If the two halves are soldered together then surely they would have to have a good deal of heat applied to split the two halves and couldnt this have caused the bubbling if it had been heated for too long.
 06/04/2008 04:25 PM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

As they say in this page (http://www.riddik.net/publ/1-2-1-25 ), it was not possible to cool the processor below 70 degrees Celsius.
The person also approves that did not overheat the processor.

It is not mine processor fortunately.
I only wanted how much qualitative copies are delivered to Russia. Impressions not so good.

Can you show me any disassembled Phenom? (any links? photos?)
 06/06/2008 01:12 PM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

Nevertheless i would be desirable to trust in AMD.
If it will not be, Intel again will start to sell Celeron's for 1K dollars and progress in development of processors will stop.
But in Phenom there is no sense to trust.
 06/06/2008 09:08 PM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

Originally posted by: Disappear

As they say in this page (http://www.riddik.net/publ/1-2-1-25 ), it was not possible to cool the processor below 70 degrees Celsius.
Oh ok. That was a bad sample then I suppose. Nearly all the rest are not like that, they are full load below 50°C and sometimes below 40°C on good air cooling. You can search IT forums and see this and I have personal experience of more than a department full of Phenom quad-cores to testify.

Can you show me any disassembled Phenom? (any links? photos?)

Yes. Hold on.. let me find one and I'll get back to you. I'm busy with some projects right now.
 06/07/2008 02:36 AM
User is offline View Users Profile Print this message

Author Icon
brane212
Member

Posts: 125
Joined: 03/24/2008

I could not prevent either of mine 9850s to heat over 60?C when under load, so it is apparently not just some "bad sample".

And even if it was, how come AMD couldn't test chips for these kind of errors, especially since tolerances are so tight ?

9850 is declared as 125W, just like my 6000+. Difference is, my 6000+ has more effective cooler and 6000+ can work at temps over 60?C and 9850 can't.

It is incomprehsible how AMD didn't care to check the finish on the lid and cooler. Both have significant curvatures.
 06/07/2008 04:44 AM
User is offline View Users Profile Print this message

Author Icon
Disappear
Junior Member

Posts: 6
Joined: 06/03/2008

In this case Phenom Black Edition was used. There can be it matters.
 06/07/2008 05:53 AM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

Disappear, check here, AMD K8 and Intel Core 2 are also shown: http://www.xtremesystems.org/f...owthread.php?t=183253

Originally posted by: brane212

I could not prevent either of mine 9850s to heat over 60?C when under load, so it is apparently not just some "bad sample".

First of all, 60°C is not bad or outside the specification but is entirely normal.

However, it all depends completely on these few parameters:

What motherboard were you using?
Which program were you using to poll the temperatures?
What cooling were you using (fan/heatsink/TIM)?
What are your ambient heatsink temperatures?
What was your load voltage?
What temperature was showing idling with 9850 BE at stock with CnQ activated?

60°C 65nm 125W is still not hot. The 65nm QX6700 ran plus 70°C stock under load air cooling.

And even if it was, how come AMD couldn't test chips for these kind of errors, especially since tolerances are so tight ?

Which errors exactly? I really am not sure what you are saying but I'll clarify.

Of course AMD and Intel all test their processors very rigorously. They do not release untested processors but their are many phases after the testing where things can go wrong. Please read about some of the testing facilities and procedures thoroughly and then you'll have some sort of knowledgeable understanding fo what it takes:
http://www.amdzone.com/index.p...ide-amds-building-312
http://techrepublic.com.com/2346-10878_11-9356.html
http://video.gearlive.com/vide...-amd-performance-lab/

This isn't a $1000 firm we are talking about to give you knock off equipment on the sly but one dealing and worth $1000000000's. Children errors are not minutely tolerated at even semi-professional small businesses never mind world recognized highest class businesses that make The Global 100 Most Sustainable Corporations in the World list: http://www.global100.org/2008/index.asp



9850 is declared as 125W, just like my 6000+. Difference is, my 6000+ has more effective cooler and 6000+ can work at temps over 60?C and 9850 can't.

