New 2.4C, but poor results

I just got a new 2.4C. It's a SL6WF, made in the Philippines. It looks like it was made week 41 of 2003, but not packed out until this August.

I put it in an Abit AI7 board, along with two sticks of standard grade Crucial PC3200. Overclocking to 3000 MHz was a no brainer. I just set the divider to 5:4, and it worked perfectly.

I expected I should be able to get more out of it that that. I'm still hoping for something in the 3200 MHz range. I put the GAT settings to auto, auto, auto, disabled, disabled. I set the processor strap to 800 PSB. I tried leaving the memory timings at SPD, which gave me 2.5-3-3-7. That looked funny to me, since I thought my Crucial memory was only spec'd CAS 3. So, I tried manually setting the timings to 3-4-4-8. I've got the Vdimm set to 2.7 in the BIOS, but it shows up closer to 2.8 in Motherboard Monitor. Hyperthreading is enabled. Temperatures look decent: 22 deg C case, 24 CPU (idle), 42 CPU (load), 52 PWM (load).

I can't seem to get anything more than 3000 MHz out of it. I tried setting the clock to 255 (3.06 GHz), but I get errors in Prime95 fairly quickly. I also tried setting the clock even higher. At 267 (3.2 GHz) I get Prime95 errors almost instantly. At 275 (3.3 GHz) it will not boot into Windows.

I tried setting the divider to 3:2 to see if a slower memory clock would help. At a clock of 260 (3.12 GHz) the memory is at 173 - well under spec. It didn't help. I still get errors.

I also tried the divider at 1:1. I set the clock at 220 (2.64 GHz), and it booted and ran fine. This seems to tell me my RAM will stand at least 220 MHz, so I don't know why it's failing so soon when the divder is at 5:4.

I tried raising the CPU voltage slightly. It defaults to 1.525, but shows up in Motherboard Monitor closer to 1.54. I raised the BIOS setting one notch to 1.55, but it didn't seem to help. Besides, I thought Northwoods didn't really benefit from a lot of voltage.

Am I doing something wrong? I expected the 2.4C should do a little better than this. Did I just get a bum chip?

It might be your Powersupply.

At some point you will need to bump up the vcore, even with Northwoods. Just don't go too high. The 1.55v or 1.575v is fine if you have stock cooling. So far your temps look ok. If you have a good CPU heatsink and good case cooling, then 1.6v or 1.625v will be ok.

I think the AI7 defaults to fixed AGP/PCI at 66/33, but double check to make sure. Another thing you should do if you haven't already is to update the BIOS. Use the FlashMenu program that can be downloaded at the Abit site and flash to the most recent BIOS update.

musawi said:

It might be your Powersupply.

Click to expand...

I've got a Coolmax 400 watt supply in there. Reports seem to suggest it's a rebranded Super Flower / TTGI unit. It's rated 18A on the 12 volt rail. I've heard it's not a premium supply, but it's not junk either. Some say they (TTGIs) are made in the same factory as Enermax and Fortron.

I've watched the voltages in Motherboard Monitor. The 12 volt reading dips from about 12.6 volts at idle to about 12.4 volts under load. I haven't stuck a meter on it yet to see where it really sits.

The funny thing I've noticed is a voltage labelled DDR VVT. What's that? It sits around 1.6 volts at idle, but drops down to 0.98 under load. It isn't the DDR dimm voltage. That's correctly set around 2.7~2.8 volts.

12.6 Volts, Batboy isnt that too much??

batboy said:

...If you have a good CPU heatsink and good case cooling, then 1.6v or 1.625v will be ok....

Click to expand...

Case cooling should be good, as evidenced by the case temp. I've got plenty of 80mm inlet and outlet fans, and the air coming out feels cool.

The processor heatsink is the Alpha 8942. It's supposed to be a good unit. Although I'm not using Artic Silver compound, I followed the instructions on their site for applying compound to a P4. Look around step 10.
http://www.arcticsilver.com/arctic_silver_instructions.htm

PCI/AGP is fixed. I did check that. Motherboard BIOS isn't the most recent, but it's only one version off. The only thing mentioned in the revision history for the most recent BIOS is "updated microcode." I'll flash it in tonight anyway.

musawi said:

12.6 Volts, Batboy isnt that too much??

Click to expand...

I think my motherboard reads way high on the 12 volt. This is the second power supply I've had in there (the first was a 250 watt unit) and the readings are similar.

