Help OCing my first AMD - an Opti 165

Ran full Memtest overnight - 19 passes with no errors. Any idea what's causing the problem or something new to try?
 
No major change. SP2004 still halted after about 45 sec on Large FFT and Blend. Small FFT still okay. :(
 
Switching to the other two slots give me all kinds of problems - no error message, but the system just hangs during boot. Switching back to the regular slot, boot is fine again.

Edit: Any idea why my CPU speed is shown as 2592 instead of 2605 like we calculated?
 
Popped ClockGen back open and started downclocking :(

At 2458 CPU, 273 HTT, and 223 RAM things seem to be stable again. At least, SP2004 Blend doesn't stop. I still get those "paging" blocks every so often, but I don't think that indicates instability.

Judging by the fact that the Small FFT test was just fine at 2592mhz, I'm assuming that either the HTT or the Ram require these lower timings? Bah. This is so weird, considering that the Ram passed Memtest at 237. I'm assuming its the RAM, and not the HTT that's causing the problem. Any idea of what to do next?

Or should I look into gettting a decent CPU clock of at least 2.5 with Ram at 223 somehow? Bah. I hate to move backwards :(
 
Uh oh. About 1 hour, 24 minutes in, Windows is telling me that my page file is to small. Maybe all that paging is a bad thing? I've not increased the size from default, but it doesn't seem logical to me that a CPU/RAM test should be paging to HDD, rather than the RAM itself :confused: :confused:

Is this how SP2004 is supposed to work, or is it indicative of a larger problem?

Edit: Test has run for over three hours with no further warnings. I wonder if it had to do with all the stopping and starting to find a low enough frequency. Still makes me edgy
 
Xaeos said:
Uh oh. About 1 hour, 24 minutes in, Windows is telling me that my page file is to small. Maybe all that paging is a bad thing? I've not increased the size from default, but it doesn't seem logical to me that a CPU/RAM test should be paging to HDD, rather than the RAM itself

Is this how SP2004 is supposed to work, or is it indicative of a larger problem?

Edit: Test has run for over three hours with no further warnings. I wonder if it had to do with all the stopping and starting to find a low enough frequency. Still makes me edgy
Click to expand...
i love that. note carefully, it tells you the minimum is too low :rolleyes:
wth is that supposed to mean anyway?

for SP2004, can you manually set how much ram it allocates? i haven't really messed with the program, as i still use the older prime95


Xaeos said:
Switching to the other two slots give me all kinds of problems - no error message, but the system just hangs during boot. Switching back to the regular slot, boot is fine again.

Edit: Any idea why my CPU speed is shown as 2592 instead of 2605 like we calculated?
Click to expand...
mmk. guess that plan is out.. haha
different speed is most likely due to a rounding error, nothing more.


Xaeos said:
Popped ClockGen back open and started downclocking :(

At 2458 CPU, 273 HTT, and 223 RAM things seem to be stable again. At least, SP2004 Blend doesn't stop. I still get those "paging" blocks every so often, but I don't think that indicates instability.

Judging by the fact that the Small FFT test was just fine at 2592mhz, I'm assuming that either the HTT or the Ram require these lower timings? Bah. This is so weird, considering that the Ram passed Memtest at 237. I'm assuming its the RAM, and not the HTT that's causing the problem. Any idea of what to do next?

Or should I look into gettting a decent CPU clock of at least 2.5 with Ram at 223 somehow? Bah. I hate to move backwards :(
Click to expand...
there's an easy way to find out if it's the ram or HTT.. drop the memory ratio down one notch. if it's still not stable, drop the cpu multi too. that'll prove that it's the htt.

if it makes you feel better, i PROMISE you won't see any difference between 223mhz and 237mhz ram. only in synthetic benches is there a slight difference :p
 
Haha. Yeah you're right about the minimum thing. I'm getting too paranoid! :p

SP2004 doesn't seem to have any manipulative abilities so far as CPU or RAM are concerned. You can have it interface with MBM5, but that seems to be about it.

Problem - I came back after 11 or so hours and it turned out that one of the cores crashed out at around the 4 hour mark, the other was still running. So bah. Still unstable even at 220 Ram.

Went back in, changed mem divider down to 133mhz which according to CPU-Z, sets the RAM at about 185. CPU and HTT still at 2592 and 288 respectively. Letting the test run...

