2.4c voltage questions.

Backstory:

1.8a northwood (18x100) running at 18x175 @ ~3.2 ghz, I think it was at 1.65 volts.

I recently snagged a SL6Z3 2.4c from ebay, and right now it's running 12x266 @ ~3.2 ghz at 1.575 volts, 5:4 fsb:ram (ddr466, I haven't tested ram capability seriously yet, but I was getting artifacts in Stronghold2 with 1:1 and 2.7 volts so I set the divider to 5:4, artifacts went away.)

Basically I've never pushed voltage hard on a processor or ram before, and I want to know what sort of limits I should be working with. I've seen people say they run ddr at 3.2 volts, so later this week I'll probably start some tests with various voltages to the ram (2x512mb corsair ddr466, spd rates it at cas3 at those speeds). Decent DDR isn't terribly expensive now, so if I cook it I'll be unhappy but I can get more.

The 2.4c on the other hand, it's a used part and I really don't want to burn it up. What's crossing a line with this processor? 1.575 right now @ 3.2 ghz leads me to believe I have some headroom here, but how much?

Cooling is an XP120, 4 heatpipe job with a 120mm fan. Thermal-control is enabled in bios but I have it set fairly aggressive, the fan goes 12v @ 44c. Case cooling / airflow is good. (full size tower, several 80mm fans pushing air in and out.)

http://www.ocforums.com/showpost.php?p=4661881&postcount=103

Seems the max safe Vcore was 1.65ish (10%) for northwoods to prevent the SNDS.

With DDR it really depends on what IC's you have.

Ok, sounds like a reasonable risk. I still have some headroom on vcore to push for a bit more.

Honestly, I expect the motherboard's reaction to a pumped up FSB to be the limiting factor here. I'm pretty happy with the 3.2ghz I'm running now, but if I can get 3.4 or 3.5 with stability, then I'll be all over it.

And I guess I can always put the 1.8a back in if things go sour, and wait until spring when I plan on upgrading to a core2 duo (or amd rig if amd is impressing me via bang-for-the-buck.)

Thanks for the response.

The motherboard should be fine, my 2.4c done 292FSB on my AI7, as well as many others. 1.65 (probably 1.700 - 1.725 in bios) should be fine. Being an M0 you should get a nice jump in speed from it. You may not even need that much voltage for 3.5 say.

Also, if you have a great CPU and you get to 300FSB the pci/agp lock will no longer work, therefore you'd need to boot at say 295 then clock gen the rest of the way if I remember right.

Jimbob7 said:

1.65 (probably 1.700 - 1.725 in bios) should be fine. Being an M0 you should get a nice jump in speed from it. You may not even need that much voltage for 3.5 say.

Click to expand...

please do not reccomend using more than 1.65v on a northwood. SNDS is a very real phenomenon, and using 1.725 in bios WILL kill that chip. maybe not over night, but it will stop overclocking like it used to, and will eventually not run at stock speeds without 1.65v.

my sl6z3 needs 1.625 in bios to do 3.4ghz stable. it gives me 1.6ish in windows as read by mbm5, and droops down to about 1.575 under load. has been running at 3.4-3.5ghz 24/7 since the day i bought it, almost 3 years ago.

i can run 300fsb for 3.6ghz with 1.675 set in bios. it is not prime stable, but i can game, run benchmarks, etc, as long as i can keep the cpu cool (not very long in the summer, unless i throw an unbearable amount of air at it...very noisy). this is on my ic7max3.

No, i'm saying to get 1.65v he may need to select 1.700 to 1.725. I'm not at all suggesting he uses more than 1.65.

Jimbob7 said:

No, i'm saying to get 1.65v he may need to select 1.700 to 1.725. I'm not at all suggesting he uses more than 1.65.

Click to expand...

i see what you are saying, but let me tell you how it works with my chip:

i set 1.675 in bios, i get (usually) 1.65v in windows, and maybe 1.63v under load. the 1.65v in windows is not static. it does fluctuate wildly. it can jump up to 1.68v at times, mostly it idles at 1.65v. i have tried setting 1.7v in bios, and it idles at 1.675 (mostly). this is still dangerous for the chip, and can quickly degrade my oc. when setting 1.7v in bios, it NORMALLY is at 1.675v, but again, it can fluctuate up to say 1.71v (as read by MBM5, have not measured with my DMM, it could be waaay higher).

really, better safe than sorry, he really shouldn't need more than 1.65v to get his max oc. setting 1.7v in bios does absolutely nothing for me, it does not make me prime stable at 3.6ghz. i can still happily bench at 3.6ghz with 1.675v, there is no need to risk more.

didn't mean to sound like i was jumping down your throat, just wanted to clarify that just because it droops doesn't mean that it won't momentarily be at dangerous levels.

