AMD 3700+ San Diego - overclocking stability - 1st timer

S

sham

n00b
Joined
Dec 4, 2005
Messages
50
Hello Everyone,

I have an AMD 64bit 3700+ San Diego. It has been overclocked from 2.2Ghz - 2.7Ghz. The HT Multiplier is at 4x and the multipier the same. I have just increased the fsb. The voltage is at stock. The idle temp is 33*c.

The problem I have is to do woith stability. When using prime 95 and doing a SMALL FFT TORTURE TEST and "memtest" together, they both work without any problems for over 10 hours - and more.

THE PROBLEM>>>Yet when I run just prime95 on small fft's without "memtest" an error comes up after 10 - 30 minutes. It says a fatal error, and Rounding Test 0.5, expected less than 0.4.

Is theer a reason why this only happens when i am not running memtest?

Does this mean the overclock is not stable - i have no problems using the PC at the moment?

Is tehre a way of solving this or is it first time builders and overclockers paranoia.

Thanks in Advance
 
I have the same problem. It's just too much of an oc with stock voltages.
 
Thank you for your reply.

yet I still dont understand why the rpime95 only fails when memtest is not running.

Could someone please clarify.

Thanks
 
It could just be instabilty error that it doesn't give errors with both running. It would be best to use both at seperate times, rather then together. So you must have instability. That's about as much as I can tell from this. Have you tried running memtest alone? Have you tried downclocking it below 2.7 to maybe 2.6?
 
yes, the memtest alone works fine

The problem is that when i do just the prime95 i get that error.

I will try and overclock it down to 2.6Ghz.

BUT, could you tell me what the 0.5 instead of 0.4 error means. Is there a difference between the small ffts and the large ffts etc.

Thanks in advance.
 
i have lowered the overclock slightly. will rport back any errors.

does anyone know why it inly happens in the above circumstances though.

Thanks a lot to everyone who has replid so fats. I really appreciate it! ;)
 
well that mean you passed the CPU max oc limit. :-

Theres a CPU stablility and Memory Stablility

Memtest is for Memory stablility, Prime95 is testing the CPU stablility, if you failed the Prime95 that concludes:
1. not enough voltages to the CPU. (increase the voltages to get the CPU to stable)
2. Unstablilize.
 
thank you

I now have just two more questions./..sorry if they sound a bit stupid!

1. Why does this only happen when i dont run memtest and prime95 together?
2. What does the "rounding 0.4, less than 0.5" mean?
What should I decrease to stabvilize, without increasing voltage?

Thanks
 
thanks for reply.

i havce been talking to a friend, and tried summin confusin to mdo with memory. the problm he found is thatr when i got to cupid, the cpu/memory thingy is CPU/15 when overclocked.

When I keep it as standard - nmt overclocked - it is still at CPU/11.

I have PC3200.PC4400 ram. Do you know why thi is happenin.
 
i do not use prime95 to test my CPU stability, because even though prime95 failed after 10 mins. but i can run 8 hrs stable with Stress Prime 2004 by Johnny Lee. maybe you can try this program and see if you will get any errors.
 
sham said:
thank you

I now have just two more questions./..sorry if they sound a bit stupid!

1. Why does this only happen when i dont run memtest and prime95 together?
2. What does the "rounding 0.4, less than 0.5" mean?
What should I decrease to stabvilize, without increasing voltage?

Thanks
Click to expand...

1. Prime95 by default sets its process to the lowest priority, meaning all other processes running take priority over it, namely memtest, which must take up most if not all of the cycles and prime95 gets ignored, therefore it doesn't have a chance to make an error. This is just a guess, but that's what I'd assumed.

2. Basically, prime95 did a calculation where it expected one answer and got another. This is an error by the cpu.

Answer to this is pretty obvious, lower the overclock.
 
Thanks

I currently have overclocked it using following

HT Multiplier=4x
Multiplier=10x
HTT=270

What do you reckon i should lower it to

thanks
 
sham said:
Thanks

I currently have overclocked it using following

HT Multiplier=4x
Multiplier=10x
HTT=270

What do you reckon i should lower it to

thanks
Click to expand...
I reckon you should lower it until it stops failing prime95 torture tests. OR RAISE the VCORE. Have you tried raising the voltage to the CPU?
 
no....i dont want to raise the voltage

I have changed the oC a bit

"i have overclocked my 3700+ from 2.2Ghz to 2.65Ghz.

I have used:

HTT=265
Multiplier=10x
HT Multiplier=4x

The question i have is that when i run occt or prime 95 at HTT=270, it is unstable - although in all the tasks I do it is fine.

I understand taht prime95 shows complete stability. I get the orunding error is 0.5 not 0.4 error.

The question I have is if I put the HTT=270, but lower the HT Multiplier to 3x, will that be better or worse.

Why does the Ht Muty x HTT have to be under 1000.

Is it essential


Any ide3as
 
drop the HTT link to 3x.

i was gonna answer the P95 + memtest thing but someone already did.. <_< lol

just bump the voltage a tad... -_- your temps are fine, voltage bump wont hurt anything (unless its already been bumped? ^_-

[edit] the HTT link has to try and be under 1000 just because thats what they built it to run at, ym next question is why did you run it over spec (HTT link) even knowing this? thats just begging for problems.... --__--
 
0mega said:
drop the HTT link to 3x.

i was gonna answer the P95 + memtest thing but someone already did.. <_< lol

just bump the voltage a tad... -_- your temps are fine, voltage bump wont hurt anything (unless its already been bumped? ^_-

[edit] the HTT link has to try and be under 1000 just because thats what they built it to run at, ym next question is why did you run it over spec (HTT link) even knowing this? thats just begging for problems.... --__--
Click to expand...

