A64 3000+ Venice Overclocking Help? (DFI Ultra-D)

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
ashmedai said:
Ooooh yeah. Marketing lingo sticks in my head as little as I can manage. I was trying to figure out WTF processor went in a Socket ZIF.

Well you're in the neighborhood, now just for the stability tweaking. Run over RGone's guide again if you haven't, also LDT 1.5V helped me a bit with the higher HT clocks. Chipset...dunno, mine's usually around 40C when I look. Don't think what memory you have matters that much as long as it's half decent and you set the ratio appropriately.
Temps are no higher than 42C now no matter what voltage or setting I'm using - at least for those that I've been able to boot to windows in.

I have left LDT at 1.4 as I had read somewhere it was an optimal setting with issues at 1.5 - I will give it a whirl tomorrow at 1.5 and see if I have any luck.

After posting this morning I cranked my HTT up to 300 and then was playing with voltages, ram divider ratio's, ram timings, etc to see how far I could get. I found a couple of combinations that allowed me to try to boot to windows (it always posts) and the rest would no even get as far as "Backing up CMOS". The ram seems really finicky as sometimes a value 1 higher and/or 1 lower in ram timings would allow it to get further and error out on loading windows (unable to find/load pci.sys, ntoskernel, etc).

I was hoping to re-stumble across the combination that allowed me to get into windows at 300 HTT - and then fine tune for stability - but no luck so far.

I even got brave and (very shortly - a few minute at a time at most) tried DRAM voltages as high as 3.4-3.5 - but saw absolutely no difference from having it at 2.90. At least with the settings I was using at a time.

So today I'll do some more reading, and tomorrow morning I'll try again :)

edit - for that matter, I think I'll take some pictures of the bios screens with my 300 HTT settings and have people critique them - perhaps there's something terribly obvious that I'm missing :)

edit 2 - and yes, HTT multi at 3 and ram divider at either 133 or 100. Have tried both.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Corsair Value isn't that great to start with and I don't think the Ultra-D likes it very much either. And most RAM doesn't need more than 2.8 or 2.9V, CH-5 is a bit of an exception that way. Even that is usually fine with 3.2V, less if the clock isn't too high.

BTW, you have all four motherboard power connections plugged in, right?

BIOS screens might be helpful, it's a pity no one has made a utility to dump one's BIOS settings to a file...at least not that I've seen yet. Anyone that knows different, PM me.

LDT 1.4 is recommended for most overclocking, 1.5 just seemed to help. As with anything if you can lower the voltage and have it still be stable after you hit your final speed, it's a good idea.

Try running just one stick in channel 2 until you get it working, it makes it a bit easier especially when the RAM can just almost make it as it is. But the fact that the RAM divider is having a direct effect on where it crashes makes me want to blame the Corsair.

If you still have trouble, un-overclock it for a bit and flash with the 5/10 beta BIOS, it might help since there are memory issues going on.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
ashmedai said:
Corsair Value isn't that great to start with and I don't think the Ultra-D likes it very much either. And most RAM doesn't need more than 2.8 or 2.9V, CH-5 is a bit of an exception that way. Even that is usually fine with 3.2V, less if the clock isn't too high.

BTW, you have all four motherboard power connections plugged in, right?

BIOS screens might be helpful, it's a pity no one has made a utility to dump one's BIOS settings to a file...at least not that I've seen yet. Anyone that knows different, PM me.

LDT 1.4 is recommended for most overclocking, 1.5 just seemed to help. As with anything if you can lower the voltage and have it still be stable after you hit your final speed, it's a good idea.

Try running just one stick in channel 2 until you get it working, it makes it a bit easier especially when the RAM can just almost make it as it is. But the fact that the RAM divider is having a direct effect on where it crashes makes me want to blame the Corsair.

If you still have trouble, un-overclock it for a bit and flash with the 5/10 beta BIOS, it might help since there are memory issues going on.
Running 510-3 right now actually.

I finally gave up, noticed NCIX had a new sale, so I grabbed this ram for a net cost of about $90 more than I paid for the Corsair Valueram. Plus, I've still got my 2x sticks of 512mb PC2700 with a pending buyer to make up the remaining cost difference - so all told, I should break out even. I ordered the ram, and will RMA the corsair stuff when it shows up tomorrow or friday (which means I still have two days to toy with the corsair stuff).

I will snap a couple of bios shots tomorrow morning of the corsair settings though - it's always a learning experience.

