Opteron 165 on NF4 Ultra-D overclocking query

[renzokuken]

Before I start, here?s my system:

DFI LanpartyUT nF4 Ultra-D (BIOS 704-2BTA)
AMD Opteron 165 @ 1.8ghz (0551VPMW)
Thermaltake Bigwater SE water cooling kit
2x Legend 512MB PC3200 RAM (generic..for now)
Nvidia 7800GT 256MB PCI-E x16
WD Raptor 74GB
BenQ Black DVD+/-RW DW1650
Enermax EG565P-VE FMA (24P) 535W ATX2.01

Hiyas,
Recently I?ve been tinkering with my Opteron setup in an attempt to get a bit more out of it. Actually, the truth is I originally built the system with overclocking in mind, but never ended up getting around to it after college started. The good news is that I?ve now got a solid 2 weeks off study and can focus my attention to overclocking this sweet sweet rig of mine. The bad news is that it?s giving me a [content edited] of a hard time. To begin, I?d like to share with you all what I?ve discovered about my system and present some results in tabular format:

http://img87.imageshack.us/img87/686/results1wk.jpg

A few points I need to make clear here is that for all tests, I was using a CPU multiplier of ?9? (the maximum supported multiplier for my processor). All voltages are at stock settings. I've tried both sticks of 512MB RAM independently and have tried them in both the yellow and the orange slot of the motherboard. Note also that all timings are at stock (auto) setting, with the only change to the DRAM menu being the divider setting. I've tried to get it to boot at 2T command rate. And finally, all four (4) power connectors are plugged into the motherboard.

The reported Mhz by CPU-Z for each test is accurate according to the following table:


http://img114.imageshack.us/img114/4057/truedividers4ci.jpg


In Figure 1, the first 3 tests are attributed to finding the maximum stable speed my RAM. I sort of knew it was around DDR460 from when I first got my system a couple of months back and was doing some preliminary testing. As you can see, the maximum stable DDR speed is DDR458. Similarly, I know (from earlier tests) that the maximum stable overclock on stock voltage for my CPU is around 2.65ghz and my motherboard will boot quite happily at FSB?s at and above 300+.

Looking at test #4, 5, 6, 7 it can be seen that my system boots up until a FSB of 275 and then it fails. This is what I don?t understand. At a FSB of 275, my DDR400 RAM is running below spec at ~DDR354, my CPU is only at 2.475ghz, and my motherboard is cruising along at a HTT of 1650 (stock rated at 2000). I find it highly unusual that my system has trouble booting at this relatively ?easy? overclock. Take a look again at my system details above and you?ll find my cooling solution and power supply are more than adequate. Does anyone out there know what could be causing the problem?

Next, take a look at the results for test #4, 8 and 9. The system boots and is stable at the 2:3 divider and the 3:4 divider. In both instances, the RAM is running below spec. At the 5:6 divider, the system does not get past the CMOS update screen and will not boot. In this case, the RAM is only very slightly overclocked and well within the maximum stable DDR speed of DDR458 tested earlier. Again, the CPU and Motherboard are well within their limits.

From these results I can only conclude that Opteron 165?s don?t like dividers, or don?t like my RAM. Someone from another forum told me to try loosening the Read Preamble (RP) and Max Async Latency (MAL) in the BIOS RAM settings and see how it goes, but when it still didn't boot at those specific dividers after making them as loose as the BIOS would let me, I assumed the problem lie somewhere else. Unfortunately, most advice I?ve been given has been from Opteron 146/148 owners and not Opteron 165 users. It?s clear that Opteron 165?s behave very different from the single core counterpart.

Unless there?s something out there about Opteron 165?s that I haven?t read about yet, I?m gearing up to buy me some new RAM. Why? Because I?ve seen people will similar rigs (except they?re using brand name RAM) and the exact same stepping CPU (0551VPMW) utilize the same dividers I?m using without any issues. This fact alone leads me to believe that the culprit in my system is this generic RAM. Again, unless I?m not tweaking a setting in the BIOS that I should be, I can?t see what else it could be. So please, I?m begging all of you fellow overclocking enthusiasts, if you know or you can suggest a possible fix to this evil evil problem I?m having please lend me your knowledge.

If you see this same post on a different forum, it is because I have gone around to all the different forums to reach as many people as possible for a solution to this problem so that hopefully one day this system of mine can finally reach its full potential.

Any help is greatly appreciated.

