XP 2500 Overclock Problem Questions

I have an unlocked XP 2500+ Barton which i bought 2 Corsair CMX256A-3200C2 modules to run in dual channel with. I ran them at PC2700 for a long time because i was unable to get my computer to run properly with the ram/FSB at 200mhz, EVEN at 5:200:200. Im trying again and asking for help.

It dosnt seem to be unstable, but i still fail prime95 tests. I can run the small fft test for 3 hours or maybe more, but i cannot run the blend or the large fft test for more than 15-30 mins without an error usually. But the system seems stable otherwise.

Recently i ran a blend test using 11*190:190 (MP*FSB:RAM) with a vcore of 1.675 for over 22 hours. I would call that stable, i did other things during that time too, i did not notice any instability. Thats roughly 2090mhz, CPUZ reports it as 2101.6 or so.

However, i also ran 9*200:200 at the same vcore and i failed the large and blend tests.

As best i can tell, it *seems* to fail the prime95 tests which also test ram, but it does not seem to fail the ones which do not (small fft.) I have run memtest86 through its cycle a time or two and never come up with any memory errors.

There are a few things that i think could be causing the problem, but its hard to rule anything out exactly, so im going to ask you guys first, see what you say. It could be related to my power supply, which is an Antec PP352X 350 Watt, supposed to be a good power supply, but under a load, the voltages fluctuate quite a bit from the 'default' settings. At this time i do not have a log of voltages not under load.

If you guys have *any* suggestions i might try, i would be thankful. I cant afford any new hardware for the time being so im just trying to make due with what i have. Built this thing to run @ 200mhz FSB, dont see why it dosnt run so without erroring myself. For some reason i just cannot run my ram without prime95 errors at pc3200 even at rated & loose timings in prime95. Even when i do not run 1:1. I have long tables of tests i have run and moderatly well documented, if more info is needed i can provide it without any trouble.


Ratha

-

Below is a voltage readout from motherboard monitor, a few current/nonrecent tests, and my system specs.


Default Vcore: 1.675
Default Vdimm: 2.6

+-------------------------------------------------------------------------+
|Total number of readouts: 215 CPU Speed: 2060 MHz |
|Running from: 12/18/04 3:53:30 PM until: 12/18/04 4:28:35 PM |
+-----------------------------+----------+----------+----------+----------+
|Sensor | Current | Low | High | Average |
+-----------------------------+----------+----------+----------+----------+
|Case | 27° C | 27° C | 27° C | 27° C |
|CPU | 42° C | 42° C | 43° C | 42° C |
|Sensor 3 | 0° C | 0° C | 0° C | 0° C |
|Core 0 | 1.62 V | 1.62 V | 1.66 V | 1.63 V |
|Core 1 | 2.69 V | 2.67 V | 2.70 V | 2.69 V |
|+3.3 | 3.26 V | 3.22 V | 3.28 V | 3.25 V |
|+5.00 | 4.95 V | 4.92 V | 4.95 V | 4.95 V |
|+12.00 | 11.67 V | 11.61 V | 11.80 V | 11.71 V |
|-12.00 | -12.19 V | -12.19 V | -12.11 V | -12.18 V |
|-5.00 | -5.19 V | -5.24 V | -5.19 V | -5.20 V |
|Fan 1 | 4017 RPM | 3924 RPM | 4066 RPM | 4000 RPM |
|Fan 2 | 0 RPM | 0 RPM | 0 RPM | 0 RPM |
|Fan 3 | 2163 RPM | 2149 RPM | 2177 RPM | 2165 RPM |
+-----------------------------+----------+----------+----------+----------+


-------------CL/RCD/RP/tRAS----------VCORE 1.675v----------Corsair DC Mode-------
MP FSB RAM - Timings - VDIMM (real) - Test HRS:MINS (# Pass) - [Tries]
---------------------------------------------------------------------------------
11 190 - 2.5-3-3-6 - 2.6v (2.7) - Stop Blend - 22:14 - (74) - [1] 42C

