Unofficial Opteron 939 Thread

[Guest LithoTech]

@LithoTech I've generally found LanParty NF4 BIOS - Cleaned up 704-2bt a little has given me best performance. It's on Tmod's CD.

 

I'm just revisiting my memory timings again after getting a replacemrent set of LA.

 

luck

 

I've been resisting real hard at updating to this bios, and it's a loosing battle. The reason I hold off is that I like to give the stock 623 a decent chance, as I need to learn it as much as it needs to learn my cpu and ram. Also, I never like updating things unless I see something in the new version I can't live without. Not knowing these bios's intimately, I've no idea what 704 has that 623 doesn't. Well, other than the huge amounts of praise it has received.

 

So yup, I'll be doing the switch to 704, and I have a Tmod CD burned from not long ago.

 

But first, I really should try each stick by itself, as well as trying the yellow slots.

 

Or should I leave the yellow slot test until I have the 704 bios?

 

/edit: I just did a google on those Gskill LA (didn't know what LA was), do they really come without a heatspreader? Pretty impressive timings and speed :drool:

[supershanks]

originals were bare then g-skill started clothing them so mine have got the spreaders.

 

re 704 i think the tables are improved over 623 maybe. This post reflects the sirt of thing. Can't get high read preamble you can also see in the original link

LanParty NF4 BIOS - Cleaned up 704-2bt a little

the improvement in running a number of modules which were problematical.

 

i'm currently playing around with settings also with this Text Tutorial - TCCD DDR600 2.5-3-3-7 twenty four seven: The ultimate DFI settings

 

I think this is exactly right:-

Kakaroto

I will not use Memtest86++ because it sucks big time when tuning for a 24/7 setup, windows is what matters so programs as SuperPI32M for initial test are going to be used and for long term stability SP2004 (RAM Stress).

Have been running with sp2004 for memory testing for a while.

 

Have got 275x10 Sp2004 stable 4 8.5hrs which is great - trying to get to a 273x11 that will be stable. Atm can get in there pretty well 1mb pi etc., HotCPU & all the benches but not prime stable.

 

I tried the yellow sockets b4 but came back to orange.

 

luck

[aowen512]

So can anyone get back to me about my core volts dropping about .040 when I run something stressful like occt and typically jumping about .015 idle. Is this a very bad drop or should I not be concerned? I am not sure how this volts normally acts as I am not too fimiliar with this platform.

 

thx fellas

[Guest LithoTech]

So can anyone get back to me about my core volts dropping about .040 when I run something stressful like occt and typically jumping about .015 idle. Is this a very bad drop or should I not be concerned? I am not sure how this volts normally acts as I am not too fimiliar with this platform.

 

thx fellas

 

Heh, sorry aowen, seriously didn't see the first post there, I can help a bit I think.

 

I have upped my core volts since the last time that I posted. I currently have set the vcore in bios to be 1.65. Cpu-z reports 1.627-1.637 idle. Once I start up OCCT the volts drop to 1.593 pretty steadily which is a diff of .043 volts. Thats a huge drop in volts I think. Is this stopping me from having a high stable overclock??

 

First, I have a couple questions.

 

1) Are you having a stability problem with a high overclock? Doh, silly me, who isn't? Push it high enough and eventually...

 

2) What kind of Power Supply do you have? Your sig doesn't mention it?

 

cheap rosewill case, does the trick!

 

Did it come with the case? :confused:

 

If you need help in identifying what it is, just say so. If it is a PSU problem, don't worry because that's actually really easy to fix and relatively inexpensive.

 

On to some details that should be helpful to you:

 

Normally, a drop or fluctuation of plus or minus 5% in voltage is OK. I think most people would prefer it on the polus side, but generally if it's within 5% you are OK. If someone could please verify, I believe the 12v rail has a somewhat different tolerance? Or was it one of the others?

 

Regardless, 5% of your high number of 1.637 is .082. Lets subtract that from your low number of 1.627 and we get 1.545. Thats still a little bit away from your 1.593 but pretty close.

 

Without knowing what PSU you have, I can only say it's definately worth investigating further or at the very least monitoring very closely.