And yes, all 4 power plugs are connected. I had also thought of dropping one stick and going 1x512 until I could get it clocked right. I'll add that to my "to-do while drinking coffee" list for tomorrow before work. (that's when I do most of my computer work / time).

Anyone have any thoughts/critiques on the new ram? Pro's? Con's? Experience? Advice?
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Functionally it's the same stuff I'm using. Run it up to 3.2V with 2-2-2 timings and go from there. On that BIOS (and the 3/16 beta it's derived from) CH-5 is supposed to work better in 1&3 instead of the usual 2&4 - but ONLY that BIOS, if you go back to 3/10 or a different one for some reason you're going to need to put it back in 2&4.

Strike that, you need the 5/10-2 for that. 5/10-3 is still based off the 3/10's tables not the 3/16 see here. Might wanna re-flash it. Not sure what the 5/10-1 is for yet, still reading. Think it's for TCCD users though.

Peter, you're probably thinking TMSP/400. Egg has it for about $52 plus shipping for a 512MB stick.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Yeah, -1 is still the designation for the TCCD special. In this case though it looks like 510-1 has had some issues so I'd stick with -3 for TCCD & everything else, -2 for high voltage types like Winbond (like the CH-5 in the VX).
 
Joined
Jan 15, 2004
Messages
705
1 = optimized for tccd
2 = for bh-5/ch-5/utt
3 = work with anything.

no, that specific ocz are suppose to carry the UTT chips, the same chips in the gold vx except these were not "tested" and rated at 2-2-2-5.

i believe the twinmos sp sticks are suppose to have either the UTT or ch-5 chips which are just as good.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
UTT != CH-5, UTT = Untested. Could be Untested CH-5, could be Untested horsecrap. It's frequently used intending CH-5 though.

TwinMOS SP/400 most definately has CH-5. It's right on its Mfg. spec page. SP/433 uses BH-5 instead though. Gold VX also uses CH-5.

That OCZ is apparently it after all, the serial is right for Value Winbond series, Newegg's just not doing a very good job of labeling it. Also they're tested by OCZ at PC3200 levels, apparently they can do it by machine that way with lower overhead cost. On the other hand it's at the same price as TMSP/400, and suspect of being VX rejects, so I'm still thinking TMSP/400 is a safer bet with no financial reason to choose this. Good to know though, gives a backup next time TMSP/400 is out of stock at the Egg.


peterhoang2002 said:
1 = optimized for tccd
2 = for bh-5/ch-5/utt
3 = work with anything.
To reiterate, the -1 version has some issues so it might be a good idea for TCCD users to default to the -3 unless they're prepared to do extensive research about it on their own first.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
peterhoang2002 said:
i heard that this ram uses the same chips that the ram u link to have except its much cheaper and doesnt have the heatspreaders.
Perhaps - but there's no guarantee. I've had enough bad luck with value series ram (roll the dice on if you'll get IC's that'll do 2-2-2-8 or 3-3-3-8).

The corsair value-select that I currently run was recommended to me by a Corsair employee no less. He said that 99% of the VSR sticks in the past 16 weeks have been UTT or CH-5 of varying makes - and the spd should report 2.5-3-3-7 to start, supporting up to 2-2-2-8 or something like that. We then discovered that I just have that bad of luck, and got some generic crud they were using as a place-holder on the lines that day - and it seems to barely like 2.5-3-3-8.

Plus - even if it is the same chip in a value ram, as Ashmedai said - it's obviously in there because it's not quite hacking it for the full spec stuff. Either way, for what's going to net me at most $20 more out of my pocket, I'll stick with something that both guarantees 2-2-2-8 @ 3.2V as well as advertises warranty as high as 3.5V.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Ahhh! So very very close now.

I figured I'd bump the thread with an update and a little snippet of interesting trivia.

Got the new ram (PC3200 2-2-2-8 @ 3.2V OCZ EV VX Gold) and installed it. Runs 2-2-2-8 just like it advertises, without a single frickin problem.

Trivia bit is that if you search google for "DFI Ultra-D Beta Bios" this thread is the second ranked link now. Neat.

280 HTT - works like a charm for the 10 or so minutes I ram prime on it to test. -Proof Here-. I was quick shocked that I managed to my ram speeds that right right out of the game too.

300 HTT is a bit more work though. Can get into windows now, and it appears to run relatively stable, but I can't get it to roll on Prime for more than 10 seconds. I'm playing with ram, voltage, etc - changing then reverting to try to tighten this up a bit.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Try putting VID Special up a notch if you haven't yet. 1.55 isn't quite stable on what I'm running, but 1.56 is rock solid. Same for the 1.5V LDT.