11 200 - 2.5-3-3-8 - 2.6v (2.7) - Fail Blend - <0:21 - ( 0) - [2] 43C

10.5:195:195 - 2.5-3-3-6 - 2.6v (2.7) - Blend - In Progress 42C ***

9 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - <0:09 - ( 0) - [5] 42C
9 200 - 2.5-3-3-6 - 2.6v (2.7) - Stop Small - 2:51 - (10) - [1] 41C
9 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Large - <0:14 - ( 0) - [5] 42C

Not Recent:

9 166 - 2.5-3-3-11 - 2.7v (2.8) - Stop Blend - 21:36 - (64) - [1]
10 166 - 2-2-2-6 - 2.6v (2.7) - Stop Blend - 21:15 - (69) - [1]
11 166 - 2-2-2-6 - 2.6v (2.7) - Stop Blend - 28:55 - (90) - [1] - DEFAULT SPEED
11 200 - 2.5-3-3-11 - 2.6v (2.7) - Fail Blend - <4:00 - (10) - [8]
11 200 - 2.5-3-3-8 - 2.6v (2.7) - Fail Blend - 10:30 - (33) - [1] - Fluke?

9 :200:200 - 2.5-3-3-11 - 2.6v (2.7) - Fail Blend - <0:08 - ( 0) - [2]
9 :200:200 - 2.5-3-3-11 - 2.7v (2.8) - Fail Blend - <0:00 - ( 0) - [4]
9 :200:200 - 2.5-3-3-11 - 2.8v (2.9) - Fail Blend - <0:00 - ( 0) - [15]
9 :200:200 - 2.5-3-3-11 - 2.8v (2.9) - Fail LARGE - <0:08 - ( 0) - [14]
9 :200:200 - 2.5-3-3-11 - 2.8v (2.9) - Fail Small - 2:07 - ( 7) - [1]
---------------------------------------------------------------------------------

OS: Win98 SE
MB: Abit NF7-S 2.0 - (Bios 2.3)
Processor: AMD XP 2500+ Barton
Ram: 2x 256 Corsair PC3200 @ 2-2-2-6-1 @ 333mhz (CMX256A-3200C2)
Video: ATI Sapphire Radeon 9500np (128 bit)
Sound: Creative Soundblaster Live
HDD1: Western Digital 120 w/8mb cache
HDD2: Western Digital 120 w/8mb cache
DVD: Samsung - DVD SD-608 - 40x8
CDRW: Lite-On - LTR-52246S - 52x24x52 - (204310007215)
Case: 3DCOOL Tornado 1000
Power: Antec - PP352X 350 Watt
Monitor: Cybervision - 15" C50
Keys: AT&T - RT101 - 120275-101 Rev B - (52250188)
Mouse: Logitech TrackMan Stationary Mouse - T-CC2-9F (LU069305378)
NIC: Linksys - NC100U Ver 2 - (213005699-E57U-AB-D)
Router: Linksys - BEFSR41 Rev 2 - (Bios 1.45.7) - (C2122034371)
Heatsync: Thermalright SK7
Thermalgoop:Arctic Silver Ceramique

BARTON:
27488
axda2500kv4d 9581488270946
aquca0306tpaw

<P ID="edit"><FONT SIZE=-1><EM>Edited by Ratha on 12/18/04 06:40 PM.</EM></FONT></P>

Increase Vcore, even when underclocking to achieve a 200mhz fsb? Even with +Vcore ive not had any luck reaching 200mhz that dosnt fail prime under any multiplier.

I dont have temps for those overclocks (or underclocks) but i will probably re-duplicate them over the next month of trying to figure this out. I am CERTAIN that they were under 60C when loaded.. in fact, i believe they were under 55C. I think thats the limit i set for myself.

Sometimes i think its the ram that dosnt like 200mhz, other times i think its the FSB.. sometimes the vcore, ive tried to isolate and figure out whats causing the problem.. but nothing seems to fit. Hell, it could be motherboard voltage for that matter, thats something ive not ever changed though.


