rustynator

After scavenging for a golden stepping on ebay for a while, I managed to get my hands on a new, OEM Opteron 146 cabye 0540FPBW 1341649J50404. I paid quite a lot for the CPU (about 20% over the cost of a new, retail version in the shops) so I was expecting quite a lot from it (i.e. 3-3.1GHz on air).

I spent the whole of yesterday doing a slow burn in on the CPU (a couple of hours prime95 followed by a 200MHz increase). On stock volts, I started to get prime errors after 10 minutes at 2.7GHz. 2.8GHz requires 1.425V to be prime stable for 20 minutes, and 2.9GHz isnt even stable for 1 minute with 1.52V.

I have read loads about this CPU on XtremeSystems.com, and just about every CPU with this stepping seems to do a prime stable 3.0-3.3GHz (depending on the cooling) with low volts.

I have tried to rule out the memory by running slack timings and a 180 divider. My temps arent too bad either....37C load at 1.35V and 42-44C load at 1.52V...so I dont think heat is the big issue.

Can anyone think of anything that I might be doing wrong to inhibit the overclock? Or do you think this is just a one off, rubbish version of the stepping, which I have paid over the odds for?

The Viper

it really is true about steppings...there r good a and bad apples in every basket, anyways go back to 2.8 and put some more volts and get that stable than you can move a little slower with the increses, try upping your chipset voltage too

Jonrezz

did u lower htt to like 3x or 2x?

just a guess

rustynator

Thanks el7145. Just about every 0540FPBW I have seen has cracked 3GHz on ridiculously low volts (quite a few with air cooling), all apart from one guy on XS.com, who can only get to 2.9GHz on water cooling with 1.50V. Looks like I have a CPU from the same/similar wafer to his.

I will knock the CPU back to a lowly 2.8GHz, with as small a Vcore as possible, and go from there. Gosh this is annoying! I paid £190 for this oem CPU when I could have bought one with a random (but problably similar OCing) stepping for £160. Damn you stepping God!

EDIT: I am currnetly using a HT multi of 3 but I will give 2.5/2 HT multi a try Jonrezz. I will be very surprised if that solves the problem with a HTT of only 290 though.

The Viper

also just for kicks make sure your +12V rail isnt fluctuationg too much

euphoricmix

Is your memory DDR-400, DDR-500, etc? I guess it really does not matter what type it is, run an aggresive divider to see if you can get stability. Just so you can rule out the memory. Some 2GB kits need to be ran at a 2 command rate, so if you have it set to 1 you might want to play around with that too. Like el7145 stated make sure your PSU is giving you clean, consistent power. A good application to track that is OCCT, but you have to get a complete test for results. So set it to an overclock you know you have stability with and upon completion you will get voltage charts. There should not be more than 5% of play between you highs and lows with your +12v.

Sidenote: Leave your HTT multplier at 3, 2.5 will be useful for 333 HTT and above.

ncsa

Please post your BIOS Version and Genie Settings ... Edit this to help you out - from an Expert BIOS.

rustynator

Hi ncsa

Genie BIOS Settings:

FSB Bus Frequency............................. - 290
LDT/FSB Frequency Ratio....................... - x 3.0
CPU/FSB Frequency Ratio....................... - x10
PCI eXpress Frequency......................... - 100Mhz
CPU VID StartUp Value......................... - 1.500v
CPU VID Control............................... - 1.500v
CPU VID Special Control....................... - Above VID * 104%
LDT Voltage Control........................... - 1.30v
Chip Set Voltage Control...................... - 1.60v
DRAM Voltage Control.......................... - 2.60v

DRAM Configuration Settings:

DRAM Frequency Set............................ - 180=RAM/FSB
Command Per Clock (CPC)....................... - Enabled
CAS Latency Control (Tcl)..................... - 3
RAS# to CAS# delay (Trcd)..................... - 3.0
Min RAS# active time (Tras)................... - 8.0
Row precharge time (Trp)...................... - 3.0
Row Cycle time (Trc).......................... - 09 Bus Clocks
Row refresh cyc time (Trfc)................... - 14 Bus Clocks
Row to Row delay (Trrd)....................... - 04 Bus Clocks
Write recovery time (Twr)..................... - 03 Bus Clocks
Write to Read delay (Twtr).................... - 02 Bus Clocks
Read to Write delay (Trwt).................... - 06 Bus Clocks
Refresh Period (Tref)......................... - Auto
Write CAS latency (tWCL)...................... - 01 Bus Clocks
DRAM Bank Interleave.......................... - Enabled
DQS Skew Control.............................. - Auto
DQS Skew Value................................ - 0
DRAM Drive Strength........................... - Auto
DRAM Data Drive Strength...................... - Auto
Max Async Latency............................. - 9ns
DRAM Response Time............................ - Normal
Read Preamble Time............................ - 6ns
IdleCycle Limit............................... - 256 Cycles
Dynamic Counter............................... - Enable
R/W Queue Bypass.............................. - 16 x
Bypass Max.................................... - 04 x
32 Byte Granularity........................... - Disable(4 Bursts)


Those are the slack timings which I used to try and reach 2.9GHz, whilst eliminating the RAM from the equation. My RAM is capable of 280MHz at 3-3-3-6 at 2.6V.

I have just posted a thread over at xtremesystems.com, and it seems that maybe this stepping isnt the golden stepping for air/water cooling, but instead it the phase cooling golden stepping.

I notice that your amazing OC is under phase ncsa. Have you tried yor FPAW's on air before? What was your max OC?

Thanks

Mastacator

Try giving your pci-e a nudge. say 103mhz, for some reason it helped me a little.

ncsa

This is what the chip would do on Air .. however it is on an Asus so did not know what it could do on a DFI as that is setup for Phase. The Asus would top out at this as going to 320 the SATA drives would drop off.



On Air it needed alot of Vcore but on Phase this was easy at stock Vcore...



Your chip may well do better with a higher Vcore but this will require better cooling.

Send it across and I will test it for you