 

There is a link or to in Supershanks' signature for PSU's or Power Supply Units, very informative reads there with many links to dozens of pages.

 

A state of the art system can be very demanding power wise, so you will find very many experts on power supplies here, ask anything you can't follow and many will be happy to explain.

[Guest LithoTech]

originals were bare then g-skill started clothing them so mine have got the spreaders.

 

re 704 i think the tables are improved over 623 maybe. This post reflects the sirt of thing. Can't get high read preamble you can also see in the original link the improvement in running a number of modules which were problematical.

 

i'm currently playing around with settings also with this Text Tutorial - TCCD DDR600 2.5-3-3-7 twenty four seven: The ultimate DFI settings

 

I think this is exactly right:-

 

Have been running with sp2004 for memory testing for a while.

 

Have got 275x10 Sp2004 stable 4 8.5hrs which is great - trying to get to a 273x11 that will be stable. Atm can get in there pretty well 1mb pi etc., HotCPU & all the benches but not prime stable.

 

I tried the yellow sockets b4 but came back to orange.

 

luck

 

Now I'm really perked to try 704. It's sure easier than pulling ram, although I still have to do that eventually.

 

I also have to see first hand how Tmod's disk works, and the whole flash and clear process goes. I even recorded all my reloaded settings last night in some sort of subconscious preparation.

 

Read that thread myself just last night. While it doesn't apply to my ram, the theory and rationalization of what he does to adjust and compensate in order to clock higher -- could still be applicable for me. At the very least, it would be fun to test and experiment. I've nothing to loose, and knowledge to gain.

 

I disagree about memtest, but it might be because it seems to work completely different for me. It either loads and works pretty much error free, or doesn't load at all. If I do manage to find a combination of settings that produce errors, there are tons of them and the ram heats up like crazy and I reboot immediately.

 

So it like, passes if it loads and gets no errors in an hour or so of test 5 and 8. Then I go ahead to XP. HexusPi first, then the modded one, then super pi 1m, 32m.

 

OCCT is next, I usually get an error from OCCT if the ram is in any way unstable, so I go no farther until it passes OCCT. In fact I often do OCCT before Superpi32, but I'll do Hexus and SuperPi1m and 2m because they are fast.

 

If it passes OCCT I'll play with some benchmarks, do some reading, surfing, work, let the machine fold in the background. I leave P95 till I go to bed. I'll let it run untill I have to shut it off the next day. That can last a while if need be, because I do have a second machine.

 

Essentially, if it won't load Memtest, it sure ain't gonna load XP. And once in Memtest, at least for me, if it is going to error, it does so very quickly. Any time I've left it for more than 12 hours, it's never errored.

 

Another thing it's good for, I had a stick of 512 generic that went bad on me after about 1 month old. I all of a sudden was getting P95 errors, and there were no changes made to the system or bios. I put the ram in another machine, same deal. Ran memtest on it, same deal, both machines. Took the ram back, store was giving me grief asking me if I did this and that, as soon as I mentioned memtest they shut up and gave me a new stick.

 

:cool:

[aowen512]

Thanks for the reply lithotech. Yea I guess my psu isnt in my sig(is now). I must have totally forgot to put that in there. I am using a 450w fortron from newegg. I have also seen this psu on some of the psu recommended lists so I figure its not all too bad. Fortron is a pretty good name.

 

http://www.newegg.com/Product/Product.asp?...N82E16817104954

 

This power supply seems to me atleast like it should be handling the load pretty well.

 

1 proc opty 148 currently at 2970 with 1.63 (1.65bios) volts

3 led 120mm yate loon fans

1 7800gt 470/1100

1 dvd burner

2 usb devices (headphones n printer)

2 sticks ram 2.62 (2.60bios)

and volts for northbridge is 1.71 (1.70bios)

no other devices such as pci, fan controller, front temp readout etc

 

Hope this helps!!

 

Thanks again

[Guest LithoTech]

Thanks for the reply lithotech. Yea I guess my psu isnt in my sig(is now). I must have totally forgot to put that in there. I am using a 450w fortron from newegg. I have also seen this psu on some of the psu recommended lists so I figure its not all too bad. Fortron is a pretty good name.

