Can memory controller problems kill RAM?

[Zap]

UPDATE #5 3/26

I'm wondering if the "crappy" Opteron has a faulty memory controller. Got in some Corsair XMS 2GB kit and it was working fine for a few hours, now one stick is dead. Mobo just beeps with it installed. I had it the whole time on the "crappy" Opteron at completely default speeds just to make sure the CPU was good at rated speed. My Geil RAM was exhibiting that strangeness with the "crappy" Opteron but is happy now with one of the other chips. However, one stick of the Corsair RAM is now dead. Tested using other Opteron in other motherboard and "beeeep."

UPDATE #4 3/17

1) Some people are getting too worked up over my "accusation" of Monarch.

2) I renamed the message title again.

3) My inquiry regarding whether the CPUs I've gotten from Monarch were used was prompted by the IHS looking different than I have seen previously and have nothing to do with how well or not well the cores themselves overclock. If I had posted "I'm building a workstation and am using a Tyan board" then perhaps people wouldn't be so quick to accuse me of being unhappy about getting bad overclockers, when indeed I had posted about the second core before I even put the chip in a motherboard for the first time.

4) At least nobody has said, "how did a retard like you become Elite" like they're fond of doing in Off Topic. :roll:

5) Also, at least math and spelling Nazis haven't shown up. When a thread dissolves into "OMG you left out an apostrophe... you suxor ..."

UPDATE #3 3/16

The two CPUs I received today (one looking new, one looking lapped) will pass Prime (just a few minutes) at 2.7GHz. Re-tested my original CPU will not even boot off disc at 2.7GHz and will fail Prime immediately at 2.4GHz. Basically I set up the new looking CPU, let it run Prime for a while, came back and saw that it was still passing, shut off system, popped in second "lapped" CPU I received today, booted up at same settings and ran Prime for a while, came back and it was still running, shut off, swapped to my original "lapped" CPU, turned on and it won't boot onto disc. Dropped multiplier to 8X and it boots fine and runs Memtest fine, but Prime fails on very first test.

Basically I was right in my assessment of the original CPU, that it overclocked like crap.

Something else I noticed is that this board doesn't like my OCZ memory at certain memory multipliers.

UPDATE #2 3/16

Decided to change the name of this thread.

I believe that two of the three Opteron 144 CPUs I've purchased as new from Monarch Computer have been previously used.

I looked at my original Opteron 144 from Monarch and it also looks like it was lapped. I had ordered it with a motherboard as an OEM CPU but had told them to NOT test it together, however when I received it I found out it was installed on the board with a Thermaltake Venus HSF already mounted on it. Well, after receiving the suspicious CPU earlier today, I cleaned the paste off the original Opteron and found out that it also looks like it was lapped.

My original CPU

First CPU received today

Second CPU received today

The pictures are from laying the CPUs on my scanner. Looking at the pics of the first two CPUs, the IHS looks lapped with some rounded corners and the texture is really smooth. The third picture (labeled "Second CPU received today") shows nice sharp edges on the corners of the core and more texture on the IHS surface. Unless someone has a good explanation as to why I've suddenly found two IHS that look like no other, my only conclusion is that Monarch Computer sold me two used Opterons as new.

UPDATE 3/16

Well, I got in my two new Opteron 144 chips from Monarch a few minutes ago. Just opened up the box and haven't tried them out yet, but... it looks like one of the CPUs have been used. Now, I've never tried "lapping" an IHS, but this one looks lapped to me.

I've seen literally hundreds of CPUs with IHS over the years (previously working at a computer repair shop) and I've never seen one like this. The corners are rounded and the sharp edges are softened a bit. Often the IHS leaves a rounded square "cut" on the bottom of heatsinks because they are so sharp and not quite flat, but this chip's IHS is totally different. So, anyone who's lapped their IHS, please look at the picture linked above and tell me... does that look like a lapped IHS?

I've previously purchased three OEM CPUs from Monarch. The Sempron and A64 3200+ were AFAIK new. My first Opteron 144 (the subject of most of this thread) was mounted in the board I ordered with it and so I couldn't tell if it was "virgin" or not, but it did have bent pins whether due to mishandling or whatever. Now, these two OEM Opteron 144 chips... one looks brand new but the second one is what has me worried.