Check with those Multi Meters

The 12v rail is fine. It's better to have a little more voltage than less.

If you have the second to most recent BIOS, then that should be ok too.

That DDR VVT is not the vdimm. I think that voltage is supposed to be about half the vdimm. Look around, you should have a vdimm voltage, I don't remember exactly what it's called.

It's possible the mobo don't like the RAM. Abit are known to be a little picky with RAM at times. Can you get a friend to let you borrow some different RAM to try? Even a 2.4C made last winter should average at least 3.2 to 3.3 gig.

batboy said:

It's possible the mobo don't like the RAM. Abit are known to be a little picky with RAM at times. Can you get a friend to let you borrow some different RAM to try? Even a 2.4C made last winter should average at least 3.2 to 3.3 gig.

Click to expand...

I can try pulling out one stick of memory, so it doesn't run dual channel. Do you think that might help? It won't be easy for me to find someone with PC3200, and impossible to find something better.

That's what I was thinking. I ought to be hitting at least 3.2.

I have a different chip its a B but doesnt matter, I got 3 Ghz without changing the voltages. So its very possible you can even reach higher than that. Dont change voltages only if you have to.

Ty, since your vcore is fairly low right now and your load temps are reasonable.... I'd recommend setting the vcore to 1.58 or 1.6V and see if that helps your overclock at all. You have temps similiar to mine and I my temps only went up to 47°C load with 1.63V, so yours should be fine.

When you're just starting to oc a chip, I've found it's better to have a little extra vcore than not enough. When I first got my 2.4C, I had it at 3.2 gigs w/ 1.58V for a few weeks..... and now look at it, 3.42 gigs w/ 1.55V .

Good Luck oc'ing...... your SL6WF should have more in it than that, so just keep with it.

-Meatball

Well, I've done some more testing.

Setting the ratio to 1:1, and memory timings to 3-4-4-8, it seems stable at 245 FSB. That gives me a core speed of 2940 MHz. Short runs of Prime95 look stable.

While I'm very pleased that my PC3200 memory will run so nicely (albeit at very relaxed timings), I'm starting to suspect this chip is not a very good overclocker.

At this rate, I'll just run it 245 @ 1:1, instead of 250 @ 5:4.

Time to start cranking up the voltage.

Yes turn up that voltage!! 1.625 should take you up into the 280 range. Crank that mem voltage also!! Pour the juice to it!! That processor will be fine up to 1.725V I would not run it there continuously but you can for benching and testing!!

WZ

That processor will be fine up to 1.725V I would not run it there continuously but you can for benching and testing!!

Click to expand...

IMO, that's taking an unnecessary risk.... especially with a brand new processor that he isn't familiar with yet. I'm not a big fan of suicide runs... I prefer to find an overclock that I can run 24/7 with no worries. 1.73V is begging for SNDS and that's not a risk I'm prepared to take b/c I don't have so much money laying around that I can afford to throw a 100 bucks away for a little higher benching score.

WeldZilla said:

Yes turn up that voltage!! 1.625 should take you up into the 280 range. Crank that mem voltage also!! Pour the juice to it!! That processor will be fine up to 1.725V I would not run it there continuously but you can for benching and testing!!

WZ

Click to expand...

WeldZilla - Welcome to the forums!

But I've got to disagree w/your statement that 1.625Vcore "should" get a 2.4 P4C to 280 FSB. I notice your running a 2.4 'MO', and yes, a 'good' 'MO' may perform substantially better than many other P4C stepping's, so your performance may be somewhat exceptional, compared to the average run-of-the-mill 2.4 P4C. Ty isn't running an 'MO'.

Ty - Yeah, some chips just do 250MHz, or so, & that's about it. Though I think you should be able to do a little better. For some reason, I've never gotten *great* OC's from Philippine chips. Good, but not great. But this is from a very small sampling size, and luck-of-the-draw often reigns supreme! I DO think I'd try a little more Vcore. I usually run ~1.65V, which some peeps think is too high. But you can always try it, and drop it back down. I personally see no probs w/1.65, but do realize it's at the high end for 24/7. I think you're pretty lucky your PC3200 is running 1:1 245FSB. Too bad you didn't have some high quality PC4000 or 4400 you could try.