 

http://www.newegg.com/Product/Product.asp?...N82E16817104954

 

This power supply seems to me atleast like it should be handling the load pretty well.

 

1 proc opty 148 currently at 2970 with 1.63 (1.65bios) volts

3 led 120mm yate loon fans

1 7800gt 470/1100

1 dvd burner

2 usb devices (headphones n printer)

2 sticks ram 2.62 (2.60bios)

and volts for northbridge is 1.71 (1.70bios)

no other devices such as pci, fan controller, front temp readout etc

 

Hope this helps!!

 

Thanks again

 

Yep, pretty solid PSU, I have two of the 400's in other systems here. Although it's possible with that graphics card you may be pushing it, I'm thinking I saw some of those GTs or GTXs had a 500w min requirement. I'm unfamiliar with that motherboard or it's power requirements, but are you sure you have all the required power connectors hooked up?

 

I'd say 2970 isn't too shabby either! Damn straight you have all the power connectors hooked up!!

 

If you wanted to investigate it further, I'd follow the instructions in the beginning part of this post of Supershanks':

 

http://www.dfi-street.com/forum/showpost.p...&postcount=1269

 

You will need a multimeter (reasonable quality as cheap as $20, tons of good ones on ebay), and a passing knowledge of how to use one (tons of tutorials on the web). It's not a complicated device.

 

More info in these liinks:

 

http://www.driverheaven.net/guides/testingPSU/

http://forums.extremeoverclocking.com/show...ad.php?t=137886

 

Basically what you want to do, is test the lines with the multimeter when the computer is at idle and then at load, and see what the readings really are. The multimeter will be accurate, the readings from software are usually not.

 

You will then know exactly how close to the 5% limit you are.

 

If you come to the conclusion that it might just be not enough power to run your rig, there is an easy way to check. Especially if using a multimeter. Simply disconnect everything in the box except the bare minimum to boot up and run:

 

CPU fan

One Hard Drive

One stick of ram

One video card

KB and mouse

 

Yank the power to everyting else, everything.

 

Boot up and load the cpu and see if you get the same fluctuation. Chances are you will, and that doesn't mean the PSU is bad, it just means that it fluctuates that much under load.

 

But, if it doesn't fluctuate anymore, then you know that you are pushing the limit!

 

It's sort of odd that with such complicated devices as modern computers, very often troubleshooting things boils down to the simplest form of process of elimination.

 

I imagine a lot of repair shops would dispense with a most of the elimination process by going for the jugular right off. If they suspected the PSU after testing with a DMM, they'd hook up a spare PSU and see if that helped. If it did, then it proves it's the PSU. Thats what I'd do here, I have a spare PSU on the shelf for just such a matter, but it's such a POS I'd never put it in someone elses computer for fear of it blowing up.

[Squid1]

So can anyone get back to me about my core volts dropping about .040 when I run something stressful like occt and typically jumping about .015 idle. Is this a very bad drop or should I not be concerned? I am not sure how this volts normally acts as I am not too fimiliar with this platform.

 

thx fellas

 

Didn't see your original post, but the SLI-DR/Ultra-D do not have their sensors reflec the true voltage in OS readings for the load/idle part.

 

In reality, there boards pump 0.05 V into the CPU when the CPU is busy, compared to idle. The sensors do not show any increase during this (apart from the fact that they are off, too low, anyway).

[supershanks]

I've actually picked up an increase in vcore under stress when running OCCT.

see OCCT which tends to support the findings of Martin & others.

 

However the accuracy of any sag/peak using an 8bit sensor value ,may well br overstated/distorted see -

soundx98 - Question: My Voltage is a bit off Posts #17 & 18

 

luck

[Clay]

Yesterday I was running @290x10 had it nearly totally stable. But it failed prime after 59min. Well I wasn't totally pleased with the memory bandwidth I was getting @that speed so I decided to Load Optimised and start OCin' all over. Running stock speeds I decided to do a few benches. I clicked on everest-reads and got an instant restart.. (Aquamark3 also restarted the mechine)I've got a few restarts in the past with this rig.(But I allway tweaked a settin' or upped a voltage or 2# problem went away til the next move up in fsb)

But at default settin's restart's ?LOL Anyway I wanted to know if anyone else has had this problem. Is it a bios-opteron thing? I'm currently using the 6/23 bios that came on the board.