I'm stable at 3.2ghz at 1.575 set in the bios. With 3 more ticks to play with, I'm pretty sure I can hit 3.4 or 3.5 then. I just have to figure out the best voltage and fsb:ram ratio for the memory.

Ah brings back memories of my old 2.4c.

If you want to get rid of the artifacts and run 1:1 with your memory screaming, then download i865/i875/i848 Tweaker, and then set the DRAM Latency Parameter from 00 to 01

You also have to remember that that chip is not really a Northwood, but a Galatin core. These are server grade chips most of the time, and I'm guessing could theoretically withstand more voltage pumped into them than a NW.

microfire said:

If you want to get rid of the artifacts and run 1:1 with your memory screaming, then download i865/i875/i848 Tweaker, and then set the DRAM Latency Parameter from 00 to 01

Click to expand...

Do you mean from 100 to 101?

edit - i just googled the tweaker and found the post from the person who wrote it, just want to clarify before I start diving in, not knowing what i'm doing

edit #2 - that's exactly what you mean, I see now. The dropdown list to change that parameter has 00/01/10/11 as the options, short for 100/101/110/101. I seee...

Yea give it a try, I had two AI7 mobos and never could fix the artifact problem till I discovered the tweaker(s).
DRAM Latency Parameter:
i865 chipset = 100 by default (That is why they artifact.)
i875 chipset = 101 by default (That is why they don't artifact.)

One thing I swear by is running 1:1, even if you have to drop some CPU speed for raw memory speed. Even so, you now have the option of both worlds. Have fun overclocking

microfire said:

Yea give it a try, I had two AI7 mobos and never could fix the artifact problem till I discovered the tweaker(s).
DRAM Latency Parameter:
i865 chipset = 100 by default (That is why they artifact.)
i875 chipset = 101 by default (That is why they don't artifact.)

One thing I swear by is running 1:1, even if you have to drop some CPU speed for raw memory speed. Even so, you now have the option of both worlds. Have fun overclocking

Click to expand...

Running 1:1 or 5:4 (or even 3:2) will depend on just how high I can get the fsb independent of the ram. It's at 266 right now, and I still haven't done the rebooting and playing around/testing I need to do to see what I can hit fsb-wise. (Massive upload right now I really shouldn't interrupt.)

If I can hit 290+ fsb, there's no way the ram will do that... I doubt it anyway - corsair ddr466. I'd still run a ratio (5:4 if I can swing it, or 3:2 if I must) and let dual-channel recover as much of the loss as it can.

Basically my gameplan is a) run 3:2 and increment fsb and voltages til I max out fsb

and

b) run 1:1 and play with vdimm and see how high I can push the ram before I start getting artifacts / instability.

Then compare benches / bandwidth / 'feel' from both and go with a gut choice.

Well... I'm pretty dissapointed. My ram has turned into a serious limiter. No matter what I tried (overclocked, default speeds, etc) I couldn't get it to boot with more than 1.60 vdimm. Any speeds over 230mhz or ANY speed with tight timings either produced errors in memtest86 or just wouldn't boot.

Attempting to strap as psb400 failed every time, absolutely could not. This meant no 3:4 divider, so I ended up going back to what I originally had, 266/5:4 (instead of something like 280/4:3 with slightly better timings, which I'm pretty sure it would have done.

While running a 3:2 ratio I also couldn't get anything more than 285 fsb, and that wasn't stable. =/ (I stopped at 1.65vcore, but I'm back down to 1.5 or .575 now, forget which.)

Maybe I'll get some better ram (this corsair ddr466, dissapointment =/) or more likely, just be happy with the working oc'd box until spring when I try to make the switch to conroe and pci-e gpu.

Always make sure your AGP/NB voltage is max out, this helps stability at high FSB, and it helps your RAM at high speeds flow through the northbridge on route to the CPU vice versa.

Ok, might try again later then.

edit -

northbridge voltage increase seems to have helped. I didn't rigorously play around like I did last night, but right now I'm at 280:224 / 5:4 stable, which is more than I got last night. Thanks for the tip!