Jump to content

Unofficial Opteron 939 Thread


Recommended Posts

  • Replies 2k
  • Created
  • Last Reply

Top Posters In This Topic

In the past month and a half, I have upgraded my machine with an Opteron 175, OCZ Platinum EL DDR500 3-3-2-8 2x1gb RAM. I haven't been able to overclock it yet because I don't have the time, but the farthest I remember having it was like 226x11 with quite a bit of voltage..like 1.45, but the temps were getting like up to 47C or so...I don't know why this dang chip won't go farther...

But, I'll have to post the stepping and stuff later when I get home from work. Anyone know what I can do about having a weak core? I can run Prime on core 0 forever and superPi, but on core 1, whenever I get past a certain point, it fails, but core 0 never does. Is there anything I can do about this, besides getting a new CPU? Thanks

Share this post


Link to post
Share on other sites

Whoa! You got some super ram there! Nice! Only .05 more and hopefully you can keep those timings. D

 

What are your load temps at with that Sonic Tower?

 

Hey Litho, I am using the same ram but could never bring it more than 262 1T with [email protected] ( seemed the voltage didn't help a bit), the rest of the settings are auto. Can you please elaborate what are your settings in your signature, such as "[email protected] 15-17-3-3-2-3-3072". I am using the RS482 board btw.

 

Thanks.

Share this post


Link to post
Share on other sites

Guest LithoTech
Hey Litho, I am using the same ram but could never bring it more than 262 1T with [email protected] ( seemed the voltage didn't help a bit), the rest of the settings are auto. Can you please elaborate what are your settings in your signature, such as "[email protected] 15-17-3-3-2-3-3072". I am using the RS482 board btw.

 

Thanks.

 

Hi Hank,

 

Hey, we have the exact same processor too! Stepping is CAB2E 0602GPKW!

 

DFI RS482 Infinity mATX S939 RS482 BIOS321

AMD Athlon 64 Opteron [email protected] 1.375v CAB2E 0602 gpkw

2x OCZ Gold Edition XTC 1GB PC-4000 1:1 3-4-4-8-12-16

WD 80G 8M ATA100

Antec Sonata II 450W SmartPower

 

I'll post my full mem settings in a sec, there are a couple other things I noticed first. I'm no expert on this, far from it, only seriously started overclocking a couple months ago with this present build. Previous overclocking was purely experimental and educational in nature, seeking flawless stability rather than optimization, top speed and good enough for 8hrs P95 stability. I still tend to prime for 24+ hours, and take pride in uptime of my systems. So what I say or question may be off track, incorrect or newbish, please forgive. :)

 

I know what you mean about extra Vdimm voltage, did nothing at all for mine, in fact produced more errors in most cases.

 

You'll see below, I am only using 2.6v on the ram, which I believe is supposed to be 2.7 - 2.75 for stock settings, and warrantied to 2.9v. I tried running less than 2.6v, no go there. So for mine, with this mobo and PSU, 2.6v is where she likes no matter what the clock speed. I still need to verify my Vdimm voltages with a multimeter, but the BIOS and MBM both report it as about .08 above its setting. Therefore, if this is correct, 2.6 = almost 2.7 on my system.

 

I tried running my Opty below 1.4, using various 1.3 settings with and without variuos special VID controls, and found that it didn't like less than 1.4v when clocked past 250 or even 200. In all cases when trying to get over 260mhz on my BH-5 ram, upping my cpu voltage reduced the number of #5 memtest errors I got. So I'd reccomend trying 1.4-1.45 volts on your CPU, it does seem to help with the memory controller and memory stability.

 

I've no knowledge at all of the Infinity board, but I'm wondering about your bios version, seems sort of old compared to the nf4 bios' that are most commonly used. Unless, of course, the date for that series is 2006! Have you tried the memory with different bios'?

 

The motherboard difference alone could easily account for the different top-outs we are getting with this ram; might be worth a look to see what others top out at because I really don't know where I am falling with 270 compared to anyone else regardless of processor or mobo.

 

I've tried several times to get 280mhz out of these, and am convinced it is not going to happen. I'm vaguely sure that I did get 275mhz stable in memtest but didn't make note of it and didn't pursue it further because I limited myself to 270 on the CPU until I swapped the stock heatsink -- more of a heat limit than a cpu wall. I'll be trying this again in the next few days or weeks as I have now done the swap, but still toying with the BH-5 at the moment.

