Trying to get a DUT3C 1700+ to 2.3ghz

[deadseasquirrel]

I have a KT333 board (8k5a2) with no 1/6 divider, and don't want to change the fsb past 166, in order to keep agp/pci in spec. Using Samsung original pc2700 synced with fsb at normal timings.

So, I wire-modded the sockets so that I can have a multi of 14x. To get it to boot into windows, I needed 1.85 vcore. However, that failed 3Dmark01 with a hard reboot. So, I upped vcore to 1.90 (the next available) and upped vdimm to 2.90--> quick question about vdimm... theoretically, in my head anyway, I can't imagine upping vdimm would help unless you are *overclocking* the ram, right? Well, it's just sitting at 166. But I did anyway.

This time, with the 1.9 vcore and 2.9 vdimm, 3dmark01 worked like a charm (even looped for 8 hours overnight w/o problems), as did 3dmark03.

The problem is that Prime95 fails on test#4 (couple of minutes).

Idle temp is 45C, load is 50C, using an sk7 and ys tech.

I really don't want to go any higher on the vcore (hell, I'm not too happy about 1.9v right now and the next jump is 1.95).

I'm thinking that maybe 2.3ghz is a bit too much for this DUT3C. Any other ideas-- any chance that a 3.0 vdimm on the ram would help?

Although I haven't given up, I did change the multi to 13.5x for 2.2ghz and it's been Prime95ing for a few hours now with no problems at 1.825 vcore and 2.9vdimm. Of course, temps are a little better-- 43C idle and 49C load. Since I think this is gonna be pretty stable, I'll try to go to 1.8v and lower the vdimm, and maybe increase the ram timings as well.

Assuming all these options are stable, which would you feel best with:
14 x 166 @ 1.9v
13.5 x 166@ 1.8v
12.5 x 166@ 1.65v

deadseasquirrel

 

[maluckey]

First, forget about the PCI bus speed, so long as you get no errors, and have good cooling. I never go below 43 mhz PCI speed.....ever. I run my KT266a as high as 2340 mhz (13 x 180 mhz), for gaming with Unreal Tournament 2003, for hour on end on a 1/4 PCI divider.

Second, this is a T-bred B? If so, run at 180 mhz FSB, and drop the voltages to 1.75 vcore (after burn-in and good thermal compound).

Third, your temps are way higher than mine. My setup is in a room that is nearly eighty degrees (farenheight) ambient. I run an Alpha PAL 8045, and the idle temps (13 x 174 mhz) are about 29 degrees C., and full load (heat soaked) is 43.5 degrees C. You may need better cooling, or to rethink your current setup to maximize what you have.

 

[deadseasquirrel]

Originally posted by: maluckey
First, forget about the PCI bus speed, so long as you get no errors, and have good cooling. I never go below 43 mhz PCI speed.....ever. I run my KT266a as high as 2340 mhz (13 x 180 mhz), for gaming with Unreal Tournament 2003, for hour on end on a 1/4 PCI divider.

I've heard bad things about NICs not doing well too high out of spec. The only other PCI is a firewire card I rarely use. I'm mainly concerned about the AGP.

Second, this is a T-bred B? If so, run at 180 mhz FSB, and drop the voltages to 1.75 vcore (after burn-in and good thermal compound).

Yeah, I forgot to mention it's a Tbred B. 180fsb x what multi? 13 like yours?

Third, your temps are way higher than mine. My setup is in a room that is nearly eighty degrees (farenheight) ambient. I run an Alpha PAL 8045, and the idle temps (13 x 174 mhz) are about 29 degrees C., and full load (heat soaked) is 43.5 degrees C. You may need better cooling, or to rethink your current setup to maximize what you have.

Ambient room is 81F, and the case sits under a desk with little ventilation. The temps might be a tad high, but I'm only another 5-6C higher than you at max. I doubt the instability is heat-related. The max load temp when I drop it to 13.5 x 166 is still around 49C (as opposed to the 50C), and it's rock stable. But your point is well-taken.

When I get back home, I'll set it up for 13x180 and see what she does. I'll try your 1.75v but I'm concerned about that since I couldn't even get it stable at 14x166 with 1.9v!

deadseasquirrel

 

[rommelrommel]

Originally posted by: maluckey
First, forget about the PCI bus speed, so long as you get no errors, and have good cooling. I never go below 43 mhz PCI speed.....ever. I run my KT266a as high as 2340 mhz (13 x 180 mhz), for gaming with Unreal Tournament 2003, for hour on end on a 1/4 PCI divider.

Second, this is a T-bred B? If so, run at 180 mhz FSB, and drop the voltages to 1.75 vcore (after burn-in and good thermal compound).