Ratha

-

-------------CL/RCD/RP/tRAS----------VCORE 1.650v----------Corsair DC Mode-------
MP FSB RAM - Timings - VDIMM (real) - Test HRS:MINS (# Pass) - [Tries]
---------------------------------------------------------------------------------
5:200:200 - 2.5-3-3-11 - 2.7v (2.8) - fail Blend - <0:00 - ( 0) - [15]
5:200:200 - 2.5-3-3-11 - 2.7v (2.8) - fail LARGE - <0:00 - ( 0) - [15]
5:200:200 - 2.5-3-3-11 - 2.7v (2.8) - fail Small - 0:51 - ( 3) - [ 1]

VCORE 1.75
11:200:166 - 2-2-2-6 - 2.6v (2.7) - fail Blend - <0:00 - ( 0) - [1]
11:200:166 - 2-2-2-6 - 2.6v (2.7) - fail Small - 0:05 - ( 0) - [1]
11:200:166 - 2-2-2-6 - 2.7v (2.8) - fail Blend - <0:00 - ( 0) - [1]
VCORE 1.775
11:200:166 - 2-2-2-6 - 2.6v (2.7) - fail Blend - <0:00 - ( 0) - [1]
11:200:200 - 2.5-3-3-8 - 2.6v (2.7) - fail Blend - 9:00 - (31) - [1] - Fluke?

These tests were conducted using a 512 meg Kingston HyperX PC3200 2-3-2-6 module. I could only use this stick for a few days, and here are the tests i was able to perform while i had it.

When i say its unstable, various programs would crash or not load, or just disappear entirely after launching them. For the most part, i learned absolutly nothing with this ram, except that the problem does not appear to be the ram i have. None of the attempts in this table have been published in my previous posts.

-------------CL/RCD/RP/tRAS----------VCORE 1.650v----------Kingston SC Mode-------
MP FSB RAM - Timings - VDIMM (real) - Test HRS:MINS (# Pass) - [Tries]
---------------------------------------------------------------------------------
11 200 - 2.5-3-3-8 - 2.6v (2.7) - stop Blend - 1:00 - ( 3) - [1] 41C
10 200 - 2.5-3-3-8 - 2.6v (2.7) - stop Blend - 1:16 - ( 4) - [1] 42C
10 200 - 2 -3-2-6 - 2.6v (2.7) - fail Blend - <0:00 - ( 0) - [2]
10 200 - 2 -3-3-6 - 2.6v (2.7) - fail Blend - <0:00 - ( 0) - [2]
10 200 - 2.5-3-2-6 - 2.6v (2.7) - stop Blend - 1:13 - ( 4) - [1] 42C
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - fail Blend - 0:17 - ( 0) - [1] 43C +Vcore?
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - stop Blend - 0:19 - ( 1) - [1] 44C +Vcore?

CL/RCD/RP/tRAS VCORE 1.675v
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - fail Blend - 7:35 - (27) - [1] 41C +Vcore?
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - stop Large - :35 - ( 2) - [1] 41C +Vcore?

CL/RCD/RP/tRAS VCORE 1.700v
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - CRSH Blend - 1:13 - ( 5) - [1] 43C Unstable

CL/RCD/RP/tRAS VCORE 1.725v
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - stop Blend - 1:30 - ( 3) - [1] 44C Unstable

CL/RCD/RP/tRAS VCORE 1.750v
10.5:200:200 - 2.5-3-2-6 - 2.6v (2.7) - stop Blend - 0:08 - ( 0) - [1] 45C Unstable
10.5:200:200 - 2.5-3-3-6 - 2.6v (2.7) - fail Blend - 4:13 - (14) - [1] 45C Unstable

CL/RCD/RP/tRAS VCORE 1.650v
10 200 - 2.5-3-2-6 - 2.6v (2.7) - Stop Blend - 2:58 - ( 9) - [1] 41C Unstable
9 200 - 2.5-3-2-6 - 2.6v (2.7) - CRSH Blend - : - ( 0) - [1] Unstable

11 200 - 2.5-3-2-6 - 2.6v (2.7) - stop Blend - 0:05 - ( 0) - [1] 40C Unstable
11 200 - 2.5-3-2-6 - 2.6v (2.7) - stop small - :31 - ( 1) - [1] 41C Unstable

<P ID="edit"><FONT SIZE=-1><EM>Edited by ratha on 12/18/04 09:23 PM.</EM></FONT></P>

I have two more things i wish to try before i do that. I am running a test right now with 10:198:198 and its about three hours in, and seemingly prime stable so far, but when/if it hits 15-20 hours i will call it one way or the other. Need to push it to its limits see if i cant lock it up a time or two during the testing. First comes the idle testing while i sleep.