 

Here's my settings from the file I made for my O/C database entry that I still haven't got around to posting:

 

AMD Opteron 146 Venus | stock air | CAB2E 0602GPKW

 

2x1024 DDR500 UCCC OCZ PC4000 Gold GX XTC | OCZ5002048ELGEGXT-K

 

eVGA EGeforce 6800GS SLI 256mb 256bit PCI-E, 500/1.1 | 81.98 drivers

 

2x WD Caviar SE16 250gb SATA2 Raid-0 Boot | Samsung Spinpoint 250gb PATA

 

Creative Sound Blaster X-FI Xtreme Music 24BIT Sound Card

 

LG 4167B 16X DvD+-RW 16X Dual Layer Writer IDE 2MB

 

Sony floppy

 

Enermax Liberty 500 Power Supply

 

Antec P180

 

 

 

==========

 

 

Genie BIOS Settings:

FSB Bus Frequency - 270

LDT/FSB Frequency Ratio - 3

CPU/FSB Frequency Ratio - 10

PCI eXpress Frequency - 100Mhz

 

CPU VID StartUp Value - Auto

 

CPU VID Control - 1.400v

CPU VID Special Control - Auto

LDT Voltage Control - 1.20v

Chip Set Voltage Control - 1.50v

DRAM Voltage Control - 2.60v

 

DRAM Configuration Settings:

DRAM Frequency Set - 200=RAM/FSB:01/01

Command Per Clock (CPC) - Enable

CAS Latency Control (Tcl) - 3

RAS# to CAS# delay (Trcd) - 04 Bus Clocks

Min RAS# active time (Tras) - 07 Bus Clocks

Row precharge time (Trp) - 03 Bus Clocks

Row Cycle time (Trc) - Auto (15) Bus Clocks

Row refresh cyc time (Trfc) - Auto (17) Bus Clocks

Row to Row delay (Trrd) - 03 Bus Clocks

Write recovery time (Twr) - 03 Bus Clocks

Write to Read delay (Twtr) - 02 Bus Clocks

Read to Write delay (Trwt) - 03 Bus Clocks

Refresh Period (Tref) - 3072 Cycles

Write CAS Latency (Twcl) - 1

DRAM Bank Interleave - Enabled

 

DQS Skew Control - Auto

DQS Skew Value - 0

DRAM Drive Strength - Auto

DRAM Data Drive Strength - Auto

Max Async Latency - 10.0 Nano Seconds

DRAM Response Time - Normal

Read Preamble Time - 7.5 Nano Seconds

IdleCycle Limit - Auto

Dynamic Counter - Enable

R/W Queue Bypass - 8 x

Bypass Max - 04 x

32 Byte Granularity - Disable(4 Bursts)

 

 

==========

 

 

Prime95 + 3dMark2001SE + CPU-Z 1.31 + Everest 1.51:

 

27ghzp9524hrs22min416200610573.th.jpg

 

 

short description:

270x10 @ 2700Mhz

3-4-7-3 @ 2.6v vdimm

 

Things I note:

 

Row Cycle and Row Refresh Cycle are higher than average (I think).

Refresh period of 3072 seems uncommon but worked better for me.

Drive Strength and Data Drive Strength left on auto because of lazyness and it worked good.

Max Async Latency and Read Preamble are Opteron sensitive settings, playing with these can help a lot. Again, my settings are higher than average (I think).

I haven't really done any testing with the Dynamic Counter off yet, at least I can't find any note of it in my stacks of papers.

 

It's obvious that although I am happy with this as an overclock database entry, there is still a lot of untested settings and I think plenty of room to get a stable 275 (or even 280 if I am real lucky) out of these sticks. As well, I've yet to push this CPU at all, but no matter what it is capable of doing I'll run it at 2.5ghz for its 24/7 operation of work/games/folding.

 

Hope this helps! I know a couple posts were paramount for me finally getting this BH-5 to behave at over 260. Damn thing is, it does me no good unless I can reach 270mhz with them because otherwise they are on a divider and running at sub 250. So I know full well how you want to get 270, and I know some of these XTC are capable of it even though most examples I've seen are also running on a divider. I think for 24/7 use a divider is best since it stresses the ram less, and must be more stable, and the difference in performance is really only needed for a quick benchtest run and completely unnoticable in a game or general use.

Share this post


Link to post
Share on other sites

Damn thing is, it does me no good unless I can reach 270mhz with them because otherwise they are on a divider and running at sub 250.

 

 

I know exactly what you mean. It seems like no matter how high the 2GB modules clock, it's never quite enough to reach 1:1. After trying 16-18 sets of HZ modules that topped out anywhere from 260 to 280 I decided to give a set of OCZ EBs a shot. I'm convinced that from a ease of use/performance standpoint this is the way to go. Here are my timings running 295 FSB, 166 05/06 divider. With HZ memory the cpu was OCDB stable at 295x10. Using the same VID with the OCZ I lose 20MHz. I'm guessing it's the strain on the memory controller with some of the timings being so tight.