I was planning to test the chips later today, hopefully get some feedback on this issue by then.

------------------------------------ original post ------------------------------------------

Fellow overclockers, please be generous and forgiving to a DFI n00b.

Purchased an OEM Opteron 144 and a DFI LANPARTY Ultra-D from Monarch and just got it in today. So far... terrible overclocks and it's the hottest running AMD CPU I've had in a while. Never had a DFI LANPARTY board with it's myriad of options, so perhaps I'm missing something.

Parts:
OEM Opteron 144
Thermaltake Venus 7+ HSF (free from Monarch)
DFI nF4 LANPARTY Ultra-D
2x512MB OCZ Premier CAS 2.5 PC3200 RAM with copper heatspreader
Sapphire Radeon X800GTO PCIe
Fortron "Green" FSP400-60GLN
Toshiba DVDROM

I think I have the 6/23/05 BIOS 623-3. To take other stuff out of the equation:
DRAM Frequency Set 100 (1/02)
CAS 3.0
LDT/FSB Frequency Ratio X 2.5
CPU VID Control 1.40 V (or Auto)
LDT Voltage Control 1.40 V
Chip Set Voltage Control 1.70 V
DRAM Voltage Control 2.80 V
everything else in BIOS is default

The stability test is running Prime from UBCD for a few minutes. I'll worry about long term stability later, just want a quick check on what's definately not stable.

With CPU/FSB Frequency Ratio set to X 9.0 (default on this Opteron) and testing FSB/HTT in 5MHz increments, maximum of 250MHz, or 2250MHz core speed. At 255Mhz Prime craps out immediately.

Lowering the CPU/FSB Frequency Ratio to X 7.0, FSB/HTT can be set to 300MHz (with all the same other settings as listed above) and Prime runs happily with the CPU at 2100MHz.

With the CPU/FSB Frequency Ratio set to X 8.0 and FSB/HTT still set at 300MHz (and all other settings as above) so that the CPU runs at 2400MHz, Prime craps out immediately.

At 8X ratio with HTT 280MHz for 2240MHz core speed, Prime runs fine.

At 8X ratio with HTT 290MHz for 2320MHz core speed, Prime fails immediately. With these settings I started boosting the vcore in 0.025v increments. 1.475v and it became Prime stable. 1.425v and 1.450v Prime failed immediately. With vcore at 1.475v, raised HTT to 300MHz for 2400MHz core and Prime fails immediately. Raising ratio to default 9X just to see what happens, and at 2700MHz it cannot even boot from disc - spontaneously reboots after "Building DMI Pool....."

I'm having such a terrible time with this that soon I'm gonna be running back to embrace my socket 754 Semprons that clock sky high with complete ease (one hit 2.7GHz).

My wife's gonna be gone the rest of the week starting tomorrow so I'll probably "borrow" her computer with the Abit KN8 board for a second opinion on this CPU.

Perhaps I need to sacrifice a chicken to the overclocking gods?

 

[TheRyuu]

Whats the hypertransport multiplyer? Make sure to back that down to 3x/4x and try to keep the hypertransport speed around 1000 (it's ok if it's less, mines at about 800mhz right now).

 

[Avalon]

I'd leave the HT frequency at 3x, you won't ever need to set it lower. Other than that, my experience with the Ultra-D is that it undervolts. Yours may or may not.

As for the chip running warmly...yes, they do run much warmer. My Venice was a much cooler running chip.

The weakest 144 I had was a CAB2E 0540 FPMW. It was able to run at stock volts up to 2.2ghz, at which point it would crap out. I had to set my voltage to 1.45v in BIOS to get 1.4v, which let me prime at 2.25ghz, and from there I gained a couple hundred mhz until I had to bump to 1.5v BIOS to get 1.47v @ 2.5ghz. At 2.6ghz, I needed 1.52v, and could only superpi at 2.7 with 1.55v, so I dunno...you may be seeing something similar to what I did with that particular chip. All of my other Opterons have been fantastic.

