For the love of... Help w/200+fsb

[Sokratz]

Hiya folks, finally starting to bring everything together on my end. This is what I have and what I've done.

Mobile 2500+ (IQYHA 0348 SPMW)
-DangerDen RBX
NF7-S 2.0
-This L12 mod http://downloads.modlabs.net/bios/pic/L12mod.jpg (I think this is the correct l12, when I load optimized defaults, it goes to 166)
-D23 3D-Fire Bios from TicTac
-Heatsinked and fanned all the Mosfets
-Put active cooling on southbridge
-Heatsinked SATA chip
-Slapped a monster copper heatsink (no fan) on northbridge
1GB Mushkin Level 1 (sigh, wish I would have opted for level 2 here) pc3500 2-3-3-? (btw, its' confirmed that 2-3-3-11 is actually a tad bit faster than 2-3-3-6 on the Nf7-S)
9800 pro
-DangerDen Maze4
-Heatsinked and fanned fall ram chips
Antec True430
-Rails (from winbond)
1.81 - 1.84 (@ Vcore 1.85)
2.82 - 2.83 (@ 2.7 Vdimm)
3.23 - 3.26
4.92
11.74 - 11.80
-12.03 - -12.11
-4.94 - -4.99
4.97
3.49

Temps under load at 1.85v
CPU 41
Case: 28

Here's my problem (and I *guess* it could be in my head):
I can run 13x200 (Vcore 1.85v, Vdimm 2.7v, Chipset 1.7v, AGP 1.5v, Ram @ 2-3-3-11, FSB/DDR Ratio 5/5, CPU Interface Enabled, all %'s disabled.) I can probably do this prime95 stable for hours and hours but as soon as I try to go above 200, prime95 quickly fails. I've tested the ram with memtest86, and have tried other ram (some hyperx but it's hard to compare) and don't see any problems.
I believe I can also run 13.5x195 (at same voltages/timings)(what i used to install winxp w/) but haven't long-term tested it.

13x204 was apparently stable when I used the Nvidia utility, but once I set it like that in the bios it crapped out.
12.5x210=unstable.
12x210=unstable
I need to try like 4x210 to see if that's stable just for shits I guess.


Any ideas?
Thanks much,
Sokratz

 

[myocardia]

Most chips have a limit on high high they will go fsb-wise. Since their isn't a 4x multiplier, try something more "normal", like 8 or 9x210. If it still isn't stable, you found out how high your chip will go. It may very well be only 200 or 202fsb. BTW, you're complaining about 2.6ghz? Now, you just need to see if it's Prime95 stable. I bet it isn't, especially if you do the custom torture test with 2048 minimum fft's and 4096 maximum fft's.

 

[magratton]

This is EXACTLY what I have also found. Anything above 200mhz fsb and suddently Prime95 unstable. I have, for all intents and purposes, the exact same setup (sans the elaborate cooling additions, etc, I do have the Thermalright slk974u with vantec tornado 92mm fan). I cant even do 202mhz (regardless of voltage and multiplier) and be prime stable with large (2k+) ffts for 4 hours. @200fsb I am rock solid no matter what I throw at it.

Thought it was all in my head as well Sokratz!

 

[magratton]

Originally posted by: myocardia
Most chips have a limit on high high they will go fsb-wise. Since their isn't a 4x multiplier, try something more "normal", like 8 or 9x210. If it still isn't stable, you found out how high your chip will go. It may very well be only 200 or 202fsb. BTW, you're complaining about 2.6ghz? Now, you just need to see if it's Prime95 stable. I bet it isn't, especially if you do the custom torture test with 2048 minimum fft's and 4096 maximum fft's.

this is thread hijacking, but just for a moment: This needs to be added to some "overclockers guide" somewhere. I definitely agree that running P95 in custom mode using high fft's (2k and above) will highlight your issues MUCH sooner than running any of the other canned modes. You will get your answer within several minutes vs. many hours.

 

[Verdant]

put me down for another unstable above 200...
i have an NF7 with a 2600+ which runs seemingly completly stable at 13x215, no crashes, etc for a few days, even under intenssive usage... but folding at home wus complete in error.
they will not complete at anything above 200 mhz... i will have to try at lower multiplier though (i have tried at 12x)
one thing i noticed which i found odd , was a lack of a 201 mhz fsb option.. it goes 199, 200, 202, 203... which makes me believe frequencies above 200mhz may be generated in some other way...