 


[b]DRAM Configuration Settings:[/b]

[color=darkgreen]

Dram Frequency Set (Mhz).............. 166=RAM/FSB:05/06

Command Per Clock (CPC)............... Enable

CAS# latency (Tcl).................... 3.0

RAS# to CAS# delay (Trcd)............. 03

Min RAS# active time (Tras)......... . 00

Row Precharge Time (Trp).............. 02

Row Cycle Time (Trc).................. 07

Row refresh timg (Trfc)............... 12

RAS# to RAS# delay (Trrd)............. 00

Write Recovery Time (Twr)............. 02

Wite-to-Read Delay (Twtr)............. 01

Read-to-Write Delay (Trwt)............ 02

Refresh Rate (Tref)................... 3120 Cycles

Write CAS Latency (Twcl)...............01

DRAM Bank Interleaving................ Enabled



DQS Skew Control...................... AUTO

DQS Skew Value........................ 0

DRAM Drive Strength................... Level 04

DRAM Data Drive Strength.............. Level 2

Max Asynchronous Latency.............. 08.0

DRAM Response Time.....................Normal

Read Preamble Time.................... 03.0

IdleCycle Limit........................016

Dynamic Counter....................... Disable

R/W Queue Bypass.......................16X

Bypass Max.............................07X

32-Byte Granularity................... Disable (4 bursts)

[/color]

Share this post


Link to post
Share on other sites

Opteron 146 CABYE 0540FPBW - Liquid Chiller CPU@+3°C

 

More to come :nod:

Great speed, especially for those temps! Any reports of cold bugs with that stepping? And what liquid are you using? I have found that ~2/3 denatured alcohol, ~1/3 H2O provides me with the best temps thus far.

Share this post


Link to post
Share on other sites

Guest LithoTech
I know exactly what you mean. It seems like no matter how high the 2GB modules clock, it's never quite enough to reach 1:1. After trying 16-18 sets of HZ modules that topped out anywhere from 260 to 280 I decided to give a set of OCZ EBs a shot. I'm convinced that from a ease of use/performance standpoint this is the way to go. Here are my timings running 295 FSB, 166 05/06 divider. With HZ memory the cpu was OCDB stable at 295x10. Using the same VID with the OCZ I lose 20MHz. I'm guessing it's the strain on the memory controller with some of the timings being so tight.

 

Ummm, 18 sets?? :eek: :eek2: Tell me you work at a retailer or wholesaler!

 

What percentage roughly, topped in the 260s? I guess I really should ask, what percentage hit 280?

 

And so the OCZ EBs topped out even less? You'd need to be mem stable at 265mhz for a 9/10 divider, so close but so far?!

 

I had a look at EB when I got the gold XTC, and nixed that idea because of the price and I suspect my Plat 3200s are the same CE-6 and I developed an aversion to them when I couldn't reproduce a 225mhz stable setting I once had -- damn things barely do 210! I do plan to give them another shot, I know a lot more now than I did when I was using them.

 

They are great at rated speeds with very nice timings and solid stability, and yes ease of use! I've got some decent benches with them on a divider. Perhaps I will try them again sooner than later.

 

I want to spend some time building a spreadsheet for my temperatures using various fan speeds, case opened or closed, load or idle, overclock, etc. I've been logging the info onto paper regularly for a while, and want to get them into Excel to have a proper look at the trends, maybe make a few charts. I'd make a second sheet for benchmark results of various overclocks and mem timings, w/ or w/o dividers. Sifting through the info by divider, or bench result, or timing, would be a lot easier than shuffling papers looking for a note, and making a new note.

 

Is there already such a beast tucked away somewhere here? Rather than recreate the wheel... I did find that spreadsheet here that listed all the G.Skill ram settings, and I've been using it as a digital record of my own bios settings, but it got me wishing that the other ram types and manufacturers that people use here had the same file or database of settings, and not just G.Skill. Then, it wouldn't take much to combine them all into one master bios setting spreadsheet for all ram and DFI boards. I looked around and couldn't find any such file for other types of ram.

Share this post


Link to post
Share on other sites

Ummm, 18 sets?? :eek: :eek2: Tell me you work at a retailer or wholesaler!

 

What percentage roughly, topped in the 260s? I guess I really should ask, what percentage hit 280?

 

And so the OCZ EBs topped out even less? You'd need to be mem stable at 265mhz for a 9/10 divider, so close but so far?!

 

I own a few different businesses but none of them are computer related.

 

2 sets hit 280 and a few went to the mid 270 range. 260-264 seems average with the sets I tested.

 

That is what I like about the EBs. With such tight timings you can run 5/6 and have the same real world performance as 9/10 or 1/1 with loose timings.

Share this post


Link to post
Share on other sites

Guest LithoTech
I own a few different businesses but none of them are computer related.

 

2 sets hit 280 and a few went to the mid 270 range. 260-264 seems average with the sets I tested.

 

That is what I like about the EBs. With such tight timings you can run 5/6 and have the same real world performance as 9/10 or 1/1 with loose timings.

 

Interesting! I must agree, especially with the crap-shoot of getting a kit that will do 270+. If you are going to be stuck with some that top at 260, you are far better off with EBs.

 

Thats pretty much what decided me on getting some of these Centon BH-5. Well, not exactly, because I didn't know this, only the theory. So I needed to find out for myself, first hand, a learning experience well worth the bucks to me. If there were a night school course that offered what I've learned from this ram, I'd gladly pay the same amount as a course fee. But there is no such course. There is only here, and other forums, and the ram itself.

 

So I'm working on building a collection, or selection of ram. All I'm lacking for now is a good kit of TCCD. Lots of that around here, reasonable price.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×
×
  • Create New...