As for the divider not helping: Once in a while, even w/a steep divider, a specific memory set still won't OC the cpu in a specific mobo combo, when technically it should. I usually run Asus mobo's, and haven't seen it myself, but have heard of it. You might try the 3:2 divider again, but w/the loosest mem timings possible, a little more Vcore (1.65), and crank the Vdimm to 2.85V. Actually, 2.85Vdimm should help your 1:1 memory speeds, too. 2.85 Vdimm may sound high, but I've been running it for a year & a half w/o probs. Many memory manufacturers simply spec a 2.5V 2-3-2-6 PC3200 module @ 3-4-4-8, and 2.75 or 2.85 Vcore, & PRESTO: They now market the exact same sticks as PC4000! So crank up that Vdimm!

Good Luck

Strat

Stratcat said:

For some reason, I've never gotten *great* OC's from Philippine chips. Good, but not great. But this is from a very small sampling size, and luck-of-the-draw often reigns supreme!

Click to expand...

I'm almost ready to add another one to your samples.

I raised all the voltages 0.1 higher than their default. I set the core to 1.625, the DDR to 2.7, and the AGP/Northbridge to 1.65. When I look at them with Motherboard Monitor, they actually show up 0.02~0.10v higher than where I set them. I set the FSB to 267, and the memory timings to 3-4-4-8. Still no good. The system boots, but Prime95 dies within seconds.

If I'm going to have to resort to crazy volts just to get 3.2 GHz out of it, it isn't worth it for an extra 200 MHz. I'll just let it run at 3.0 with stock voltages.

Oh well.

Wow, I'm sorry that your oc didn't turn out the way you hoped.... it always sucks when that happens. I agree with what you say though..... 3 gigs is still plenty fast, most of the 3.0C's don't even get a 600mhz overclock, so don't feel too down. When I got stuck with my oc, I just switched over to Folding and doing case mods to lower my temps. Oc'ing the processor is just one part of overclocking...... there's many more things that you can do that are just as rewarding and fun.

Good Luck with your computer

mtb856 said:

Wow, I'm sorry that your oc didn't turn out the way you hoped....

Good Luck with your computer

Click to expand...

It's not so bad, and I haven't given up yet. Although I'm not a believer in burn-in, there's always a first time for everything. Maybe if I let her grind for a while at 3.0, she will soften up and let me get a little more.

I've still got a few other tricks to try. I still haven't swapped my DIMMs around into the various slots. I found a friend of mine who also has a 2.4C, and I'm going to try to talk him into swapping processors/ memory/ motherboards.

Ty Bower said:

If I'm going to have to resort to crazy volts just to get 3.2 GHz out of it, it isn't worth it for an extra 200 MHz. I'll just let it run at 3.0 with stock voltaget.

Click to expand...

That's the right 'tude, IMO. No sense burnin' up the Vcore for 200MHz. Didn't catch your cooling set-up, but there's nothing wrong a w/a low Vcore stable 25% OC on air. I think we just have higher expectations from the 2.4's. I set my machines up for Distributed Computing 24/7 on air, where they MUST be rock stable. I usually get around a 25% stable OC. But my stability requirements are probably a lot more rigid than some of the 'highest numbers possible' Oc'ers. I've got a machine that hadn't been re-booted for maybe 5 or 6 months, prior to a recent DC client upgrade. Otherwise, it just sat in the corner crunching Cimate Prediction 24/7/365. And when it went down, back then, it was due to a 110V AC Line Feed power glitch - not the machine's fault.

FWIW - My 2.6 Philippine chip will do 3.3 when messing around, and 3.2 when validated for running as a DC cruncher. That 3.2 is a little LESS than a 25% OC.

FWIW2 - If you're willing to get rid of your current chip, and are willing to pay a premium for a 'hand picked' 'MO', you might try calling Azzo Computer. They currently show out-of-stock on the 'MO's, but that's where I got my 2.8 MO, that shot straight up to 3.6. I had to wait a few days 'cuz they were out-of-stock then, too. But the peeps there were pretty good w/communication, and gave an accurate ETA. Many peeps believe the 'early' 2.4 MO's were the best, but there was a thread recently, about a month ago, where a guy just got a newer one, & posted it here. He got around 3.6, IIRC. But then again, luck-of-the-draw always plays its hand. Just thought I'd mention it.

Strat

<edit>

If you live in a place that gets cold over the winter, and wanna' challange, you can try for the 4G club while also helping HEAT your place too, by joining us in this S478 P4E 'multi-unlock' thread HERE!

Well, at least you'll get a Hi FSB AND X-tra winter heating in the same deal!