What makes you say 9850 BE can't run over 60°C?

It can run at and above 60°C very easily and it does safely with many users. One of the first tests we have to do at work whenever we bring in new systems is to see if they can handle summer ambient temperatures. Summer ambient temperatures at 30-35°C which is 45-50°C within the case. All our samples completed testing perfectly on stock coolers and yes, they were all above 80°C after 48 hours load with those case ambient temperatures. 60°C is not even close tow hat they can tolerate even though your 125W processor is 2 cores and these are 4.

They are rated for 115.5°C as you see in software, 60°C is far below what they can handle.
 06/07/2008 06:57 AM
User is offline View Users Profile Print this message

Author Icon
brane212
Member

Posts: 125
Joined: 03/24/2008

KAB:

First of all, 60°C is not bad or outside the specification but is entirely normal.


It is UNACCEPTABLE if your CPU max allowable temp is 60?C internal protection inside your CPU triggers at 61?C.

What motherboard were you using?


Gigabyte GA-MA790FX-DS5 with latest BIOS (F6 IIRC)

Here is link to specs:

CLICK

Which program were you using to poll the temperatures?


Program was a gadget inside GNOME which was using kernel infrastructure to do this ( nf-something, I can check it out if its relevant).

What cooling were you using (fan/heatsink/TIM)?


I used what I've got in the package with CPU. Strictly as it was meant to be used, unmodified.

What are your ambient heatsink temperatures?


Well, my room is in the attic and case is very well ventilated.
Sensor on the board shows now 30?C and it probably was something around that with Phenom. I don't place big confidence in this reading since it is very dependent on airflow path etc.

I've got free-running temp 38?C with BOX cooler, which went over 60?C under load.

What was your load voltage?


Strictly default. No overvolting, overclocking etc.

What temperature was showing idling with 9850 BE at stock with CnQ activated?


I've got free-running temp 38?C with BOX cooler, which went over 60?C under lload. It wasn't with fully enabled CnQ, since Linux kernel 2.6.25 doesn't have it yet- it is included in 2.6.26. IIRC CPU was idling at 1GHz and 1.07V according to readings...

60°C 65nm 125W is still not hot. The 65nm QX6700 ran plus 70°C stock under load air cooling.


Yes, but QX can run hot. Phenom obviously can't. It faints so hard that it needs 10 minutes or so to be able to POST again...

Which errors exactly? I really am not sure what you are saying but I'll clarify.



How about cooling system capabilities ? Stick CPU in test socket, press on top its cooler, start some CPU intensice program asnd see what happens ?


Or maybe checking that cooling surfaces are actually plain ?

If I can press CPU against cooler, look at it againsta backgund lam and SEE THE LIGHT COMING BETWEEN CPU AND HEATSINK how hard it is for AMD to figure that physical tolerances are WAAAAY out of specs ?

What makes you say 9850 BE can't run over 60°C?


Well, readings for pone thing. 10 seconds after temp reaches 60 it dies.

Specs are another. After my initial complaint on some forum someone has linked Phenom's data ( IIRC it was a Whitepaper) and it clearly listed max temp as 60?C

It can run at and above 60°C very easily and it does safely with many users.


Well, maybe AMD is selling better stuf in US and we get borderline stuff, but NONE of my Phenoms can cross 60?C...
 06/07/2008 07:14 AM
User is offline View Users Profile Print this message

Author Icon
brane212
Member

Posts: 125
Joined: 03/24/2008

What makes you say 9850 BE can't run over 60°C?


Yeah, I managed to find link to AMD's data that states this:

IN_AMDs_OWN_WORDS
 06/07/2008 10:29 PM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

Originally posted by: brane212

KAB:


First of all, 60°C is not bad or outside the specification but is entirely normal.




It is UNACCEPTABLE if your CPU max allowable temp is 60?C internal protection inside your CPU triggers at 61?C.


Do you have Sisoftware Sandra installed? Can you install it please and I'll show you what your maximum tolerated temperature is. You'll need Windows, no temperature support for Phenom in Linux at all yet.

