New AXP-M build problems + UPDATE2

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Ok, I tried with the CPU at 8x200MHz and the RAM at 100MHz (200DDR) to rule them out as a possible cause for the failures and just to test how the FSB behaves at 200MHz (400DDR). Prime ran for 8hours before I aborted it - it's pretty obvious it doesn't have any problems with that FSB speed I guess, right?

Next I will try the CPU at 11x200MHz and the RAM still at 100MHz to see if it's the CPU and I have a feeling it will be fine again. If it indeed has no problem with the CPU at 2200MHz and the FSB at 200MHz (400DDR), then that would leave the RAM, right? But why then Memtest completed successfully 32runs with no errors, before I aborted it at the 8th hour of running?

Original post below:

===

Hi! I just got an AXP-M 2400+ with an EPOX 8RDA+ Pro motherboard and 512MB DDR400 Kingston Value RAM. The machine is running at 3200+. Thing is - Prime 95 runs into errors regularly (sometimes getting an error pretty quickly at the first test, sometimes at the second) and I've seen a couple of random restarts (Windows XP blue screens - I disabled Automatic Eestart).

When I run Prime 95 the CPU gets at most to 45C - 46C which I think is acceptable. I checked voltages and with Prime 95 running the CPU hovers ~1.568-1.584, the DIMM voltage ~2.704-2.736. What are the acceptable ranges for these values? Should I up the voltage for the DIMM or CPU or both? If yes, by how much? I intend to try a friends DDR400 512MB stick and see how it goes with that, but I'm willing to try other options if there are any.

About the random restarts, I have seen two so far, and both were during regular Windows usage and not playing games or torture testing. After the first one, I exchanged the CPU fan for a better on (GlacialTech, all copper, 2300RPM which keeps the CPI pretty cool I think) and I switched power supplies from a generic JNC 350W to a 120mm fan Fortron 300W which I think is good enough for my system (1 HDD, 1 CD-RW, Radeon 9000 Pro). The second restart occured tonight, and I'm not sure whether this caused it in any way, but I queued up like 4 files for download in Firefox, along with the loading of a QuickTime movie inside the browser... seems like pretty routine usage which shouldn't present any problems to the system but I thought I'd mention it. Should I try updating the motherboard drivers? What can I try to fix this?

I'm not very concerned about Prime 95 as long as I don't get any automatic restarts, but I think that somehow the two problems are connected and I'd like to fix them somehow. Thanks in advance for your replies and ideas!

UPDATE:
The system completes Memtest with no errors (running for 8 hours).
CPU settings are 11x200MHz, 1.600V (tried it at 1.65 but without any success, i.e. Prime still bombs).
RAM timings 3-3-3-11, voltage at Default.
Haven't experienced a restart for awhile... granted I'm not using it much during the week.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Crap! Just got blue-screened again. This time the error was in ati3dnag.dll. The previous restart was not caused by this (i.e. I got a different error), but I'm not sure exactly by what.

A bit of info I forgot to supply in my previous post:
++ CPU is running 11x200MHz (which I think equals 3200+, right?).
++ RAM timings are 8/3/3, CAS3.
++ CPU and DIMM voltages are at their default / auto values (what are they by the way?).
++ AGP is at 1.5V and the chipset is at 1.6V.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Snakexor, Thanks! Are those values safe? I sure don't want to end up with a dead CPU. ;)
What are the default voltages? And upping the CPU voltage will increase the CPU temperature, right? What is the highest temerature (under load) I should shoot for?
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Just got that same ati3dnag.dll blue screen. Both times trying to play the same AVI file. Will try the latest ATI drivers, do you guys think that'll help? I'm running some pretty old Catalyst 3.7s.
 

volrath

Senior member
Feb 26, 2004
451
0
0
um... its an ati error. and you are running old ati drivers. i think the solution is obvious?
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Volrath, W00t, thanks! I installed the latest Catalysts from ATI's site and now this particular error is gone and the AVI plays fine.

What are the default voltages for the CPU and DIMM? Right now they are at their auto values in the BIOS.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Crap, just got another error, this time in ati2cqas.dll. I'm beginning to regret I chose to go with an AXP system. :(
What should I do? Will an update of NForce drivers help?
 

