Athlon 1600+ AROIA unlocking probs

I've been working with this Athlon 1600+ for about two weeks now, since the AROIAs hit, trying to overclock it to the speeds I've been seeing on these forums. I'm beginning to think I'm in the matrix and these insane speeds are just not attainable Here are my specs:

Athlon XP 1600+ (unlocked) AROIA
Swiftech MCX462 w/ Delta big *** loud fan
Soltek 75DRV5 (KT333) mobo
512MB Mushkin PC2700 2.5-3-3
~45-50C CPU temps under load, ~28C ambient, 1.85v CPU, 2.8v memory

I tried pushing the FSB beyond 151 and it just won't have it. Even between 148-151 it can be unstable in Warcraft 3 and I simply cannot get through the entire 3dMark2001SE benchmark (except one time). I just unlocked the processor today using a rear window defogger repair kit as per advice of an Anandtech member and while it unlocked the chip, it isn't quite acting right.

When I put it back in and booted up it came up with 1266 MHz at default (10.5) multiplier. I tried to see if I could underclock some more and checked 5.5x and it didn't even post. I tried 11x and no post. At 12.5x I got a post of 1800+ (???) What's odd is that it is basically shaving off 1 off each multiplier setting and seeing as I'm limited to 12.5 on this board I can't go any further :/

What's the deal with this? I'm happy that I can now push it to 2000+ without having to ramp to the unstable 156 MHz (I'm at 142 right now) but that is still far from the OC's I'm seeing from people on this board. Plus, my multipliers are messed up and I could potentially get even more if I could boost to 12.5x133 and then push the FSB even higher.

I still can't get all the way through 3dMark2k1SE at this speed, but I'm hoping Prime95 will run all the way when I go to sleep tonight. Is my memory crap? Is the processor whack? Does my mobo suck? Did I screw up on my unlocking work?

I need some help here. I want to hit at least 2100+ before I'm going to be happy.

By not being able to use all the miltipliers means you did not unlock the cpu properly. You'll have to redo the bridge connections.

I've had a really difficult time unlocking my AGOIA Y 0213 1600+. I've unlocked several other XP's without any problem, but no matter how carefully I try to unlock this one, I can't use any multiplier but default.

I'm using the same High Speed PC kit I've always used, and I know that I'm doing it correctly, even cerifying my result under a 30x microscope.

Is there a difference between the chips with a brown PCB and the chips with a greed PCB?

Is there a known issue with the newer 1600+ chips?

Yes some of the green PCB make unlocking difficult I've only unlocked 3 of 6 green XP's but I'm 100% with the brown ones I dont remember if its 5 for 5 or 6 for 6 or what.

Unlocking my brown XP1600+ AGOIA K only took my nephew approx 20mins. He didn't even need a magnifying glass for the most part. What a set of young eyes can do, I tell you.

What I'm trying to get at is this: are there changes in the green chips that require different methods? I've tried 5 times with no success, but I've never failed to unlock a brown chip on the first try.

There are some CPUs which come with the L1s coated with a cleaer material which can be sanded down to reveal the L1s and get a good unlock, this procedure could be dangerous, cause I heared of ppl sanding too much and getting to the copper underneath the CPU, and thus, killing it, if you try to sand it down slowly and carefully, it could help.

What MOBO do you have?

If you have one with a 1/5 divider, you need to avoid fsb speeds from 150 to 166 because you're PCI devices will be out of spec, causing some instability. Once you hit 166 fsb, if the 1/5 divider kicks in, you'll be right back in spec and you should be ok until 180ish fsb

PS...I hope the green PCB ones aren't that tough to unlock...that's what I have!

Well, I managed to successfully unlock my green AROIA. I cleaned it up and started over and its unlocked all the way through the spectrum. Problem is now, I can't use the 12.5x multiplier (2000+) because I have major stability issues. Even when I'm at 1900+ (12x) I can't get Prime95 to go beyond 2 hours of testing. This is ridiculous! I'm not OCing the FSB at all and I'm having these issues. I've tried two sticks of RAM, both with the same results (256MB of Crucial PC2100 and 512 of Mushkin PC2700, both set to 2.5-3-3). I'm beginning to think Soltek sucks :/

I'm going to try one more thing before I pull out this mobo and sell it on Ebay. I have a Soyo Dragon Plus (KT266A) sitting in my living room as a set top box/divx system and it has been one of the most stable boards I've ever used. I'm going to try my 1600+ in that system and see what kind of results I can pull. If I can't get a stable 1900+ that runs Prime95 for at least 12 hours, I'm just gonna sell my Soltek mobo and 1600+ and get me a P4 or something. :/ Been dying to try these new P4's as I've had nothing but poor OC results from the last 3 Athlons I've owned. Plus a P4 would rock my 80+ DVD collection I have to encode into Dvix A 1.2 Tbird just doesn't cut it.

But I probably can't afford a P4 + mobo so I'll end up just buying an Epox board and another 1600+ Damn Intel for being so gd expensive.

If anyone else has some more suggestions as to why my processor is so unstable let me know. BTW, after checking temps some more I'm actually lower than I thought. Around 43-47C under load @ 1.85v (max voltage)

So does your Soltek board have 1/5 pci dividers?
If it does, the divider will kick in at 166, that will put your pci bus back to stock speeds therefor not stressing them.
At 150, your putting too much to the pci, agp, and usb devices..

Yes my board has the 1/5 PCI divider. Thing is, I can't OC beyond 133 FSB when I bump the multiplier to 12x or 12.5x It isn't even stable at the stock FSB with a higher multiplier :/

That's really weird considering I should be able to at least get a stable OC at 12.5x133

And I've already tried going to 166 and that's a total joke. I slowly go up in small increments and while it will usually post and go into Windows at 156, after that I have to skip 10 mhz and go straight for 166, but even then it won't post. The problem isn't in the PCI bus going too fast. I'm not even going out of spec on the FSB right now. I'm at 133 for god's sake :/

I think you're going about this all wrong...