If i can run 10:198:198 and 11:195:195 then there should be no reason i cannot run 10:200:200 without failing prime tests. It would seem that the 200 barrier is just that, a solid barrier. No matter what multiplier. The 5:200:200 tests show that.

Do you think that if either the northbridge or the southbridge were 'overheating' that it would cause prime instability, but not system instability? Do you think the margin between 198 and 200 could really trigger that? To me it just seems unlikely, but what do i know.

Keep the ideas comming in, ill put em on the list and start working against them to see if i cant break this barrier. You know, im half tempted to try something like 10:205:205 just for the fact that with my luck, 200:200 is just not possible for me to be prime stable on, but 205:205 could be. X_X

Lol.


Ratha

My processor did not like 13:166:166 for some reason at all. Wasnt heat related, it just did not like it, even at 1.725vcore

An update as to the tests over the last day or so. 10:198:198 failed after 5 hours. I started a test 10:205:170 6/5 ratio, and so far its been running nearly 10 hours. Will put some outside stress on it in a little bit to see if i can cause any problems in the testing or stability.

Whats this say to you guys? Ive used two types of PC3200 ram that have REFUSED to run prime stable @ 200mhz, but i can run 22 hours prime stable @ 11:190:190 and 15 hours stable @ 10.5:195:195, yet i cant run 5 hours stable at 10:198:198.

After reviewing all of the tests ive done, again, with all the new data ive collected in the last week, i am starting to think its something related to ram. 11:200 is too much using near stock voltage settings for my cpu it seems. However the tests i have done which were 9:200:166 and 10:200:166 both came out above 20 hours stable.

I have multiple tests with 11:166:200 which were not prime stable for more than 4 hours except for one, which was ten and i think that was probably a fluke. I think the next test i should try is probably a 10:166:200 just to be sure. I believe that will fail pretty quickly because of the ram being run at 200mhz.

What would cause there to be a problem with both the single channel 512 meg stick of kingston, and the two 256 meg dual channel sticks of corsair in this thing?

Ive tried upping ram voltages before, that didnt seem to help, tried vcore voltage, that didnt seem to help. The only thing i have not tried is upping the chipset voltage. Do you think that could/would be the cause of this problem? I know nothing about chipset voltage and have no way of measuring its temps so im not so likely to play around with it.


Ratha

-

-------------CL/RCD/RP/tRAS----------VCORE 1.675v----------Corsair DC Mode-------
MP FSB RAM - Timings - VDIMM (real) - Test HRS:MINS (# Pass) - [Tries]
---------------------------------------------------------------------------------
10 198 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - 5:45 - (19) - [1] 42-45C 27-30C

Just checked Abit website, seems there have been some bios revisions, flash to the newest version. You can also find a memory compatibility list on the website.

Just to let you know ratha the corsair memory your using is rated for 2-3-3-6 @ 3200 on intel systems. It is only rated for 2.5-3-3-6 on AMD systems. I had similar memory in my system then I switched to Patriot memory for better timings. As far as your CPU goes it probably needs more voltage like most people have suggested already. I have a bad chip and I couldn't hit 3200 speeds until I put my vcore up high. Look at my sig for settings. Just remember that higher vcore will shorten the life of your CPU.

Xaser 3 case with 420W TT PSU
AXP2500 (11*200 at 1.85v)
9800 Pro 128MB (418/372)
NF7-S v2.0
PDP DDR400 1GB(2-3-2-5)2.6v
Lite-On SOHW-832S
WD 80GB 8mb buffer caviar edition.

Hmm, where to start..

