Epox 8RDA+(nf2)@247fsb, need help to get to 250

[***Deimos***]

Hello folks,

Might remember me as the skeptic who doesn't believe in burn-in. After 6hr+ of "burn-in" in the form of looping Memtest86 1.51, I was still stuck at 218. A bit higher sometimes worked, but randomly gave errors. However, I made a breakthrough today, when I was searching for a jumper to make one of my 2 drives master (neither has jumper (slave), so I was searching for an extra on the mobo).

The JCLK jumper was not set, so I thought I would give it a try. Before you know it, I am up to 224. Epox 8RDA+ v1.1 with 1.76V vdd, XP2500+ using 8.5x and generic Samsung 512MB DIMM, (using 2.5-3-3-11 auto precharge), I got pretty cheap.

Max 1:1
Next, I tried CAS3 but this DIMM doesnt like that. However, 2.5-4-4-11 worked like a charm.. all the way to 236Mhz (Memtest86 1.51 test#5-3runs, SuperPi 1M, and Windows Memtest-up to 30% coverage used to test). At 238, I tried bumping up the Vcore on the CPU, the Vagp, CPU interface Optimal instead of Agressive, turning off auto precharge, widening timing to 2.5-5-5-14, trying lower multipliers, but nothing helped. I even tried putting some aluminum foil covered with seran wrap on one side, around the memory chips, and the north bridge area to provide some "grounding planes", but alas it didnt help.

Max fsb
From previous experiments I know that 75% and 60% dividers work best for high fsb. Previously got it up to 240/75 and 244/60 (before I figured out the JCLK jumper). Since memory is <<200Mhz, timings don't really matter, so I didnt change them from 2.5-4-4-11. Set 8x to eliminate CPU prob, and got up to 240/75 and 247/60. BUT, the strange thing is, SuperPI, Memtest86 1.51, Windows Memtest, they all pass flawlessly. However, 248,249,250 all cause NO POST. This is strange because problems always creep up as a few errors, escalate to many, and then finally no post over at least 3-4Mhz range.

Why would the mobo suddenly fail over 247fsb?? Any suggestions?

Currenly trying to "burn-in" at 224 1:1 2.5-3-3-11 with SuperPI 32M and Windows Memtest. Is it better to burn in at 224 with tighter timings or 247 with async mem at 149Mhz?


BTW: nf2 Vdd varies 1.744-1.76, and Vdimm varies 2.912-2.928, and without additional mods, cant raise either. CPU is not a factor since <2.0Ghz, and works up to 2.47Ghz.

[saaya]

more vdd and better nb cooling, and try the l12 mod

[***Deimos***]

more vdd and better nb cooling, and try the l12 mod

Yeah, I think you are right about more vdd.. but captain.. I'm giving her all she's got. I dont have my soldering stuff here on campus, and too expensive to go home just to get it (might as well buy a new mobo for the price).

I would agree with you on the cooling, except, even at 6am, when the air is still cold, with cold start, the same problem exists.

The L12 mod sound interesting... although I thought it only affected certain processors, and was reason they couldnt run 200fsb.. and my can go all the way to 247fsb.... I'll google it... btw: do you know any "magic" bios for the 8RDA+? I thought it was the problem so updated from 3827 to 4729 -> no effect. The new BIOS has 9x multiplier, but it is totally defective.. anything >200fsb and it resets BIOS... totally useless.

[***Deimos***]

L12 Mod at xtremesystems.org

it works also for me : 225Mhz before mod then 240Mhz stable
It helped some guys too here

But , you must have in mind : you will lose the 9.0X multiplicator , because all AMD 166Mhz/200Mhz cpus do not support 9.0X

it seems I already have this MOD. It also says its only for 133fsb CPU, I have 166fsb Barton...

This is very similar to my 166MHz XP2600+ chip on an Epox 8RDA+. I can not run high FSB and will always crash entering windows with the 166FSB jumper on past 190MHz. I take the jumper off and the chip defaults to a XP2000+ and maximum stable FSB of 225MHz with ease.