I have tried the drop the HT link to 3x

so the settings were

htt=270
multiplier=10x
and ht multipier=3x

occt=not stable
prime95=rouning error, 0.4 epected 0.5.

I can get a stable overclock at

htt=265
multiplier=10x
ht multiplier=4x

and an unstable at either
htt=270
multiplier=10x
ht multiplier=4x

AND ALSO UNSTABLE AT

htt=246
multiplier=11x
ht multiplier=4x.

The memclock is at 400Mhz on all f these, and CPU Z shows either

CPU/11, or CPU/10.

based on these results, what should I tweak to obtain a stable overclock without raising voltage,

Thanks
 
you're asking for a lot to get 2.7 out of a 3700 on stock voltage. im impressed you even get into windows like that. that rounding error is typiclly either the max oc or a sign that you need more voltage. my 3700+ does 2.65 at 1.52v, and maxes at 2.75 at 1.62 volts (even 1.8v wont get 2.8ghz stable). keep htt link to 3x, theres no need for it to be that high, just between 600 and 1000. please run the memory divider lower while testing to find the highest cpu overclock. please remember that just because your memory can run fast, does NOT mean your cpu's memory controller can run it. my cpu, for example, can only run to 250mhz stable, even tho my ram is rated for 275 (on my old winchester I could run it to 281mhz). Take the memory out of the equation with a 100 divider, then take the cpu out of the question and see how fast your ram can go. from there, decide on the best tradeoff of cpu mhz and ram frequency.
 
arnemetis said:
you're asking for a lot to get 2.7 out of a 3700 on stock voltage. im impressed you even get into windows like that. that rounding error is typiclly either the max oc or a sign that you need more voltage. my 3700+ does 2.65 at 1.52v, and maxes at 2.75 at 1.62 volts (even 1.8v wont get 2.8ghz stable). keep htt link to 3x, theres no need for it to be that high, just between 600 and 1000. please run the memory divider lower while testing to find the highest cpu overclock. please remember that just because your memory can run fast, does NOT mean your cpu's memory controller can run it. my cpu, for example, can only run to 250mhz stable, even tho my ram is rated for 275 (on my old winchester I could run it to 281mhz). Take the memory out of the equation with a 100 divider, then take the cpu out of the question and see how fast your ram can go. from there, decide on the best tradeoff of cpu mhz and ram frequency.
Click to expand...

My CPU was stable when I put a 1:2 divider on my ram "200Mhz memclock".

The settings were
HTT=270
Ht Mutli=3x
HT Multiplier=10x

I ran OCCT and it was stable

I was told that I need to adjust settings

Here are my me settigns

Min RAS# active time (Tras) 5T
Ras# to Cas# delay (Trcd) 3T
Row Preelayer? time (Trp) 3T
Row Cycle Time (Trc) 16T
Row Refresh Cyc Time (Trfc) H/2T Memory Timing 2T

Hope you can help
Thanks
 
what is your ram rated for timing wise? my g.skill pc4400 is for 2.5-3-3-7. you are showing 5 for your tras, which is too low, try setting that to 9 or 10, which is best for a64's. also, if you say it was stable at 270x10 with 1:2 divider, but not with ram 1:1, then its too much for the cpu to run at a high mhz AND at a high ram speed. you will keep running into this kind of wall until you throw some extra volts at the cpu, try 1.45v at least, thats safe, altho why bother unless you start at 1.5 :p
 
arnemetis said:
what is your ram rated for timing wise? my g.skill pc4400 is for 2.5-3-3-7. you are showing 5 for your tras, which is too low, try setting that to 9 or 10, which is best for a64's. also, if you say it was stable at 270x10 with 1:2 divider, but not with ram 1:1, then its too much for the cpu to run at a high mhz AND at a high ram speed. you will keep running into this kind of wall until you throw some extra volts at the cpu, try 1.45v at least, thats safe, altho why bother unless you start at 1.5 :p
Click to expand...

I have done following

Cas=3
Trcd=3
Trp=3
Tras=8

I have just tested the memclock at 400Mhz.

It is at 2.7Ghz, and the Ddr voltage is at 1.85V

It i stable with memtest and OCCT

Do you think it is possible for me to lower the Ddr voltage any more, or tighten any of the timings.

If I were to go about overclocing more....what would I just in terms of memory.

Thanks

Look forward to your response
 
set the cas to 2.5. the rest looks good. finally, run prime95 for at least 12 hours straight (I prefer 24 hour final burn-ins) as the ONLY program running to make sure it uses 100% of the cpu...if this passes, you indeed have a golden cpu (and could probably hit 3ghz with some voltage). you wont get better than 2.5-3-3-x out of the ram, its already quite good, you are running ddr540 really with it 1:1 at 270. the timings are good for this, consider yourself EXTREMELY lucky. Once you have finalized the overclock with at least 12 hours of prime, take some screenshots and post in the revision E overclock database. I still cannot believe you got 2.7 on stock voltage AND an awesome memory controller (grumble grumble)
 
Will do so and post back.

What voltage should i put the ram on.


The Geil say 1.95V max, and I have it on 1.85v - should I go any lower.

Thanks
 
You must log in or register to reply here.
Back
Top