Theorists to the rescue?

[alienware]

I've been working with computers for twenty years and to this day I have never seen anything quite like this. I am stumped.

About four months ago a friend of mine in the U.S bought himself a Phenom 9950 black and Foxconn Destroyer (SICK !) from Alienware when they were ridding stock. He paid about $220 which was very cheap (the board retailed at too much. Way more than $220).

When it arrived he put it together and it wouldn't boot at all so he sent back the motherboard.

He got the board RMA but decided to go with a Crossfire capable board (as by then he had ordered a 5850) and got an Asrock A785GXH. It came and he built it. Immediately it started to reboot and shut down. Even when he ran something like Windows rating. After hours of messing around we put in an Athlon X2 BE he had (forget the model but it was 2.8ghz) and it sprung to life becoming stable.

Fast forward to today and I have the CPU here (he sent it). I have fitted it in my H50 cooled Crosshair 2 ROG and it's stable enough. It hasn't rebooted, hasn't shut down and I am seeing 30c idle with 42 max when giving it a thrashing in Orthos.

Before we assume it's fine though and put it all down to him it isn't. On the very very odd occasion Orthos will fail. However, I am seeing no temp spikes, no shut downs and no reboots. I simply rerun Orthos and it carries on where it left off.

Things to bear in mind...

I checked his Asrock for 140w CPU (which this one is) support and it states 100% compatability.

The Foxconn Destroyer was indeed buggered and replaced under RMA and he sold it for $240.(so it wasn't that).

The 9950 is not overclocked , overvolted or anything else.

Any theories as to what the hell is going on? and are there any other tests I can run to try and nail exactly what the heck is going on here?

As I have said, in my board under my cooler it seems to be fine. Orthos does fail from time to time but Fallout 3 ran without a hitch and I got no less than 30FPS min. There have been no temp spikes either.

 

[alienware]

Does any one know how an AMD cpu is made? Is the top layer (the alu looking stuff) simply a peel off thing? (like some kind of heat spreader?)

 

[Diablo]

You can remove the heat spreader, but don't. You risk knocking bits off the PCB, and worst comes to the worst, you might take a chunk out of the core.

The temps aren't the problem, perhaps the memory or memory controller is fubared?

 

[alienware]

Excellent ! That explains the mystery and I think I have it solved. I wasn't going to remove it I just wondered if it would be a problem if it was dented.

When he ordered the parts for the machine I found him a cooler that was on one day sale on Newegg.com (he's American, old friend of mine from when I lived there). The cooler was a Vigor Monsoon III and came with dual 120mm fans. IIRC it was almost half the price of the Noctua 12f thing.

Well when he came to fit it he was having serious troubles (I was on Skype with him). Basically after a lot of cursing it went on but as soon as he tried to run even Windows rating the machine would shut down. We removed everything and in the end ran Speedfan and saw that one of the cores was hitting 130c right as the machine shut down. He did fit the stock cooler back on but to no avail, it was still buggered. He removed it and put a Athlon X2 in his system (2.8ghz black ed) and it's been fine.. The Phenom was a mystery though.

This has bothered me for months since because it's very rare that I have ever seen a bum CPU. I mean yeah, it happens, but rarely. I asked him to send me it (he sent me a Xonar D2X that was surplus too) and it arrived today.

Immediately I noticed that the corners of the CPU were pressed in. So here where you see this bit it was kinda pushed in.

I pressed it straight with my fingers as best I could and put it in my Asus with my H50. Now I know the bottom side of the H50 is copper so this would allow some give as opposed to hard alu or billet.

The first few Orthos passes flagged errors but the PC was perfectly stable as were the temps (44 was the highest I went after two hours of Orthos). No restarts, no shut downs nothing.

I've run 3Dmark06 now three times and each time it passed with flying colours so I would imagine that the early Orthos failures were down to the CPU still bedding. It's fine now.

I think that ghastly cooler warped the surface of the CPU so his heatsink wasn't making contact.

Oddly I went off and found a Youtube channel for the guy who makes the Vigors and challenged him over his TERRIBLE manual. The guy can't even spell or hire some one to spell the manual. It was full of spelling mistakes and makes no sense AT ALL. I'm not the only one to complain about this either and I would imagine Newegg wanted rid of them hence the low price.

Absolute junk.