BTW there's a special beta of Prime95, the normal version is screwy and gives errors on A64 when it should work fine. Version is 24.6, lost the link though.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
ashmedai said:
Try putting VID Special up a notch if you haven't yet. 1.55 isn't quite stable on what I'm running, but 1.56 is rock solid. Same for the 1.5V LDT.

BTW there's a special beta of Prime95, the normal version is screwy and gives errors on A64 when it should work fine. Version is 24.6, lost the link though.
I've been using VID's ranging from 1.30-1.45 with VID Special increasing it as high as 1.69 - just to see which ones run a bit better than the others.

The Prime errors I keep getting are rounding errors, but I'll quickly check the revision.

I'll try the 1.5LDT again too. I still think it's less stable at 1.5 than 1.4 - but I may be imagining things.

edit - oh for crying out loud. I backed up the settings that DID let me into windows, played for a bit with no luck and reverted to them. What do you think happens? They error now. Yeesh.
 
Joined
Jan 15, 2004
Messages
705
hey. so ram did make a difference for you when overclocking? i dont remember but where u able to get your HTT up to 300 and boot into windows before with a divider?

i ordered these sticks just yesterday:
http://www.newegg.com/Product/Product.asp?Item=N82E16820227213
i called up ocz to get an idea of what im getting. basically the rep i talk to said that these are "untested" and not "rejects". so basically, testing cost money and thefore the gold vx sticks cost more while they use the same sticks. the "rejects" are sold at a lower price, i think its the "premier" ram, but i wasnt sure.

ill post my OC results when i get them next week. wish me luck! :)
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
I think it's generally better to use the normal VID setting when possible and the Special strictly as a supplement. Don't forget to set VID Startup either.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
peterhoang2002 said:
i called up ocz to get an idea of what im getting. basically the rep i talk to said that these are "untested" and not "rejects". so basically, testing cost money and thefore the gold vx sticks cost more while they use the same sticks. the "rejects" are sold at a lower price, i think its the "premier" ram, but i wasnt sure.
Wouldn't load for me (probably my damn firewall acquired Newegg's IP in its block list again, it does that every so often for some reason) but I do know OCZ has a "value" line that's substantially similar to Gold VX but is instead machine tested at PC3200. I also have my suspicions that if it doesn't pass muster at the Gold VX testing bench it gets sent over to see if it passes for the value stuff...so technically it's not rejected since it passed the PC3200 machine test...it just couldn't pass the Gold VX binning.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
peterhoang2002 said:
hey. so ram did make a difference for you when overclocking? i dont remember but where u able to get your HTT up to 300 and boot into windows before with a divider?

i ordered these sticks just yesterday:
http://www.newegg.com/Product/Product.asp?Item=N82E16820227213
i called up ocz to get an idea of what im getting. basically the rep i talk to said that these are "untested" and not "rejects". so basically, testing cost money and thefore the gold vx sticks cost more while they use the same sticks. the "rejects" are sold at a lower price, i think its the "premier" ram, but i wasnt sure.

ill post my OC results when i get them next week. wish me luck! :)
No, I wasn't able to even get in at 1:2 300HTT previously.

Good to know about the OCZ - I nearly bought the premier stuff then the stuff I got went on sale.

I guess the one major thing I'm doing wrong here is not finding the limit / speeds / voltages on this new ram. It's early, I have work in a half hour, and just wanted to give it a quick test. Considering my luck, I may just postpone another day or two and redo my max HTT, CPU and RAM tests again.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Oh, and before anyone asks - running 510-2 for the BH/CH improved support and have tried both 2/4 and 1/3.

Should VID and VID Startup always be identical, or is there reasoning between having different and/or staggered values?
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Mine are both at 1.55 with a 1.04% Special.

And I assume there's a purpose to having control over the two seperately, but I'm not sure what it is, and I don't know at what point the voltage control switches from A to B either. I think it might have something to do with operating it with unconventional cooling systems, but that's pretty much a shot in the dark. About to go out or I'd look it up now that you reminded me. Hopefully someone will have enlightened us by the time I get back home. ^_^
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
okay, I shouldn't have been lazy and have avoided memtest up until now. 300 HTT, 5:6 ratio (166) resulting in 245mhz @ CAS 2 2-2-8 Dual Channel @ 3.2v. Running through the first pass now and not a single error.