Lonyo

Lifer
Aug 10, 2002
21,938
6
81
Default Athlon XP voltage is 1.65 for a normal chip, 1.45 for a mobile chip.
Try it at 1.65 and see how it goes.

Also, get Super Pi and run the 2 million test, and see what error it gives.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Lonyo, Super Pi gave me a 'NOT EXACT IN ROUND' error during loop 18.
I reduced the CPU clock to 10.5 x 200MHz, but the CPU voltage remaining at 1.56. Just to make sure, is 1.65V a safe value for the CPU? And what are the idle / load temps I'm shooting for? Thanks again!
 

CZroe

Lifer
Jun 24, 2001
24,195
857
126
Sounds to me like you need the good ol' L12 mod. It fixed my problem.

Please run MEMTEST86. I'm willing to bet that you'll get errors at higher FSBs even though both the CPU and the RAM is fully capable. The problem is that the motherboard thinks that you are way out of spec, farther than it thinks you COULD be and something is tuned for a MUCH lower FSB. This is because the motherboard does not recognize the mobile CPU and the bridges that signal a default FSB are not interpreted correctly. I tested this theory by doing the L12 mod myself. All the L12 mod does is make the CPU's default seem to be 400FSB (200MHz) and yet without it, the FSB can't pass MEMTEST86 at anything higher than 333FSB. So by simply telling the motherboard that this is a 400FSB CPU it works. All it takes is a tiny little wire dropped into two specific holes in the CPU socket.

I now have that same CPU and motherboard but I'm not using them together.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
I tried the other two tests in Prime 95 and Small FFTs passed with no problem the first two series of tests (after which I aborted it), In-place large FFTs was working okay for 20mins before giving me an error, and Blend craps out in the first 10mins consistently. Should I assume that since in the latter 2 tests the RAM is tested more thoroughly the problem is in it (RAM that is)?

Will run Memtest if I ever find a way to install it.. gives me a pretty wierd error: "Can't figure out how many sectors/track for this diskette." Perhaps this is because I'm using an USB floppy drive? Anyway.. will find a way around that.

CZroe, thanks for the imput, but I'm not sure I'm ready to drop a wire into the CPU socket just yet :) The CPU starts okay, and Windows runs pretty good, only Prime 95 crapping out. Maybe I should try a BIOS update or something to work around the problem? Or reduce memory timings?
 

imported_whatever

Platinum Member
Jul 9, 2004
2,019
0
0
You can get ISOs for bootable versions of Memtest if you want, so that you just burn it to a CD. I would definately reccomend using one.
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Whatever, yes.. that's what I'll do :)

2 Questions:
++ What is this Memory Auto Precharge setting in my BIOS and should it be enabled or disabled?
++ In the BIOS I have set the CAS Latency to 3, but CPU-Z still read it at 2. Why is that?

Is it just me, or is the latest BIOS posted on EPOX's webpage corrupted? AWDFLASH.EXE says that the BIN is not an Award BIOS.. Wierd!
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Even more wierd.. I'm pretty sure my DIMM stick is in Slot #1, but CPU-Z reports that it's in in Slot #2.
Also the SPD Timings Table specifies that the CAS latency for my memory at 200MHz is 3, which is what I have set it to, but CPU-Z reports that it's 2. Help!
 

CZroe

Lifer
Jun 24, 2001
24,195
857
126
It may be reporting the capabilities as reported by the SPD chip on the module and not the current setting. Also, the wire in the socket is even considered necesary for attaining the highest FSB with the 2 best overclocker's boards; nf7, A7N8X (Though they seem fine without it at 400FSB and a little above, they can go sky-high with it). It's really simple. It may feel like the CPU wont fit but I just pushed mine in and it worked great. Most people suggest that you shred a phone cord and use the thin-guage wires from it. I used the boot CD for MEMTEST86. Remember, it's not only testing the memory, it's actually testing the FSB. I'm willing to bet that your system is rock-solid at the same frequency with a lower FSB and higher multiplier. That would prove that it's the FSB and not the memory as well as prove that the motherboard is to blame (All AXP-M CPUs should be 400FSB capable at 3200+ settings and voltages).
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Ok, so what happens if I mix up the holes in which I have to put the wire? Do I end up with a fried CPU / Motherboard? :)
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Also, how much of a performance loss should I expect if I decide to run the CPU instead at 11x200MHz (2200MHz with a 400MHz bus), say at 14x166MHz (2300MHz with a 333MHz bus)? And with that in mind, if I decide to do this I can still keep the RAM at 400MHz, right?
 