What week code is your chip?
Also, disable CPU spread spectrum.

 

[firewolfsm]

mine hit 2.7 with 1.55V, higher voltage than i originally wanted, but any lower and i had to go with 2.4GHz

 

[rise]

in addition to what avalon said, why do you have your chipset and ldt set so high? i don't think you need to touch those yet, ime anyway. i've actually always used the lowest on those.

also, some say using a 105mhz pci-e on ati cards helps stabilty. i don't know why but it may be worth a shot.

and the obligatory "you have all the boards power connections plugged in? "

i'd definitely see how it runs in your wife's compy.

 

[F1shF4t]

Just an advice set all the memory timings in bios (eg 3-3-3-7), using some dividers the auto option set some timings too low and made it unstable completely.

U dont really need to up the voltage on the chipset and the LDT, mine are running at stock no problems.

Also check those voltage regulators (the ones with those 3 heatsinks) i had my board crash once from them overheating when i uped the voltage and had the cpu at 2.8ghz.

 

[Avalon]

Originally posted by: Dark Cupcake
Just an advice set all the memory timings in bios (eg 3-3-3-7), using some dividers the auto option set some timings too low and made it unstable completely.

U dont really need to up the voltage on the chipset and the LDT, mine are running at stock no problems.

Also check those voltage regulators (the ones with those 3 heatsinks) i had my board crash once from them overheating when i uped the voltage and had the cpu at 2.8ghz.

Yeah, those can get real hot. I'd lower the voltage on your chipset a notch, you shouldn't need to increase that.

 

[furballi]

-Memory divider at 5:6 or 166MHz
-Vdimm at 2.7
-Vcore at 1.5
-RAM timing at 2.5-3-3-8-1T
-HT at 3x...also try 4x

Start at 250MHz FSB and move up in 5 MHz increment. The average Opteron should run out of gas around 265 to 275 MHz FSB at these settings.

The A64 and Opteron run much hotter than my 3100 Sempron (Paris). I was able to push the Sempron up to 273MHz FSB with 1.55Vcore. Have two 512MB sticks of Corsair VS. Vdimm is at stock. Timing is 2.5-3-3-6-1T. 1M Super Pi is 37.7 sec. Max temp of 45C.

Please post your best 1M Super Pi time with this rig. I suspect you'll see a time of 34 to 35 sec.

 

[Zap]

Whats the hypertransport multiplyer?
...
I'd leave the HT frequency at 3x, you won't ever need to set it lower.
...
Just an advice set all the memory timings in bios (eg 3-3-3-7), using some dividers the auto option set some timings too low and made it unstable completely.
...
Also, disable CPU spread spectrum.

As I mentioned, set to 2.5X. I'll try raising it to 3X and manually setting all timings. Don't think it'll help because the board will run at 300MHz HTT fine if I lower CPU multiplier. Spread spectrum all disabled.

in addition to what avalon said, why do you have your chipset and ldt set so high? i don't think you need to touch those yet, ime anyway. i've actually always used the lowest on those.

I'll try putting those down to default.

some say using a 105mhz pci-e on ati cards helps stabilty

I've read about that and have tried it both at 100 and 105MHz, same results.

What week code is your chip?

I'll report back with the code later, after I try some of these suggestions in the DFI board and try the CPU in the Abit board. Hopefully I'll get a chance by this evening. I'll report back.

My gut feeling is that my CPU is like the "weakest 144" that Avalon had. Pathetic. After this I'm gonna go back to playing with my Sempron 3300+ @2.7GHz near default vcore. :roll:

 

[5t3v0]

My 144's a week 40 CABNE & its a gash overclocker an' aw. Luck of the draw.

 

[Zap]

No go... same lousy results with the suggestions. My week is 0548 - that sound right?

OSA144DAA5BN
CAB3E 0548FPEW
138109AL51007

I'm guessing the bolded part is the week code?

Only thing I haven't tried yet is to run this on the Abit board - been busy today with my programming assignment, looking into internships, studying for midterm, looking up all the rebates I've sent off since Black Friday (three of them are MIA and one was denied for bogus reasons :| and now they want me to fax them what I already mailed them).