Third, your temps are way higher than mine. My setup is in a room that is nearly eighty degrees (farenheight) ambient. I run an Alpha PAL 8045, and the idle temps (13 x 174 mhz) are about 29 degrees C., and full load (heat soaked) is 43.5 degrees C. You may need better cooling, or to rethink your current setup to maximize what you have.

A) Do you format your hard drive bi-weekly? I'm amazed if you have no file corruption problems at 45mhz on the PCI bus.

B) I sincerely doubt that your real core temps are 29 C, even at at idle. You claim to have 27 C ambiants with 29 C core temps???

Deadsea,

More vdimm will not help, nor will ram timings. DUT3C's don't go to 2.3gHz all the time. Maybe try 14x160 or something to back it off a bit.

 

[maluckey]

Yeah! NICs are designed with only offices, and not enthusiasts in mind, and do not handle any overclocking gladly, so it may well be your weak link. If this is the case, the your best, and possibly only option, may well be by multiplier unlocking, combuned with mild FSB raise.

To respond to another statement directed at my post, and stay within the thread......

1. I have never reformatted my HDD since install, nearly eight months ago. I have only had mechanical disk failure on a drive from dropping it, about two years ago, so don't drop them anymore.

2. I also disagree about cooling. Granted, the two thermostats are not of the same source, so their overall accuracy must be scrutinized, but a wall thermometer, my buttocks, and the sweat on my arse say it's close to eighty in the room. I have only the Motherboard resourses to rely on, for in-case temps, and as anybody knows, they are for apple-apple, and safeguard comparison only. The fact is, that 50 C. no matter how you slice it requires more voltage to maintain stability, than does 45 degress Celcius. It's basic electronics 101. Drop the temps, and you decrease resistance, to a point.

So the whole disagreement unravels, and you stand at not wanting to raise the FSB? That's great! So do I, (when the need not to, arises). But the facts are that you get more by raisg the FSB than simply the multiplier. If you had checked it out firsthand, instead of listening to Urban Myth, you'd know that PCI maximum is less an absolute than the ability of the 1700+ to reach a certain speed on any given motherboard.

 

[Killrose]

Be happy with 2.2gig, or even 2.25gig which it will probably do. Just because temps appear ok does'nt mean the chip can still acheive it, which you have proven through testing it.

 

[deadseasquirrel]

Originally posted by: rommelrommel
More vdimm will not help, nor will ram timings. DUT3C's don't go to 2.3gHz all the time. Maybe try 14x160 or something to back it off a bit.

Yeah, I knew I was pushing it, but just wanted to get opinions on whether I was missing something else I could try. And 14x160 is outta the question, because at 160 my pci/agp would be way outta whack. It resets at 166. So maybe slight FSB bumpage from 13.5x166 would do. It's perfectly stable right now there.

Originally posted by: Killrose
Be happy with 2.2gig, or even 2.25gig which it will probably do. Just because temps appear ok does'nt mean the chip can still acheive it, which you have proven through testing it.

You are right, and I'm very happy with the performance of this TbredB. My first overclock was a 1600+ agoia palomino and I thought that one blew my mind. It's kind of funny, but with this KT333 mobo, Epox has stated that it doesnt officially support the 333 Bartons (though many have been doing so without issues). So, the end of the upgrade path for this mobo is with an XP2800+ TbredB... which this 1700+ has oc'ed to already.

deadseasquirrel

 

[rommelrommel]

Originally posted by: maluckey
Yeah! NICs are designed with only offices, and not enthusiasts in mind, and do not handle any overclocking gladly, so it may well be your weak link. If this is the case, the your best, and possibly only option, may well be by multiplier unlocking, combuned with mild FSB raise.

To respond to another statement directed at my post, and stay within the thread......

1. I have never reformatted my HDD since install, nearly eight months ago. I have only had mechanical disk failure on a drive from dropping it, about two years ago, so don't drop them anymore.

You're really lucky then. Running the PCI bus 36% over spec is a real good way to corrupt files.

[/quote]2. I also disagree about cooling. Granted, the two thermostats are not of the same source, so their overall accuracy must be scrutinized, but a wall thermometer, my buttocks, and the sweat on my arse say it's close to eighty in the room. I have only the Motherboard resourses to rely on, for in-case temps, and as anybody knows, they are for apple-apple, and safeguard comparison only. The fact is, that 50 C. no matter how you slice it requires more voltage to maintain stability, than does 45 degress Celcius. It's basic electronics 101. Drop the temps, and you decrease resistance, to a point. [/quote]

He's at 1.9Vcore. 50 C at load is pretty good. Yes, colder is better, but even a SLK-900 doesn't have a huge advantage over a SK-7. Maybe some case cooling will help, but not a lot.