 

[AMDxx165xxfx60]

www.dfi-street.com good tutorial on how to overclock your DFI there, test your max HTT by using the lowest divider and most stable voltage , and keep giving your cpu voltage up to 1.5v. i wouldn't go any higher then that, if you end up hitting at least 290FSB/HTT then it's your ram.

 

[renzokuken]

I've tested my RAM up to DDR458, my CPU up to about 2.65 on stock voltage and my board up to 350HTT. As I said in the initial post, I've posted this on various other forums, including dfi-street but I am yet to recieve any advice from dfi-street. It seems nobody really knows how to solve this very odd problem...

 

[jiffylube1024]

Increase your voltages man!

First of all, you're trying to find your max FSB and stable RAM speed all in one go, BEFORE you even know your max stable CPU speeds - a big no-no!

First, find out your max CPU speed: drop your LDT multiplier to 3X, increase your LDT voltage 2 notches, and drop your RAM dividor so it becomes not a factor (1/2 dividor works well). Start by setting the BIOS to 1.475V. This will give you 1.44V at full load on the DFI board with a good PSU, and most users get 2.4-2.6 GHz out of their chip at this voltage setting. If you wish to go further, bump it up to around 1.5V in Windows (give it ~1.525 or 1.55V in the BIOS) and test it with Prime95 or another CPU test.

Once that's done, then you can find your max HTT speed and then your max stable RAM speed. By trying to do everything all at once, you're doing it all wrong...

 

[Munky]

I remember reading some posts from users about the DFI board not liking cheap RAM (Corsair VS rings a bell, in particular...), so it could be your mem that's causing the problems. Also, which test do you run to determine stability? To test the mem I usually run the built in Memtest utility. To test the cpu I run Prime95.

 

[JSFLY]

I can tell you have a 4.0 gpa ;-p

DFI boards are picky indeed. Get rid of that ram pronto! Its like Making a spoiler out of cardboard and putting it on your ferrari.

I suggest the G.skill DDR500s on newegg.

Also These boards are weird sometimes. Eg: I can't boot at FSB 280 but I can at 279 or 281. Weird.

 

[renzokuken]

Actually, I've already found the max of each of my components. I went about doing so as you suggested by eliminating 2 of the 3 components in each test and then finding the maximum of each. The maximum for my CPU is 2.56ghz on stock volts, for my RAM (1 stick in orange slot) its DDR458 and my motherboard hits as high as 350HTT on the LTD 1x, RAM/FSB 1/2 divider, CPU set to 5x multiplier. So I know the limits of my components. In fact, I DID say this in my initial and second post but I guess you didn't read it or misinterpreted it. Either way, from responses on other forums and by doing more reading the problem here is almost certainly associated with the RAM.

Originally posted by: JSFLY
I can tell you have a 4.0 gpa

What does this mean?


Since posting the intial thread, I've changed some memory settings in the BIOS. Take a look at test #7 in Figure 1 (the first link). I no longer get the BSOD on boot. No doubt this is great because it means I'm doing something right, but doesn't it seem strange that on test #7 the RAM is running below its rated speed and yet it was giving me a BSOD? If it runs stable at DDR400 on tighter settings than I'm running it now (underclocked), how does it make sense that loosening the settings made it more stable? it's running underclocked, and it was running stable at tighter settings at a faster speed. I can't really think of an analogy to make what I'm trying to say more clear, but if you just think about it for a sec, it doesn't make any sense. Theoretically I should've been able to TIGHTEN the settings even more because I'm running the RAM at a slower speed but that isn't the case for some reason....could it be that dividers screw around with the effect of timings? Perhaps running a module at DDR400 in a divider is like trying to run the same module at DDR500 1/1? Can anyone shed some light on the subject?

 

[Elfear]

The best way I found to getting the most out of my chip was to first get some ram that a lot of people had tinkered with so that I could have better starting point with the slew of dram settings the DFI's have. You probably don't want to spend more money on ram so I'd do a little research (you should be good at that) and see what kind of modules Legend uses. Than look over some posts here or here to try and find someone who has ram with the same modules and has tinkered with dram setting a lot.

That will get you very close as far as where those settings need to be. The drive and data strength are especially important.

From there it's just testing little by little until you find your max oc. Like jiffy mentioned, you'll probably want to bump your vcore up a little to get much higher than 2.56GHz. My brother has the same stepping as you and is Prime stable at 2.6GHz@1.35V. He needed 1.45V to be Prime stable at 2.7GHz.

EDIT: Doh, just noticed you posted over at the exact place I linked to.