I am attempting to try more voltage, but so far nomatter what multiplier and fsb i have set, when the ram is set at 198 or 200mhz, it will fail prime tests. The case of 5:200:200 proves this.. thats what an 800mhz underclock from 2500+ default speeds? I shouldnt need 1.8vcore in order to underclock 800mhz. But i am attempting to try the vcore to solve the problem. Windows becomes unstable at 1.75v using 10:200:200. It does not seem logical that i need more vcore to run 5,6,7,8,9, or 10:200:200 when i can run 10.5:195:195 for 15 plus hours prime stable, and 11:190:190 for over 22 hours prime stable.

If anyone can tell me why i cant get 5:200:200 to run prime stable for more than 1 minute on the blend tests, and less than an hour prime stable on the small fft, well i would appreciate it

I upped the ram voltage a test or two ago to 2.8, since my power supply kindly supplies .1v over what i set it in the bios to supply, and anything over 2.9 voids the warranty on my ram. However that did not solve the problem either.

I know the timings on this ram are 2.5-3-3-6 for AMD boards @ PC3200 speeds, but thanks. For All of my recent tests i have been running it at those speeds to rule out timing problems. However it runs 2-2-2-6 perfectly at pc2700 speeds, and since neither previously, nor now have i been able to get my ram to run stable at pc3200 speeds, pc2700 (default cpu fsb) was the next best choice for leaving my computer at until i was able to 'fix' it.

I will keep the ram, since i can afford no other, and paid out the ass for this stuff when i got it. Unfortunatly Corsair kindly forgot to mention with their nifty little "*" like they have now on their site that this ram is NOT rated for CL2 on AMD boards @ pc3200.. Otherwise i would have bought some different ram, or even the non CL2 stuff. Since the ram works, its hard to believe that both the Kingston module i borrowed from a friend, and my Corsair are bad. Its more likely to be possible my motherboard is bad, or my cpu is bad.

The power problem is not related to a bad power switch connection. The computer turns on fine, and i have checked that 'jumper/wire/switch' multiple times to make sure i didnt have it connected wrongly. When i hit my power button to shut down, or use windows to shut down, it goes into an endless loop. Never any other time. No one so far has been able to figure this one out. Im THANKFUL that its not like my other computer, where i could not turn it on (with my videocard plugged into the AGP slot) without hitting the reset button upwards of 300+ times.

(old motherboard) Ran fine with the videocard removed from the motherboard and not powered, would post properly except for the no videocard beep codes. When the videocard went into the slot however, (powered or unpowered) thats when it would take multiple resets (300+ sometimes) in order to get the thing to post. It wouldnt even post enough to tell me that the videocard was not powered like the 9500 will do if you unplug the power to the card and try to post the system.

Anyway, back to the problem at hand.. The solution to this problem still eludes me. At this point i have run about every multiplier:fsb:ram configuration that i can think of to run, and it will not run prime stable with the ram at or above 198mhz. 195 tested fine. Ive tested with an FSB of 166 and the ram at 200, and it fails the tests.

I would love to have someone else see whats going on here to prove that im not just entirely incompetent lol. I have explained whats happened and whats happening so many times to so many different people and on so many forums and tested so many different things, i am unsure that i will ever find a solution to the problem.

Cut and dry. No ram i have tested in this board has been able to run 200mhz without failing prime usually within an hour. Bad ram? No.. probably more like bad motherboard, or bad chipset, or bad something, but if the ram was bad memtest86 would've given errors.

I cant imagine what the problem is.

I have a few more things i intend to try, play around with the CPU Interface option a little bit see if that dosnt result in any positive results (has not so far.) I believe i have changed the FSB:RAM divider manually to a non "Auto" setting, but for this battery of tests i have left it set to auto for all the 1:1 tests. I may try 3:3, 4:4, 5:5 and 6:6, as i have heard people getting better results trying different 1:1 dividers before.

If you guys have any more ideas, i would certainly welcome them. I need all the ideas i can get at this point cause i have about run out.


Ratha

-

Here are the most recent tests i have done. Unless otherwise noted Vcore is 1.675. All these tests have failed. Ones that fail quickly i have tested multiple times. Generally after 5 times i give up and try a different set of tests. If i am able to get one to run longer, i try to figure out how i can gain on that time. On the unstable test in this table, i didnt even bother to let prime run. Windows were crashing down around me.