 

[noobieocer]

the cpu is fualty/bad chip, some cpus just wont even run under stock. if orthorus is failing even once in a while, then the cpu is not stable and if its not overclocked then its a really bad chip and you need to rma it. also orthorus only stress 2 cores, so not even the full 4 cores was bieng tested. try using prime, but really u need to get it rma'd

note: just becuase nothing crashes dont make the cpu "stable", it is stable when theres no errors when its bieng used like in orthos or prime.

 

[alienware]

I ran Orthos for over two hours earlier, solid as a rock.

Will run Prime later....

The CPU cannot be RMA because it's out of warranty. Also, if it is screwed (which I highly doubt now) then it would be down to damage whilst being fitted. Something AMD won't cover.

 

[Diablo]

Try using intelburn test, (linpack testing), that will spot a problem in about 5 minutes normally, although make sure you have excellent cooling (this will get the chip 10C hotter than prime)

 

[alienware]

OK Prime95 runs without a hitch (30 mins and counting and 51c). Will leave it running.

 

[AMDFTW]

try OCCT that will tell you if you have had errors on the cores during the test,run on auto

 

[alienware]

OK I have now noticed a recurring theme. If I put on the test that uses the ram I get errors. I have long suspected that this ram is duff given that I got it free. I have been running Prime95 for 2 hours now on the CPU only test (well, some ram) and it is still going strong.

I don't think there was ever a problem with this CPU apart from an ill fitted cooler.

I did pause Prime95 to play Fallout3 and played it for an hour with no problems at all. So tbh even if it does flag errors in Prime or Intelburn (ran that too and it was fine for 25 mins) then the CPU is more than good enough to do what it is supposed to (run Media Center and play Fallout 3).

It's not my main computer and never was so as long as it performs those functions then it's fine.

As I say, the ram was given to me (2gb Corsair XMS) and I have long suspected that it's a bit Donald (Donald Ducked) because when it was in here (my gaming rig) I used to get restarts from time to time for no reason (without a Bsod or anything).

Any ideas on a good app to test the ram?

 

[AMDFTW]

memtest

it takes ages tho

 

[Diablo]

Yeah, make a bootable cd, then leave it on overnight. As you only have 2GB, it shouldn't be as bad. about 10 passes miniumum.

 

[alienware]

Thanks Diablo. And thanks to all for your help !

I have decided to wrap up the CPU with an air of trust. Basically if I run Orthos it's fine. If I run Prime95 with the blend test it bombs accross random cores. If I run Prime95 on the first thing (FPU hardly any ram) then it runs fine. I literally had it running all night and for some of that had Media Center running with it so I could watch TV. When I awoke thismorning it was still hammering away, CPU at 53c and me a sweaty mess

Intelburn states what it does. I am going to thrash your PC. And due to that I would imagine it leans on the ram as once again I get a failure.

Last night I played Fallout 3 for a couple of hours in the end (I'm on my fourth playthrough with all DLC enabled) and wanted to get to the ending and beginning of Broken Steel. It was absolutely fine (apart from the D2X giving me the dreaded looped sample crash). As soon as I disabled DX on the Asus Xonar the PC ran like a champ.

Sooooo, before I become obsessed with proving this CPU foul I think I will leave the CPU tests there. The silver machine that I have put the Phenom in isn't relied on very heavily. I literally switch it on in the late evening before I retire to bed and it puts itself to sleep when I have fallen asleep and I wake up and shut it down in the morning. That's literally all I do with it apart from playing Fallout 3 on it from time to time when I don't want to sit at this PC (my main one).

I will definitely check out the ram though and time is not a problem as I have a PC sitting within feet of it that I use most of the time (this one).

So thanks again ! All very useful info for the future no doubt

 

[Diablo]

Yeah run memtest, see how that does. If it is dodgy ram, corsair has a lifetime warranty, so just get on and RMA it, it takes about a week normally. Also the D2X audio loop is the reason I gave it up and got a creative. Sound isn't quite as crisp, but on the other hand, it doesn't crash with the frightening frequency and make that annoying dugadugaduga.

 

[alienware]

Yeah I have found Corsair to be very good tbh. Especially Ramguy, a true gent.

I made the mistake (?) of buying an X-32 just before Christmas and hadn't done my research on TRIM and so forth. God, the wars that kicked off on Corsair's forum over a bios flash !

In the end my patience paid off (I had the secure erase unlock and restore with Acronis nailed any way within 15 mins) and the bios flash worked for me.

But just watching their RMA happen as people bricked their drives was quite cool.

I wasn't aware the ram was lifetime, so thanks for that. Bit of a result really

 

[thestepster]

the corsair RMA is flawless they are an excellent company to deal with yellowbird was a great help