Put your multiplier at 8, 8.5, or 9, and try your fsb at 170. Tell us what that gets ya.

And there is a much greater benefit to high fsb then purely high clock speeds (thats the whole theory behind amd's success up to this point). Don't rely on your multiplier so much...maybe even try default multi and 170ish fsb....

And how are your temps??

Garaski - as per your advice I tried a few lower multipliers with higher FSBs...to no avail.


Here's what I tried and whether or not it posted. I didn't bother going into Windows and testing stability with any benchmark because for the most part, if it posts it works unless I'm ramping it to 2000+ or a high FSB.

8.5x133 - post
8.5x140 - post
8.5x145 - post
8.5x150 - post
8.5x155 - post
8.5x160 - no post
8.5x166 - no post
9x133 - post
9x140 - post
9x145 - post
9x150 - post
9x155 - post
9x160 - no post
9x166 - no post
7x133 - post
7x140 - post
7x145 - post
7x150 - post
7x155 - post
7x160 - no post
7x166 - no post

(Then I went REALLY low to see how it would do)

5.5x133 - post
5.5x140 - post
5.5x150 - post
5.5x166 - no post

I've noticed a pattern and it is that if I go to anything between 150-155 will usually post and go into Windows but will have stability issues - as long as its default multiplier or under. Anything above 155 usually won't post. Every MHz increment over 150 gets more and more unstable as well.

It's not pure stability in the high 140's either. It has some issues every now and then, but for the most part I can run it at default multiplier at 148 and it is good as gold.

I've also noticed that even at default FSB I cannot go beyond 1900+ without major stability issues and even 1900+ fails at 2 hours on Prime95, and cannot go past the 8 lights test on 3dmark2001SE. It will run PCMark2k2 just fine 90% of the time.

My temps are great especially compared to the old Tbird I used to have. Under load the max I've seen is 47C with the case closed. And when I have the side panel off I usually get about 45C max.

By the way, these are all at 1.85v, but even at that voltage my mobo keeps it at around 1.81-1.82v Memory voltages are at the default 2.5v

Oh...well, I just remembered something, and I really hope this is the case cause I think it'd solve your problems...What is your memory speed set to?? There should be like 3 options (under advanced chipset options) : SPD, HCLK, HCLK+33 (or it might be SPD, 266, 333).

You need to make sure that it's at HCLK/266 otherwise your memory is running way too fast once you hit the 150 mhz fsb area. If you have it at HCLK+33/333, then set it back and I bet your problems are solved! Gosh I hope that does it...

And for curiousitys sake I'd like to see you try 170 fsb too (or at least something past 166)

I hope its that memory speed thing that's holding you back....

PS...good work with the in depth testing and reporting. It makes it so much easier to help when people give out all those details!

Nah the only settings I got for memory are 133/166/SPD

I currently have the Mushkin PC2700 in running at 166 with 2.5-3-3 timings. I can switch the interleave and a few other minor options. And I can switch the CAS from 2.5 to 2 (although 2 crashes with either memory in) but I don't really have much else to tinker with memory-wise. I got all my BIOS settings optimized to the best possible according to the BIOS guide somewhere on this site...

This mobo doesn't have as many memory options as other boards I've seen in the past. For being a board "made for overclocking" it kinda sucks IMO.

I gotta switch the fookin mulitplier with dip switches, how's that for high tech? :/

As soon as Devils Own is done encoding (approx 2 hours) I'm going to pop my Tbird outta the Soyo board and pop this and the fan in and see what kinda results I get. If I get better results than I'm getting now I'll just throw the Tbird in this board and use the Soyo for my XP. Back to the 266A chipset though if it does work. :/

Whoa Whoa Whoa

Dont give up yet! You do have it set at the wrong memory speed! This is a third variation I hadn't seen yet... 133=266=HCLK which is what you want! 166=333=HCLK+33 which is too aggressive!!

Switch it to 133, and see what happens

*fingers crossed*

Alriight I'll try, but it won't work

If it does, you can kiss my ***

Gimme 10 min or so to test it.

How bout if it works, I'll just be happy for you...






HURRY

Ok you fag. You won

I tried 8.5x166 and lo and behold it worked. So its running 1600+ at 166 FSb. I did a PCMark and it raped my 12.5x133 2000+ mem and HD scores, but was 600 points below on the CPU score.

I figured 166 would give my memory more bandwidth therefore giving my CPU more head room to OC. Damn, now I gotta try 9 and 9.5 multis and 170 FSB. If that **** works, then I'm gonna go for 10 and default multi and if that works WOOT!

You da man. I'll keep you posted on the results. Gimme 10-15 min.

Just keep pumping the fsb man. When it craps out, then back it down 5 or 10, then hit up your multipliers.

What happens is, with the KT333 chipset, the RAM can run at a different speed then the fsb. When you're running at 133 fsb, and have that ram speed set to 166, then it runs the ram at 166 fsb (333mhz total cause its double data rate ram). It always adds 33mhz to the system clock. So when you moved the system clock up around 150, that meant the ram was clocked at a whooping 183 fsb, and thats when it started to falter (which is totally normal). So when you move that setting back to 133, it runs syncronously with the fsb (ie it no longer adds that 33mhz and it just runs at the system clock, times two since its double). You also typically run into PCI slot problems around the 150 to 155 area, which is what everyone else was onto. You should see some bangin scores outta sandra memory bandwidth now as well as 3dmark!

Congrats!