Strange that in my case, JCLK jumper off it was stuck at 220, but with it ON (eidt) I got up to ~236 (1:1).
In that thread, people who cut L12 on Barton couldn't get it to boot anymore

[***Deimos***]

bump

[gulp35]

L12 Mod at xtremesystems.org
Strange that in my case, JCLK jumper off it was stuck at 220, but with it off I got up to ~236 (1:1).

What?

I just got the RMA of my 8RDA back from epox and I am trying to see if this one overclocks any higher than the last (max 189FSB stable), as of now I am running at 12.5x203 on my 1700+ DLT3C 0310@1.825volts (1.8 set in BIOS b/c my board overvolts)

If this is an older board from Epox I would check the capacitors for leakages, because several of mine were puffy and leaking stuff....

Sounds Like you need some better memory, the Mushkin Blue Stuff is pretty Cheap and it has been doing well. You may want to try that.

[***Deimos***]

What?

I just got the RMA of my 8RDA back from epox and I am trying to see if this one overclocks any higher than the last (max 189FSB stable), as of now I am running at 12.5x203 on my 1700+ DLT3C 0310@1.825volts (1.8 set in BIOS b/c my board overvolts)

If this is an older board from Epox I would check the capacitors for leakages, because several of mine were puffy and leaking stuff....

Sounds Like you need some better memory, the Mushkin Blue Stuff is pretty Cheap and it has been doing well. You may want to try that.

Obviously you haven't read the whole post so I will clue you in.
CPU agressive, 100% divider, Vdimm=2.912, Vdd=1.76, Vcore=1.53, auto precharge on, spectrum off, 3827 BIOS: 2.5-3-3-11: 223
2.5-4-4-11: 236
using 75% divider: 242 (mem at 181)
using 60% divider: 247 (mem at 149)
Using Memtest86 1.51 #5, SuperPi 1M/32M, Windows Memtest to test stability.
doesnt like CAS3, and 9x multiplier doesnt work.
Also, these are fsb in BIOS, in real life CPU-Z shows +1 higher.

Now, using default Vdd=1.55 this same stick of generic Samsung 512MB could only go up to 210, regardless of timings. But, the problem is that although 247(60%) is 100% stable with any program, 248 and higher cause system to immediately fail to POST. If I go over 223 with 100% 2.5-3-3-11, it gradually starts failing, first SuperPI 32M, then Windows Memtest, then SuperPI1M, then Memtest86 1.51 over 3-4Mhz... ie, I can still boot with 227Mhz...

[gulp35]

No, I read your post, but you said that with the Jumper off you got 220 and then said with the jumper off you got 236. That does not make sense.

From what I've read on other posts, you should never run async on NF2 Mobos because it causes drops in overall performance and has higher chances of causing BIOS corruption.

Have you tried using a clockgen like 8rdavcore?
Another possiblility is that you have reached the limit of your board, though it is wierd that the errors come up so suddenly.

[mcnbns]

I agree with gulp35. Running asynch is a bad idea, even if you can get 30-40MHz higher FSB.

If it makes you feel any better, my NF7-S doesn't like going above 225MHz stable, even with every VDD, VTT and VDIMM mod known to man. That's why I'm upgrading.

[***Deimos***]

No, I read your post, but you said that with the Jumper off you got 220 and then said with the jumper off you got 236. That does not make sense.

From what I've read on other posts, you should never run async on NF2 Mobos because it causes drops in overall performance and has higher chances of causing BIOS corruption.

Have you tried using a clockgen like 8rdavcore?
Another possiblility is that you have reached the limit of your board, though it is wierd that the errors come up so suddenly.

Thanks for the very good advice. But, I haven't been able to SuperPI 32M in a whole week. First thing I did is lower the fsb.. by 5, by 10, by 20! Still cant do 32M. Tried 1.5V 6X 166... thats only about 996Mhz. Wonder what is going on... Even tried different BIOS. Then I noticed it:

If this is an older board from Epox I would check the capacitors for leakages, because several of mine were puffy and leaking stuff....

I'm surprised its running at all. All 4 capacitors to the left of the MOSFETS (closest to the bracket), are leaning over, due to their buldging bases ( and tops ). So, here I am writing this at 960Mhz. Trouble is, this is already a RMA mobo, since the first one just decided to die overnight one day (while I was dl some stuff overnight). I also added a SB heatsink which I'm not sure if they will like or not.