 

[96redformula]

Nope not a bad week at all. Look at the thread below you called "One Heck of An Overclock"

 

[shinzwei]

My Opteron 144 is at 1.47V 2.60ghz Stable.

Try:

mem div @ 100
FSB @ 325
HT @ 3
Voltage @ 1.5 which dfi undervolts to 1.47

check to see if these work for u

 

[Zebo]

Never had a DFI LANPARTY board with it's myriad of options, so perhaps I'm missing something.
--------------
Pretty crazy huh?
make sure all four power connectors are connected to mobo.

I think I have the 6/23/05 BIOS 623-3. To take other stuff out of the equation:
---------------------
DL latest BIOS to properly work with Optys CPUID


DRAM Frequency Set 100 (1/02)
CAS 3.0
LDT/FSB Frequency Ratio X 2.5
CPU VID Control 1.40 V (or Auto)
LDT Voltage Control 1.40 V
Chip Set Voltage Control 1.70 V
DRAM Voltage Control 2.80 V
everything else in BIOS is default
---------------------------
-You can use "auto" LDT on the DFI or just use 3, never use half LDT's.
-Chipset needent be over 1.55V.
-Since the DFI "undervolts" CPU by .05 I'd give a startup of 1.45 and CPU VID of 1.45 = 1.40
- Try DRAM of 133 or 140 I could'nt even pass memtest with 100 or 166 @ 200 HTT.


The stability test is running Prime from UBCD for a few minutes. I'll worry about long term stability later, just want a quick check on what's definately not stable.
----------------------------
You try memtest? Tests 5 and 6 more specifically?

With CPU/FSB Frequency Ratio set to X 9.0 (default on this Opteron) and testing FSB/HTT in 5MHz increments, maximum of 250MHz, or 2250MHz core speed. At 255Mhz Prime craps out immediately.
------------------------
Virtually impossible to find 144's clocking that low. Somethings up most likly mem or chipset is crapping with all those volts/heat.


Lowering the CPU/FSB Frequency Ratio to X 7.0, FSB/HTT can be set to 300MHz (with all the same other settings as listed above) and Prime runs happily with the CPU at 2100MHz.
With the CPU/FSB Frequency Ratio set to X 8.0 and FSB/HTT still set at 300MHz (and all other settings as above) so that the CPU runs at 2400MHz, Prime craps out immediately.
At 8X ratio with HTT 280MHz for 2240MHz core speed, Prime runs fine.
At 8X ratio with HTT 290MHz for 2320MHz core speed, Prime fails immediately. With these settings I started boosting the vcore in 0.025v increments. 1.475v and it became Prime stable. 1.425v and 1.450v Prime failed immediately. With vcore at 1.475v, raised HTT to 300MHz for 2400MHz core and Prime fails immediately. Raising ratio to default 9X just to see what happens, and at 2700MHz it cannot even boot from disc - spontaneously reboots after "Building DMI Pool....."
---------------------------------
but then you say this..

Abit KN8 board for a second opinion on this CPU.
--------------------------------
Exactly what I'd try. My feeling is a) board does'nt like Memesetting you're giving it or B) the chipset volts ex: I use 1.5 @ 313 HTT C) can't be possisble to have such a bad clocker. You always get the good ones.

Edit someone suggested 3-3-3 - absolutly. Should look like this to find max CPU.

CPC - Disable.
TCL - 3
TRCD - 3
TRAS - 8
TRP - 3
TRC - 9
TRFC - 16
TRRD - 3
TWR - 3
TWTR - 2
TRWT - 3
TREF - AUTO
TWCL - AUTO


All these rest below these leave after "loading optimized defaults"

In addition to trying 133/140/150 memsettings (aka DRAM Frequency SET).. I'd also lower volts to mem like 2.7 max since DFI overvolts mem settting up mem controller for fail at low CPU volts.