Ze Mad Doktor

Banned
Dec 9, 2004
137
0
0
How do you know the CPU is capable of running those speeds? If you set it to default settings do the errors continue?
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
Well... I don't know how exactly to answer this Q... The CPU doesn't run too hot, it completes the memory-light Prime-95 with no problem (it was me aborting the test). And.. I tried it at 10x200, which I think is 2800+ and what kind of AXP-M 2400+ whould it be if it can't run at 2800+, right? :)
 

CZroe

Lifer
Jun 24, 2001
24,195
857
126
Originally posted by: mooojojojo
Ok, so what happens if I mix up the holes in which I have to put the wire? Do I end up with a fried CPU / Motherboard? :)

I just followed some images I found online. Don't remember it being hard to identify them.

Originally posted by: mooojojojo
Also, how much of a performance loss should I expect if I decide to run the CPU instead at 11x200MHz (2200MHz with a 400MHz bus), say at 14x166MHz (2300MHz with a 333MHz bus)? And with that in mind, if I decide to do this I can still keep the RAM at 400MHz, right?

Well, FSB overclocks probably yield the most tangible benefits so you should try to get as much as you can out of it. If you keep the RAM at 400(200MHz) and the FSB lower then the motherboard chipset will have to buffer memory reads and writes causing higher latency. Synchronous operation is faster than async. Single channel DDR400 is faster than dual channel DDR266. But the fact remains: Every AXP-M can do 400FSB. It's the 400FSB-capable motherboards doing something weird that sometimes prevent them.

Originally posted by: Ze Mad Doktor
How do you know the CPU is capable of running those speeds? If you set it to default settings do the errors continue?

Because every AXP-M is a bin-sorted cream of the crop 3200+. The 3200+ CPUs that could run stably at lower voltages (The better ones) were clocked even lower, the voltage further lowered and sold as AXP Mobile CPUs. Because it IS a 3200+ CPU, it can handle 3200+ settings as long as you put the voltage back up to 1.65 (3200+ voltage).
 

Ze Mad Doktor

Banned
Dec 9, 2004
137
0
0
Yes, but before you can worry about problem solving an overclocked computer you have to know that it works when it is not overclocked. Soooo, set it to it's normal settings and run your tests and see. If that works, keep it at the same speed but use the 400MHz bus. If that works then work on your increases and try and figure out where the errors start. When they do then give it more juice until they stop. then continue.

Because every AXP-M is a bin-sorted cream of the crop 3200+. The 3200+ CPUs that could run stably at lower voltages (The better ones) were clocked even lower, the voltage further lowered and sold as AXP Mobile CPUs. Because it IS a 3200+ CPU, it can handle 3200+ settings as long as you put the voltage back up to 1.65 (3200+ voltage).

this is not a guaranteed thing. It may be true 95% of the time but there will always be exceptions.

Edit:fixed UBB code
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
CZroe - Thanks again, though I'd resort to the wire-in-the-socket mod only as a last measure. ;)

Ze Mad - I'll give it a shot, bumping gradually the OC settings, but wouldn't it be strange if the CPU can't take that FSB / frequency and still run at ~35C idle and ~45C loaded?
 

mooojojojo

Senior member
Jul 15, 2002
774
0
0
CZroe - Just a quick clarification. This is one guide I found on making the L12 mod.

http://www.geocities.com/trats102002/nf7sFsbsense.html

If I understand correctly, and from looking at the table provided in the link. Since my CPU is a Mobile, I only have to bridge ah30 to ah28, right? Excuse me for being so thorough but it's just that this is all oh-so-confusing! Thank you in advance!