So the whole disagreement unravels, and you stand at not wanting to raise the FSB? That's great! So do I, (when the need not to, arises). But the facts are that you get more by raisg the FSB than simply the multiplier. If you had checked it out firsthand, instead of listening to Urban Myth, you'd know that PCI maximum is less an absolute than the ability of the 1700+ to reach a certain speed on any given motherboard.

I don't even know what you're blabering about here. I think he should try tinkering with the FSB because it allows for more precise control of final speed. At a 14X multi and 166 FSB, one more multi = 83mHz. One more FSB = 14mHz. Get my point? I've owned three 1700+ b's and one 1800+ b so I don't need a lecture about what they can/can't do.

 

[rommelrommel]

Originally posted by: deadseasquirrel

Yeah, I knew I was pushing it, but just wanted to get opinions on whether I was missing something else I could try. And 14x160 is outta the question, because at 160 my pci/agp would be way outta whack. It resets at 166. So maybe slight FSB bumpage from 13.5x166 would do. It's perfectly stable right now there.

What's wrong with 160 / 5? Slightly underclocking the PCI bus would be fine.

Or even trying some moderate overclocking of it would be ok. I'd say any fsb up to 180 should be very safe.

 

[soja]

Originally posted by: rommelrommel

Originally posted by: deadseasquirrel

Yeah, I knew I was pushing it, but just wanted to get opinions on whether I was missing something else I could try. And 14x160 is outta the question, because at 160 my pci/agp would be way outta whack. It resets at 166. So maybe slight FSB bumpage from 13.5x166 would do. It's perfectly stable right now there.

What's wrong with 160 / 5? Slightly underclocking the PCI bus would be fine.

Or even trying some moderate overclocking of it would be ok. I'd say any fsb up to 180 should be very safe.

Yep. I've used 160/5 before (32mhz pci) and everything worked fine. My vid cards have always been able to handle more then my pci (nic, soundcard, HD, etc..). I'd say up to 175 is fairly safe. One nic I had was really sensitive to above 176 vs another that could handle 190/5.

 

[deadseasquirrel]

Originally posted by: soja

Originally posted by: rommelrommel

Originally posted by: deadseasquirrelAnd 14x160 is outta the question, because at 160 my pci/agp would be way outta whack. It resets at 166. So maybe slight FSB bumpage from 13.5x166 would do.

What's wrong with 160 / 5? Slightly underclocking the PCI bus would be fine.

Yep. I've used 160/5 before (32mhz pci) and everything worked fine. My vid cards have always been able to handle more then my pci (nic, soundcard, HD, etc..). I'd say up to 175 is fairly safe. One nic I had was really sensitive to above 176 vs another that could handle 190/5.

Maybe I'm doing something wrong, but when I adjust fsb to 160 in BIOS, it doesn't keep the 1/5 divider setting (underclocking pci). Rather, it overclocks it from 133fsb to 80/40. It seems the /5 doesn't kick in until 166 and I don't see a way to force it on the Epox 8k5a2.

deadseasquirrel

 

[maluckey]

On many boards, the 1/5 is automatic, and without the ability to reprogram your BIOS, there's not much to do about it. If Your NIC is the weak link, you may indeed be limited to a lower OC because of it.

I still believe that 1.9 V core is WAY too high for the mhz you are at. I'm at 1.725 V core stable. Raising the Vcore did almost nothing but produce heat. I took it to 1.80 V and got 26 mhz for my efforts. That wasn't worth it, so I dropped back down.

Try starting with NO PCI cards in the board, and only one stick of memory. Start at default CPU voltages, and raise the FSB/multiplier mhz at a time till the CPU requires voltage. When you raise it, only raise enough to stabilize. When you can raise the voltage no more, that is your CPU stopping point. Next, add the NIC, and then the next most sensitive pieces, and so on. Whan all is done, you've maximized your system.

One final tip. Do not use more than 1 stick of memory until the last addition. this reduces the memory as a possible source of errors.

 

[rogue1979]

13.5 x 179MHz for 2237MHz at 1.825v sounds very reasonable. Your AGP/PCI bus will be at 35.8MHz which will not be a problem. As a general rule of thumb, up to 37.5MHz should be safe. You might even give 12 x 186 a try, adjusting some memory timings and vdimm may get you that.

Samsung loves vdimm, so try 2.9v for that. You won't notice a speed difference between 2240 and 2300MHz, so don't sweat it.