Edit: Crash after 5 min. Going back to set the CPU Multi down a step to 8x. HT Link Multi is still at 3x. Should the HTT or FSB still be 288 after making those changes? Clock Gen shows CPU at 2304, HTT at 288, and RAM at 192 now, with Mem divider still at 133.
 
almost sounds like it's not the ram.. or the cpu? you have weird symptoms.

and yeah, leave the htt at 288. it's possible that the board just doesn't like going that high.. but that would be kinda odd.
 
Alright, the test has run for more than 10 hours without stopping. So...HTT is the problem somehow? CPU is 2304, HTT is 288 (?), and Ram is 192.

Ram crashed at 185 and not 192, so it can't be the ram's problem CPU Small FFT test works just fine all the way up to 2605 :confused: So... what changed here? Did something related to the HTT change when I moved the CPU multi from 9x to 8x? HT Link Multi is still at 3x. According to ClockGen the HTT is still at 288, the same number it was when it was crashing before :confused:

This seems...odd.
 
I'm thinking there is something wrong with the settings you have for your RAM. What can you change ram-settings wise on your mobo?
 
Hmmm... I can change the standard CAS etc.. settings (3-4-4-8), Timing to 1T or 2T, Voltage,and a number of the smaller tweaks like TRRD (if I remember correctly). This stuff is supposed to do DDR500 at 3-4-4-8, between 2.6 and 2.8v.

To my realtive inexperience, it seems that if the thing crashed at 185mhz but is stable at 192, the ram is not busted and it must be something else huh?
 
Went back in and upped the memory divider to 166mhz from 133, but leaving the CPU multi at 8x. New stats are...

CPU - 2304mhz
FSB/HTT- 288mhz
RAM- 230mhz

So far, SP2004 has been stable for an hour...lets see if it keeps up. Arctic Silver must be kicking in. My load temps are down to 48 or 49 at CPU :)

If this is stable, then it stands to reason that the RAM is okay, yes?
 
Alright. That's 14 hours of tenative stability. So what's the deal here? The 8x CPU multi must be afffecting something else, huh?
 
Make that 19hours. Should I just concentrate on raising the base clock while leaving the CPU multi at 8x? To get it up to 2600, I'd need to raise it up over 300, and I get a warning at POST about doing that. "please set your base clock lower etc...". I'd still like to know what exactly, besides CPU speed, is affected by an 8x multiplier as opposed to a 9x?

Ideas as to how to progress?
 
To be complete, I stopped the test and switched it over to Large FFT, as that was giving me problems as well as blend, previously. Currently, stable for 2 hours.

Is there no way to get to 2.5/2.6? :(
 
Nope - in order to get to 2605 on the CPU test I changed the voltage up to 1.500 from 1.350 a long time ago. All the current tests are at 1.500
 
hmmm, interesting.

i guess stay away from small-fft then, the limiting part of your oc must be in the memory controller or something.

try this.
what's the maximum ~1-2 hours prime95 blend stable at 1.35v, 1.4v, 1.45v and 1.5v?
the way it scales with voltage should give us some useful info
 
So far small FFT hasn't given me any problems. Should I go get Prime 95 or is SP2004's blend (2 instances) equal?
 
Okay, Voltage tests...

Base stats across the board for all tests...

Memory Voltage = 2.8v
Base Clock = 289
CPU Multi = 8x
HT Link Multi = 3x
Mem Divider = 166 mhz
Mem Timing = 3-4-4-8 1T

Which Means...

CPU = 2304mhz
FSB/HTT = 288mhz
RAM = 230mhz

Now the results -

CPU 1.500v = STABLE (19 hours)
CPU 1.450v = STABLE (2 hour test)


I'll edit with new results until I get down to 1.350 which is the system default.

Edit: I just read your post another way. Did you mean that I should go back to the CPU overclocking step of the OC guide at the different voltages? Ie. Change all the specs back?
 