 

[myocardia]

Well, mines actually stable for over 12 hours with 2048 minimum and 4096 maximum at 12x209=~2505 mhz. And I'm pretty sure that it's either my memory or motherboard holding me back from a higher fsb, because when I raise my ram timings to 3-4-4-11, it's fairly stable at 11x222. But, I also have to raise my vdd to get that high, and since I've got an 8RDA3+, I have to raise it from 1.6v to 1.8v, which I don't like doing. 1.7v vdd would be fine, but 1.84ish isn't, in my book.

 

[Sokratz]

What are your folks' rails? I'm very tempted to do the PSU 3.3 mod.

Check http://s6.invisionfree.com/underclock/index.php?showtopic=31 and
http://www.bleedinedge.com/guides/psu_3.3v_mod/psu_3.3v_mod_pg1.html

 

[myocardia]

Linked: link #1 and link #2

 

[Verdant]

i am running a cheapo antec 350W SmartPower psu
my voltages are:
4.95 V
11.85 V

my cpu voltage is set to 1.8


i got probably nudge this machine higher, but it is being used as a SQL server for development... as well as folding... so cpu power isn't that important... and its rock solid now for 14 days no reboots.

 

[Sokratz]

Originally posted by: Verdant
i am running a cheapo antec 350W SmartPower psu
my voltages are:
4.95 V
11.85 V

my cpu voltage is set to 1.8


i got probably nudge this machine higher, but it is being used as a SQL server for development... as well as folding... so cpu power isn't that important... and its rock solid now for 14 days no reboots.

Can you list your 3.3 rail?

 

[Verdant]

Originally posted by: Sokratz

Originally posted by: Verdant
i am running a cheapo antec 350W SmartPower psu
my voltages are:
4.95 V
11.85 V

my cpu voltage is set to 1.8


i got probably nudge this machine higher, but it is being used as a SQL server for development... as well as folding... so cpu power isn't that important... and its rock solid now for 14 days no reboots.

Can you list your 3.3 rail?

oops! lol
its given as 3.37v

 

[myocardia]

Yeah, the reason you're able to clock that high with that little psu is because you don't have a 9800 Pro video card in it. If you did, you would have run out of power much sooner.

 

[Sokratz]

Originally posted by: Verdant

Originally posted by: Sokratz

Originally posted by: Verdant
i am running a cheapo antec 350W SmartPower psu
my voltages are:
4.95 V
11.85 V

my cpu voltage is set to 1.8


i got probably nudge this machine higher, but it is being used as a SQL server for development... as well as folding... so cpu power isn't that important... and its rock solid now for 14 days no reboots.

Can you list your 3.3 rail?

oops! lol
its given as 3.37v

so what's your multiplier and fsb currently? Are you having the same 200fsb problem?

 

[magratton]

NF7-S v2.0 mobo
AXP2500M - 200x12@1.65v
512 mb Mushkin PC3500 Level 1(2 x 256) in slots 1 and 3 for dual channel - Memory set to 2-3-3-11
Bios Settings:
- All % set to diabled
- Video and Bios shadowed
- CPU Interface enabled
Gigabyte Radeon 9600Pro
Antec TruePower 430 (3.3v rails are fluctuate between 3.28 and 3.31)

Sokratz, I believe the only diffs between you and I is I did not do the cooling mods to mobo and you are running 2x512 mb of the same ram.

So, I adjusted my settings: 8x210
Raise my vCore to 1.8 (1.76 - 1.78 actual) and my vDimm to 2.9 (2.99 - 3.01actual)
Prime95 is running stable for now, but still only within a few minutes testing.

 

[Verdant]

Originally posted by: Sokratz
so what's your multiplier and fsb currently? Are you having the same 200fsb problem?

13x200, if i jump too 202 or higher on any multiplier folding wus are corrupted, windows seems stable upto around 211 though

 

[magratton]

So poked around a bit this morning after running prime stable (large ffts) all night long. So we have the following (all of which are at 210x8 @1.8v vcore:

vdimm at 2.9 (2-3-3-11 timings)
prime stable

vdimm at 2.8 (2-3-3-11 timings)
prime failed

vdimm at 2.8 (2-4-4-7 timings, recommended by spd)
prime failed

vdimm at 2.8 (2.5-4-4-7)
prime failed

My take on this is when you drop the vdimm below 2.9 then the memory is not stable enough at >200fsb. Mushkin's web site says yeah it should go to 2-3-3 PC3500 but I noticed that they say they only test up to 200fsb and "overclocking is not supported".
Others on this site have said that mushkin ram "eats voltage for breakfast", perhaps this is really our limiting factor. I will try keeping the vdimm at 2.9 and raising the multiplier to see where I can get to later tonight when I get home.