-------------CL/RCD/RP/tRAS----------VCORE 1.675v----------Corsair DC Mode-------
MP FSB RAM - Timings - VDIMM (real) - Test HRS:MINS (# Pass) - [Tries]
---------------------------------------------------------------------------------
10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - <0:10 - ( 0) - [3] 41-44C 27-29C
10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - 1:08 - ( 3) - [1] 41-45C 27-30C
10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - 1:23 - ( 4) - [1] 41-45C 27-30C

10 200 - 2.5-3-3-6 - 2.8v (2.9) - Fail Blend - <0:00 - ( 0) - [5]

10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - <0:07 - ( 0) - [4] - CPU Interface Disable

10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - <0:07 - ( 0) - [2] - Vcore 1.700v
10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - 1:15 - ( 0) - [1] - Vcore 1.700v
10 200 - 2.5-3-3-6 - 2.6v (2.7) - Fail Blend - <0:08 - ( 0) - [4] - Vcore 1.725v
10 200 - 2.5-3-3-6 - 2.6v (2.7) - - Unstable Vcore 1.750v

I've got a hunch that that motherboard just can't run the RAM at 200. But... What if you were to (if you can) get some PC3500/3700/4000 ram and see if it will do it at 200. Could be some timing in the mobo itself that can't be adjusted in BIOS that pushes the ram too hard, while faster-rated RAM may survive.

It's really a WAG, but if that doesn't work, I'd have to say the mobo is the problem.

Mike.

Getting higher rated ram unfortunatly is not possible at this time, limited finances and all. I have considered that it could be the motherboard a few times, but again i dont have any way to rule that out unfortunatly.

I checked about two weeks ago to make sure i had the latest version of prime95, seems the one i was using earlier in the year was the latest.

The next thing i intend on trying after the current test is finished is upping the chipset voltage to see if that wont fix the problem. Right now i am running a 250mhz overclock with a 166mhz bus (10.5:166:166), on stock voltages for everything, using the tightest ram timings i have tried.

This was just one of those 'because i know it would work' kind of tests. I needed to run a test i knew wasnt going to fail, so this was to help paint a picture of what i can do. As opposed to ruling out everything i cannot. Ive done a lot of that already.

Anyway, still working on it. I dont expect to find a solution to the problem unfortunatly, but i can probably get by for the time being with a moderate overclock till ive found a better solution to the problem.

Thanks for all the assistance and ideas. Will still be open to more ideas if anyone should come up with any.


Ratha

Well, unfortunatly, the motherboard voltage did not make a difference. It defaults to 1.6 and 1.7 is max. For a test i set vcore to 1.75, vdimm to 2.8, and the voltage for the motherboard to 1.7. I had no luck running 10:200:200. Or any other iteration of 200:200.

Ive already proven that the 200mhz fsb can be broken, something to do with the ram, but not the ram itself seemingly is the cause of the problems which im having. Im tempted to just try a random out of the blue test of 235:195 (6/5) Roughly 9*235.

I am wondering how much difference there would be in performace using 9:235:195 instead of 11*195*195. If i can even run it for that matter. I have not been able to get 11:195:195 to run properly yet for some reason. I think the area around 2100mhz is the drop off point for lower than 1.75 vcore.

So as soon as i get a few good stable numbers i can pit against eachother that i want to compare ill follow that path.

Hard to say whats next to test. Will see what happens over the next day or two and check back if anythings new.


Ratha

Had this rig for a year. I bought the 2500+ from Addiamadar for a back-up. Time to retire the T-bird 1.3. It's been a lot of fun to play with, but, I wish I had just bought a Mobile chip. Not all of the XP's overclock well unlocked or not.

Abit IS7 - 3.0C @ 3.6ghz - Mushkin PC4000 (2 X 512) - Sapphire 9800Pro - TT 420 watt Pure Power
Samsung 120gb ATA-100 - Maxtor 40gb ATA - 100
Sony DRU-510A - Yellowtail Merlot