Let me clarify for you.

60°C is not the maximum temperature you are seeing in software nor does it trigger internal safety protection. The temperature you see in software is not even in °C. Software does not show Phenom temperatures, AMD BKDG mentions this very clearly. The value you see has a correlation to temperature though and is known as 'TControl' or 'Tctl'. The maximum tolerated temperature you can fetch as a hex value from the core registers of Phenom and for 9850 Black Edition it is 115.5°C. That is what AMD has coded manually and that is where the safety limit is set before processor is damaged and processor shuts down using the safety circuitry. It will never shut down at 60°C or even 80°C in software due to overheating, not until you are topped 115°C. If it does, you have some other problems elsewhere. I hope that's clearer now.

Program was a gadget inside GNOME which was using kernel infrastructure to do this ( nf-something, I can check it out if its relevant).

Oh OK. The latest lm-sensors which is how any tool in Linux fetches the temperatures, does not correctly detect Phenom temperatures, it is usually reading them 10-15°C higher than the core reports. Was the gadget called Sysmon?

Well, my room is in the attic and case is very well ventilated.

Sensor on the board shows now 30?C and it probably was something around that with Phenom. I don't place big confidence in this reading since it is very dependent on airflow path etc.

I've got free-running temp 38?C with BOX cooler, which went over 60?C under load.

If motherboard sensors shows around 30°C your heatsink ambient temperature can't be high, I would say near 25°C minimum though and as a base rule, Phenom will be at least 10-12°C higher idling. Under load it will increase more but for those ambient and idling temperatures, I didn't expect it to top 60°C. Is your motherboard overvolting the vcore under load? Have you measured this?

I've got free-running temp 38?C with BOX cooler, which went over 60?C under lload. It wasn't with fully enabled CnQ, since Linux kernel 2.6.25 doesn't have it yet- it is included in 2.6.26. IIRC CPU was idling at 1GHz and 1.07V according to readings...

Linux doesn't support CnQ for Phenom yet, so it was running full speed and full voltage.

Yes, but QX can run hot. Phenom obviously can't. It faints so hard that it needs 10 minutes or so to be able to POST again...

Not really, no.

How about cooling system capabilities ? Stick CPU in test socket, press on top its cooler, start some CPU intensice program asnd see what happens ?

There's no error in that case.

Or maybe checking that cooling surfaces are actually plain ?

It cools as intended, again, there's no error here.

If I can press CPU against cooler, look at it againsta backgund lam and SEE THE LIGHT COMING BETWEEN CPU AND HEATSINK how hard it is for AMD to figure that physical tolerances are WAAAAY out of specs ?

You can see light down there?
Have any pictures of this to show it? That'd be a very first to see...

Well, maybe AMD is selling better stuf in US and we get borderline stuff, but NONE of my Phenoms can cross 60?C...

Where are you based? I was talking about US and UK based systems.

Can you post a screenshot of your Phenoms under load with a temperature monitoring program running. Even though none of the applications in Linux gather Phenom temperature anywhere near accurately.
 06/08/2008 04:56 AM
User is offline View Users Profile Print this message

Author Icon
brane212
Member

Posts: 125
Joined: 03/24/2008

Do you have Sisoftware Sandra installed? Can you install it please and I'll show you what your maximum tolerated temperature is. You'll need Windows, no temperature support for Phenom in Linux at all yet.


Which is more credible source Sisoft Sandra or AMD's own document that I gave you the link to ?

60°C is not the maximum temperature you are seeing in software nor does it trigger internal safety protection. The temperature you see in software is not even in °C.


That is practically always the case. Unless you have some smartsensor ala DS1820 or somesuch you will always get reading of soem other quatinty ( tiome, freq, here usually voltage) and will have to use tables or formula to get temperature from that.
This is nothing new.

I don't really trust 100% that the temp I get is really accurate, and I don't care.
What I do know that crashes are temperature related.

I care that it took me 5 days installing, reinstalling and troubleshooting the machine after CPU+MOBo swap.
I do not really care whether CPU can actually take 90? and it was because of the poor cooler or maybe bad thermal patch or if Phenom is actually incapable of working over 60?C.