So my assumption (confirmation would be great) is that those speeds and voltage are fine for 300 HTT and I should completely ignore the ram now. It's likely a voltage tuning thing for the VID now to hit the right value to get it stable - correct? Or am I just way off base, and although the ram's testing fine - once it's put into practice with the machine running full bore, even ram that tests fine can cause windows to BSOD on startup?

edit - whoops, spoke too soon. 40% into first pass and I'm getting errors out the wazoo. Stepped divider down to 3:4 (150) and it's going better.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
I just left the RAM at 133 until I was done with everything else and dealt with it last. Think I might be running 150 there too. Are they in 1&3 right now?

Test 5 tends to be pretty significant...if it can pass through the end of that one it's probably stable. However this doesn't mean to just run it and skip the rest, just that it'll usually get a slew of errors there if something's wrong.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
ashmedai said:
I just left the RAM at 133 until I was done with everything else and dealt with it last. Think I might be running 150 there too. Are they in 1&3 right now?

Test 5 tends to be pretty significant...if it can pass through the end of that one it's probably stable. However this doesn't mean to just run it and skip the rest, just that it'll usually get a slew of errors there if something's wrong.
I set everything back to stock (and left ram at 2-2-2-8 @ 3.2v) before I left for work this morning.

When I had it at 150, I made it up to about 73% the default test on the two sticks (in slots 1 and 3) before I started getting errors.

It'll probably have to wait until monday now, but what I'm going to do is run each stick through seperately at both 2.8v 2-3-3-8 and 3.2v 2-2-2-8 - then if both pass (ie; I'm hoping one of those sticks isn't bad) then I'll put them dual channel 1&3 at both 2.8v 2-3-3-8 and 3.2v 2-2-2-8.

I am wondering if perhaps there's a ramp-up time and/or burn-in period of sorts needed for these sticks (are they BH or CH???).

I'm confused/concerned because I WAS getting in to windows to a stable desktop at 3.2v 2-2-2-8 timing at 300HTT. I saved those settings, played around some more with no luck, then reverted - only to find it no longer wanted to work (IRQ not equal with memory dump generic BSOD).

I hate having to break up this testing into peices like this, doing a bit each morning and whatnot :)
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
CH-5. They should work at 2-2-2 at relatively lower voltages, 3.2V is if you want it to do so at higher clock rates. How lower...don't know yet.

Definately always test new memory modules seperately so you can determine if there are any defects. When I first got my TwinMOS one of the sticks had a bad chip or something, so no matter what I did this one specific range would always get errors. Had to RMA it.

Also (and this is a shot in the dark, it's more likely one of the settings you changed is causing it) at one point my BIOS image became corrupt, and needed to be re-flashed before it'd stop acting weird.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
ashmedai said:
CH-5. They should work at 2-2-2 at relatively lower voltages, 3.2V is if you want it to do so at higher clock rates. How lower...don't know yet.

Definately always test new memory modules seperately so you can determine if there are any defects. When I first got my TwinMOS one of the sticks had a bad chip or something, so no matter what I did this one specific range would always get errors. Had to RMA it.

Also (and this is a shot in the dark, it's more likely one of the settings you changed is causing it) at one point my BIOS image became corrupt, and needed to be re-flashed before it'd stop acting weird.
It looks like "burn-in" isn't as much of a myth as i had hoped/thought. At least as far as the ram is concerned. I jacked it up to 2-2-2-8 3.2v as I mentioned before, and it was erroring all over the place. So I dropped it down to 3-3-8 2.6v and ran it through memtest for a couple of hours. I would come back every couple of hours to reboot and ramp the voltage up a notch, then when I hit 3.2 @ 3-3-8 with no errors - I tightened the timings to 2-2-2-8 @ 3.2v. So far, so good - no errors.

Still can't boot into windows, which leads me to beleive this Venice I have wants more juice than I'm comfortable giving it. I've gone as high as 1.69-171 and gotten BSOD's (instead of it just not even trying) but that was also when the ram was misbehaving.

I'm justing going to finish up these memtest settings @ 300 HTT / 2-2-2-8 / 3.2v (225mhz) and then see what I can do to tighten things up to get into windows.