Good luck Zap LMK know how it goes... I mentioned awile back I gave up on the DFI rather quick due to just getting bored trying to tweak maximus with all it's settings.. Hit the jackpot running 180 mem setting and conservative timings with ballistix and gave it up to my son to play with. But definity experianced some weird behaviors like you discribe with different and sometimes even lower memsetting with that board.

 

[Zap]

Updated BIOS to latest (11/05) version and tried the various settings suggested, no go. Tried CPU in Abit board, same results. With vcore set to default on both boards, will Prime at 250Mhz HTT but not 255MHz HTT using default 9x multiplier. Also tried different RAM on the DFI board, and tried with a Zalman 7700AlCu HSF in case it was lack of cooling.

To reiterate, on the DFI I tried lowering the multiplier and can easily run at 7x300MHz, but 8X300MHz fails. This chip doesn't want to do 2.3GHz or higher on default vcore "setting," unlike my wife's A64 3200+ @2.4GHz and my x2 3800+ @2.5GHz.

The strange thing is that this Opteron will POST at near 3GHz on the DFI board - tried it just for the heck of it with somewhere around 1.55v - of course it wouldn't even start booting off disc at anything over 2.6GHz or so.

I'll just have to live with cranking some voltage.

*SIGH*

 

[Zebo]

I know who not to buy a used opteron from in FS forums.

Geez you must have to worst one is history.

 

[Zap]

For $130 I'm not gonna bother selling it. That's cheaper than any other socket 939 CPU.

There's a couple of other chaps around here who report not being able to hit 2.3-2.4GHz on their Opterons... mostly they're drowned out by the chaps screaming how awesome their Opterons are for being able to get a CPU-Z screenie at 2.7GHz. I just saw one post like that in the past two days on a low overclock like mine... somewhere around here.

One other thing, the CPU had some bent pins... when I was switching it to the Abit board I noticed that in two sections some of the pins (about 4-5 in a row) were leaning over a bit making it difficult to seat into the socket. It's as if someone squeezed the CPU too hard while holding it and their fingertips bent the pins inwards in two spots. Was probably whoever tested it at Monarch (they tossed in a free HSF so the board arrived with CPU/HSF installed).

 

[Avalon]

I've never seen a CAB3E, a week 0548, or an FPEW. Maybe it's a combination of the three. If you really do have a CAB3E and didn't mistype CAB2E, it could be the new bum stepping.

 

[robertk2012]

I just wanted to make sure you loosened the memory timings. Sometimes the intergrated memory controller is what causes the problems since it runs at the same speed as the CPU. Thats why it would make a difference even though the memory alone doesnt cause a problem when the cpu is at a 7x multiplier.

 

[furballi]

Sounds like Monarch sold you a refurbished CPU. Had an OEM board from them with about a dozen resoldered joints on the back of the board.

 

[Zap]

Originally posted by: Avalon
If you really do have a CAB3E and didn't mistype CAB2E

You're right. Just looked at it again right now after seeing your post:

CAB2E 0548FPEW

This time I double checked!

Dunno about a refurbished CPU - it's either new or used. Since I know for a fact that they handled the CPU (since they installed and tested it) then can't really tell if new or used. I don't think CPUs get refurbished, just "recertified" if it works or trashed if it fails.

 

[nealh]

what color slots is the ram in...mine ram does best is the yellow slots

also the OCZ guys have posted they think yellow slots are best with 623-2 bios or 704-2bta bios

the ram settings can have a big impact on the how you do

 

[robertk2012]

Yeah try slots 2 and 4. I think its 2 and 4. Probably what the previous poster is talking about,

 

[Zap]

The yellow slots are 1&3. I was using the orange. The CPU does exactly the same in the Abit board.

I'm giving up on this. Spent some time today building my mATX gaming box with the Sempron running 2.75GHz on my Tforce board. Easy as pie!

 

[96redformula]

I have the EXACT SAME CPU AND Motherboard as you. I Spent 2 whole days getting it to overclock right, their are Sooooo many options with DFI, each one will make a difference. I got mine prime stable at 2800MHZ running 312x9 on 1.55volts. I am still working on it, but my damn ram went out and screwed me up, the thread on me being on my 4th set of ballistix. I got my best results on the orange slots also.