It's really not relevant for me. It can work at 900?C and be cooled with liquid sulphur for all I care. As long as it works - which it doesn't.

You can see light down there?
Have any pictures of this to show it? That'd be a very first to see...


I know what you mean. I was really shocked too. Too surprised to think about taking the time to take a picture or two and post short article on Slashdot asnd maybe something on youtube.
Not to worry though. I have to write an article about Phenom, so after having first Phenom DOA( it wouldn't even fart in the board), its replacement being unable to work under load, buying the third one just to see the same result, I am about to BYU ANOTHER BOARD - with another, FOURTH Phenom.

IN best case scenario, I would be getting one for the price of 3+1.
What a deal....

Where are you based? I was talking about US and UK based systems.


Slovenia, EU.

Can you post a screenshot of your Phenoms under load with a temperature monitoring program running. Even though none of the applications in Linux gather Phenom temperature anywhere near accurately.


All my Phenom unit is dead. After failing to perform in regular services I sent them to serve their country on special missions. Their death was swift and glorious. Their bodies were examined and will be displayed on my wall of fame...

Until next week, when new recruit is coming...
 06/08/2008 04:58 AM
User is offline View Users Profile Print this message

Author Icon
brane212
Member

Posts: 125
Joined: 03/24/2008

Is your motherboard overvolting the vcore under load? Have you measured this?


Well, IIRC SW was showing 1.07V when system is idle on 1GHz and 1.3V when working at 2.5GHz...
 06/08/2008 09:22 AM
User is offline View Users Profile Print this message

Author Icon
Kab
Senior Member

Posts: 1349
Joined: 02/03/2007

Originally posted by: brane212

Which is more credible source Sisoft Sandra or AMD's own document that I gave you the link to ?

Maybe you misunderstood me.
What you are calling 60°C temperature is not, I repeat, is not temperature. It is just an arbitrary value named Tcontrol to represent temperature indirectly. The program I referenced retrieves information programmed directly into the processor itself by AMD like clockspeed and name is (registers). It follows the best documentation and most authoritative accurate technical information about Phenom from the AMD BIOS Kernel Developers Guide and that has all the information I've posted. There is no source higher than this, all your BIOS and accurate applications are made using it.

You can read it here and it'll explain everything including temperatures: http://www.amd.com/us-en/asset...d_tech_docs/31116.pdf

What I do know that crashes are temperature related.

Now this is where we need evidence and I'll explain why. The internet is full of people topping 60°C and even 80°C with Phenom processors without a hiccup, I own a department full of them since November without a hiccup, I've provided my nephew at least eight Phenoms and quite a few motherboards to test that never had any such problem, so to say 60°C causes it problems will need a lot of testing and evidence before you can assert you have a faulty chip and hence, return it to AMD. Furthermore, I know from AMD that 60°C core temperatures in AMD OverDrive will not be a problem for Phenom if its working fine.

Firstly, you'll have to get into Windows or we can't know your core temperatures, at all. I would estimate 50°C is where they're at. Linux, Debian, Gentoo, Ubuntu variants just don't support Phenom sensors properly yet, although it is planned in one of the new kernel releases but lm-sensors developers don't have the required information from AMD yet.

Slovenia, EU.

Only been there once... is there a local AMD branch there?

All my Phenom unit is dead. After failing to perform in regular services I sent them to serve their country on special missions. Their death was swift and glorious. Their bodies were examined and will be displayed on my wall of fame...

LOL!

That would be extreme bad luck or extreme misdiagnosis if you have two dead chips in a row, nevermind four, honestly. Never happens, especially three or four times in a row and is very highly likely another faulty part you have, such as the mainboard BIOS or the RAM. If you have incompatible RAM, Phenom will find it very difficult to start.

Can you list which power supply you have?
Statistics
112018 users are registered to the AMD Processors forum.
There are currently 0 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.



Contact AMD Terms and Conditions ©2007 Advanced Micro Devices, Inc. Privacy Trademark information