What's wierd is if I leave memtest running on the standard default test, it will get through an unknown number of passes and then freeze up at 80% partway through a pass. I've never thought to look at what pass # it was on, just hit reset and carried about business as usual - but it's likely a couple dozen passes in that it does it.

edit - okay, what gives. 7% into test # 8 in memtest86 and I get a splash of ascii characters and it locks up. I thought memtest was supposed to show errors on bad ram/clocks not freeze up. Might this mean I have a bad copy of memtest (bad bios/flash/memtest86 integration) or is it my ram?

edit 2 - what in the hell. Test #10 in memtest86 is currently Pass % 200, Test % 205 and climbing. I'm going to clock back down to stock and see if both of those tests error out there as well.

edit 3 - ok, it was just the ram settings. guess I haven't worked this ram in quite as much as I had hoped. Reset to stock and it will get past 8% in test 8 at least.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Interesting. I've been playing with settings because although I was able to get 2-3-3-8 timings at 300 HTT to start to load windows, it would freeze up on the splash screen.

Dropped everything I could think of one notch at a time.

Only to discover that it does the same damned thing on stock values anyways.

I wonder if I borked either my windows install, or something a little more serious.

edit - windows works in safemode.
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
Wow, you're having an unhealthy amount of "fun" with this thing aren't you? :rolleyes:

Seriously though, I hope things do start to come together for you...check my sig for what I've done so far with my combo.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
cornelious0_0 said:
Wow, you're having an unhealthy amount of "fun" with this thing aren't you? :rolleyes:

Seriously though, I hope things do start to come together for you...check my sig for what I've done so far with my combo.
.. 335x8 eh? Hmmmm.

Bah, I'm a stubborn sort so I'm going to keep pushing for 300x9. I've pretty much got it narrow down, just need to get the problem solved :)
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
If this wasn't so much fun, I'd be having a nervous breakdown now.

I literally panicked about 20 mins ago as I couldn't get my machine to boot into windows no matter the setting. Then I tried safemode and discovered it worked, and realized I must really have (As per my previous speculation) borked my Windows install.

That may also have been creating this artificial ceiling / issue after I've established that I can do 2-4-2-8 at 3.2v rock stable for 24hrs - I just couldn't get the machine to boot into windows at 260/275/280/285/290/300 HTT.

So here goes a repair install, then I'll try AGAIN. :)

I refuse to give up 300 HTT without a fight.'

edit - damn, hoped that chkdsk from repair console would save me a repair install, but no such luck.

edit 2- what the frip... no matter if it's booting into an existing install, or try to repair install windows, when I get to the windows splash screen it locks up after a little while (about 6-8 passes of the bar).
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Wow what a mess.

Can't even re-install windows (on the same HD, or a different one) without getting a BSOD after the "hit f8 to accept eula" portion of the install phase. Everything is at stock values, default/auto - and I've tried the ram at both 2.6 and 2.8 just incase it needed more juice.

Temps are all perfectly fine (cpu 32c, pwm 36, chipset 40) and power all seems normal enough to me.

I'm going to download and burn that DFI bootable bios cd and reflash - but this time I'm going to take a step back to the 4.14's and see if that helps at all. Maybe all of this is because 5.10 doesnt like my ram or something... mind you I did remove the new stuff and temporarily replaced it with my old pc2700 - and had the same problem.

Any suggestions if reflashing does help?
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Well so far so good with the reflash of the bios. Must have been corrupted or something.

I'm just doing a repair install just incase something else got buggered then it's back to where I started :)

edit - works fine @ stock values now. Running 414-2 bios. Don't know if it was the 510 bios that didn't like me, or if the flash started off or became corrupted after my tweaking.
 

Herulach

Gawd
Joined
Dec 24, 2004
Messages
997
I dnt know about DFI boards, but i know on mine i have only 2 sata ports that are frequency locked, the otehr two i wouldnt trust at all while overclocked, god knows what kind of frequency theyd be running at. That might be whats causing your corruption, although i imagine dfi, being kings of overclocking, have the sense to lock their sata ports.
 
Joined
Jan 15, 2004
Messages
705
hang in there Lukano! dont give up! you cant let the cpu win.

well for my cpu, its only doing 280 * 9 stable @ 1.664 volts. primed over 24 hours. dropping volts down one increment (1.64 i think) made it unstable so im fairly convince that if i give it more volts, my chip can go to 290 or maybe even 300. only barrier limiting me is my cpu @ 280 * 9, 1.664 volts gives me 38c/61c idle/load temps on my current zalman cooler.

today i went ahead and ordered an xp-90 and so by the upcoming weekend, i should have a new cooler and some new ram to test out and then hopefully itll reach at least 290 after i give it some more volts.