Xaeos said:
Edit: I just read your post another way. Did you mean that I should go back to the CPU overclocking step of the OC guide at the different voltages? Ie. Change all the specs back?
Click to expand...
yes ;)

if you follow the guide to the letter, you should know the max mhz at stock up to the highest voltage tested in small voltage increments... something like this:

1_77.png
 
I have the same memory as you, and I had problems getting it over 235 at first, but after I set my chipset voltage and ldt voltage up a bit, it had no more problems. Was able to get well above 250 at 1T. If you have that option in your BIOS, give it a try. Usually when trying to find the upper limits of my memory, I keep my CPU clocks low to eliminate it as a possible failure point. That way I know for sure what component I've maxed out.
 
I'm not sure I have a chipset or LDT Volt option on my mobo - I'll have to check. It isn't with the other OC options.

Okay... CPU SP2004 Tests - Small FFT

Stats- (I keep including these so nobody has to refer back, and also so eclipse can see if I goof up somewhere heh)

Memory Voltage = 2.8
Base Clock = 200
HT Link Multi = 3x
CPU Multi = 9x
Mem Divider = 100mhz
RAM at 3-3-3-8 T1

CPU Voltage Test - 1.350 - stock Highest stable.

CPU = 2445
FSB = 271
RAM = 135

CPU Voltage test - 1.400

CPU = 2525
FSB= 280
RAM=140

CPU Voltage tes 1.450

CPU = 2559
FSB= 284
RAM = 182

CPU Voltage Test -1.500

CPU=2589 ???
FSB= 289
RAM = 144

I may have to keep editing. I'm going to run the tests for an hour for stability - sometimes they havn't crashed out until the 10 or 20 minute mark before.
 
Well there we have it, tests complete. I still may need to edit the 1.5v test again. I'll check to make sure the test is still running in an hour.
 
Bloody hell. 1.5v test crashed after 1 hour, 45 min at 2605. For some reason, even if I get clock gen reading 2600, when I hit apply it goes back to 2605. Anyway, downclocked it to 2589 - seeing if this holds. In either way, the max is still under 2.6 :(
 
don't use small fft's, use blend or in-place... sorry for not saying that earlier :p
 
*Head Explode* Okay, I'll start over again :p :D

Blend Test - going to run 2-3 hour test cycles. Sometimes it seems things don't crash immediately, but rather after 2 hours. Makes the process take awhile :eek:

1.5v = 2574mhz
1.45= 2544mhz
1.4= 2486mhz
1.35= 2417mhz

Woo. Finished
 
Whew. All done. Unless the 1.35 throws an error in the next hour or so, those seem to be the finals.
 
So where should we go from here?

I was thinking moving back up to 1.5v and setting base clock to 284?

That way... (lets see if I've finally got this right)

cpu= 284*9= 2556
ram divider = 9 / (5/6) = 10.8 > 11
RAM= 2556 / 11 = 232.36mhz

Or is there something else I should do?
 
Tried setting the above... which gives me

CPU 2548
FSB 283
RAM 231

Crash within first 30 sec of prime blend. What the hell? :confused:
 
ARGH! Lowered base clock even more to 280 (still at 1.5v)
CPU 2518
FSB 279
RAM 228

Four minutes before crash. What the hell is going on here?
 
Tenative stability by setting the ram divider to 133mhz instead of 166

CPU 2518
FSB 279
RAM 179


I really hate having the ram so slow. This is getting weird.
 
Set the dividers up to 166 again. General failings all around. I've tried upping and lowering everything a few mhz with clockgen, and I get different times of failure. Sometimes its fairly immediate, like in the first minute. Other times its 45min-2 hours.

Way out in left field theories of mine, probably with little basis in reality:

1. Is it possible I've overvolted the RAM? I have it set at 2.8v, but the specs call for 2.6-2.8. Am I running it too slow for the voltage or something?

2. HT link multi is at 3x. I suppose I could change to 4x or 5x? think it would make a difference?

I can't believe this ram has to be run under 200mhz :(

Any ideas from the Guru (s)?
 
sounds to me like a ram problem......

have you followed the guide to find the max ram stable runnings along with your CPU tests?

baldy
 
Check earlier in the thread for more data, but just for the hell of it I'll run memtest again.

Highest Base Clock that doesn't cause test 5 to error

CPU Voltage = 1.5
Ram Voltage = 2.8
Cpu Multi = x5 (temp for the test, as indicated on the OC guide)
HT link multi = x3
Mem divider = 200mhz
Timing = 1T
Config 3-4-4-8

Base clock = 232mhz ??
 
You must log in or register to reply here.
Back
Top