I did a very nice job puttin' it together. (one of my best) I really don't think it's a hardware problem. Well anyway thanks in advance."Clay"

 

Update: I flashed to Tony's 7/04 bios.(Nice bios Tony!) Loaded optimised and promptly failed a 1meg superpi! Hehe it seem my rig doesn't want to run at default settings.(Except memtest) Default changes trc to 16 and trfc to 24. So i changed them to 7 and 14 respectivly and still failed a 1meg superpi before the first loop. Then keeping all the setting the same. Except 275X10 1.375 +110 vid everything is perfectly stable. She didn't start running right til I'd clocked up to 240x11. Makes me wonder LOL Kinda' weird my rigg doesn't want to run straight til' it's got a pretty decent ocverclock on it. LOL (I think I like it!)

Has anyone else had big trouble's with running at optimised settings?

 

Edit: I wonder if the memory doesn't like running slow. But it memtest prefect @200.

[aowen512]

Thanks alot for the replies fellas. Helps me out a lot. I think im going to test my psu with a multimeter, and go from there. I know that 2970 is pretty good, but I just want to get that 3gig mark so damn bad!! My stepping is really good, and with my cooling etc I feel that I should be able to get there. Time will tell, time will tell.

 

I will reply back once I have tested my psu.

Have a good one guys.

[douthers]

I've finally got my rig to a level that I'm happy with My 146 was at 2.9 and was 8hrs + prime stable, but that was putting my Gskill ram at a level that it didn't like and Memtest was returning errors galore . I've now dropped back a touch to 2.85 and things are looking good. Prime stable / Memtest stable / Gaming stable :nod: Temps under load are mid > high 40's in a warm room. I've decided to take AG's advice and not spend every waking hour looking for tiny improvements on this set up. It's doing everything I want of it and for a £160 cpu that's not bad going......

 

Settings as follows: -

 

Opteron 146 Cpu @ 2.85ghz / 2Gb Gskill PC4000 HZ Ram @ 259mhz

 

 

 

Gpu - Nvidia XFX 7800GT @ 470mhz / 1.09ghz (84.21drivers)

 

 

 

FSB Bus Frequency > 285

 

 

 

LDT / FSB Bus Frequency Ratio > x 3.0

 

 

 

CPU / FSB Frequency Ratio > x 10.0

 

 

 

K8 Cool n Quiet Support > Disabled

 

 

 

CPU VID Control > 1.450v

 

 

 

LDT Voltage Control > 1.30v

 

 

 

Chipset Voltage Control > 1.60v

 

 

 

DRAM Voltage Control > 2.60v

 

 

 

DRAM CONFIGURATION

 

 

 

DRAM Frequency Set > 180 / 09/10

 

 

 

CPC > Enabled

 

 

 

Tcl > 3.0 Bus Clocks

 

 

 

TRCD > 04

 

 

 

TRAS > 08

 

 

 

TRP > 04

 

 

 

TRC > 08

 

 

 

TRFC > 17

 

 

 

TRRD > 04

 

 

 

TWR > 03

 

 

 

TWTR > 02

 

 

 

TRWT > 04

 

 

 

TREF > 4708 cycles

 

 

 

DRAM Bank Interleave > Enabled

 

 

 

Dqs Skew Control > Auto

 

 

 

Dqs Skew Value > 0

 

 

 

DRAM Drive Strength > Auto

 

 

 

DRAM Data Drive Strength > Auto

 

 

 

Max Async Latency > 08.0 Nano Seconds

 

 

 

Read Preamble Time > 06.5 Nano Seconds

 

 

 

Idle Cycle Limit > 256 Cycles

 

 

 

Dynamic Counter > Enabled

 

 

 

R/W Queue Bypass > 16x

 

 

 

Bypass Max > 07x

 

 

 

32 Byte Granularity > Disable (4 bursts)