well, how are your temps lukano at the highest stable overclock you can get?
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Herulach said:
I dnt know about DFI boards, but i know on mine i have only 2 sata ports that are frequency locked, the otehr two i wouldnt trust at all while overclocked, god knows what kind of frequency theyd be running at. That might be whats causing your corruption, although i imagine dfi, being kings of overclocking, have the sense to lock their sata ports.
I'm not using any SATA driver though, all PATA. So unless one of the channels is SATA->PATA converted and I'm not aware of it - or PATA drives are inversely affected as well, I can't see that being a big problem. Also, bear in mind I was also having the same problem with the HTT multi set lower, and it was fixed when I reflashed to a new bios.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Update - I can get desktop stable 300 HTT but not prime stable. I'm not sure if it's vcore or ram that's holding me back (ram doesn't seem to like more than a 2:3 divider, and vcore apparently MUST have 1.50 + 123% or better (resulting in 1.77 or higher).

The temps even at those outrageous vcore levels are around 50C at the highest, which I suppose isn't too terribly bad.

Maybe more dram voltage and less cpu will even thing out a bit, and maybe stabilize these Pi errors.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Okay, so here's my most stable so far. I haven't primed or pi'd it for 24hrs to absolutely prove it - but it's leaps and bounds better than I've been getting so I'm taking any indication of prime running for more than 5 seconds without error to be "stabler" and therefor a launching point for tightening timing and stabilizing if need be.

http://www.neuro-tek.com/300x924410.PNG

So any recommendations on what to tweak from here? A few things I'm unhappy with that I'd like to tighten are;

- vcore is 1.79. That's resulting in 50C loads temps which I'm not 100% happy with, but also realize that it's not really life threatening either.
- dram voltage is 3.2v which is retardely excessive for those loose timings of 2-4-4-10
- paired with the ram timings is that bloody divider. I know this ram can do at least 240 at 2-2-2-8 at 1:1 stock speeds - I had it running like that to warm it up when I first put it in. But now it doesn't like more than the current 192mhz at 2:3 divider which I'm really displeased with.

So suggestions are welcome. I'll tune a bit more tonight, but then I'll be leaving it for 24hrs to prime prove it at the current specs.

edit - OH!!! And I forgot to mention that I was running v23.xx of Prime95 which was giving me inaccurate rounding errors, which was solved by updating to 24.11 .
 
Joined
Jan 15, 2004
Messages
705
holy sh!t! you're running at almost 1.8v and your load temps are only 50c? what kind of cooling are you on?

edit: nm, i just saw artic cooling freezer in your sig.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
peterhoang2002 said:
holy sh!t! you're running at almost 1.8v and your load temps are only 50c? what kind of cooling are you on?
Arctic Cooling Freezer64, Coolermaster Aero casefan venting right behind it and Fortron power supply venting above (freezer positioned to vent out of the Aero).

I was actually hoping for LOWER temps believe it or not.

I'm absolutely stumped here.

I've got my ram timed at 2-3-3-8 3.4v (yes, high I know, but remember the VX is warranty safe up to 3.5v) but it's only 192mhz due to the divider.

Now if my ram is supposed to do 2-3-3-8 @ 2.6v - is that only 1:1 and do the requirements to hit that become higher/harder when your using a divider to inversely affect it from the HTT? Or should it be the same?
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Ooh, nice. Got it desktop stable at 1.66v now (1.55 + 110%).

Still not Prime stable though damnit all anyways. It's these ram timings that are killing me.
 

Lukano

Limp Gawd
Joined
Mar 11, 2003
Messages
344
Last self-serving bump/update of the night for me.

Was able to get desktop stable at 2-4-4-10 and 2.5-3-3-10 but neither were prime stable. RAM heavy prime tests were erroring, where-as the CPU heavy prime's would run significantly longer before giving a hardware/rounding error (usually ram related I'm going to assume).

Still can't convince it to boot to desktop with anything but 3:2 133 - which might have to do with some timing settings, and may just be the ram.

So I'm going to set everything back to stock/auto then adjust dram voltage up to 3.40 (with NO +0.3v enabled, that was giving me issues before) and then clock the ram to 1:1 2-2-2-8 and let it burn-in in memtest overnight. Maybe then I'll be able to force it to at least try 150 or 166 2-3-3-8 or 2-2-2-8 - or worst case scenario 150 or 166 2-4-4-10 like my current desktop stable is.

If this doesn't help though, I'll need a bunch of suggestions to try to acheive prime stable without dropping my HTT at all. I'll be back with updates in the morning.
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
I'm still with ya on this one bro...I'm lucky all I had to do was mess around with a few settings to optimize some of my clocks.....I feel for ya.
 
Top