• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Easy as 1, 2, 3, Overclocking the A64, Opteron, X2

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
you can but it leaves the posibility for system instablity open. By testing each componant independently of the others, you reduce the chance of having any one componant hold your system back. It also makes finding the weak componant in you system much easier.
 
I know I should know how to OC by now, but I didn't know much about dual-core. I was trying to run stress tests as if it were a Hyperthreading P4, and it kept freezing, lol. I realized that it wasn't working out when I ran two instances of SuperPi at once, and also when I ran Prime95 two instances (one small FFT on affinity 0, and one large FFT on affinity 1). It wasn't working out, so I read your initial post, infinitevalence, about the S&M running in background.

Now, here is are my questions about my Opteron 170:
  1. How do I choose affinity with SuperPi and/or how should I be testing it?
  2. In SP2004, what is the checkbox next to "No Affinity" for, considering that you can already choose "0" or "1"?
  3. Can I run CPU Burn in as the background instead of S&M (cuz I don't know how nkow S&M)?
  4. Any other pointers about dual-core testing as opposed to single-core?
Thanks. :)
 
Well the way i have been doing it has been using clockgen and two instances of prime. I set both instances up to small FFT and start them going with the windows open. I then use clockgen to raise 5mhz at a time, let it run through a full itteration of numbers (4-5 lines you will see what i mean) then bump it again, when one finaly fails i back it off 5 mhz and then prime for 12-24hrs depending on the amout i think it needs or reboot and bump the voltage.
 
Dude!! I was talking to you on AIM at the same time you posted this, and you didn't even tell me you posted here, lol.

Well, I figured out the affinities last night for SuperPi and taskmanager. I then figured out the affinities with SP2004.

I guess I'll start using clockgen also... I'm still old school BIOS boy...

I've tried A64 tweaker as well with my ram, but it's confusing since the #s are always off from what it says in bios :(.

Okay, so I want to develop this windows-based testing instead of rebooting over and over to test. I will use clockgen for the cpu, but what do you recommend I do with the ram? I want to use A64 tweaker, but don't know what prog to test the ram with for that doesn't take forever everytime...
 
What priority level should I use in SP2004 running two instances?

I'm trying to run one small fft and one large fft on separate affinities, and with priority 10, the system is non respondant to mouse.

First, what priority should I set each to? Second, should I run large large fft, small small fft, or large small fft and then reverse it?

I think I should lower priority cuz I'm running smart guardian, and cpu-z on the desktop as well... priority 5? :/
 
Ram: for the ram i still do it old school and find its max with stable settings. I use memtest to find where it tops out or as high as i think i need to go. Then i set it to the settings i used, and drop the frequency so i know im never going to hit it when testing the cpu.

CPU: When i test the cpu in windows for my quick testing i do not set priority 10 because then i could never change the fsb in clockgen. I set the voltage i want to test at in the bios, then like i said let prime run for a few ittereations then bump it up 5mhz. When it finaly craps out i back it off 5 mhz and let it run for 12 hrs. If its still stable after that, then i reboot into the bios set the memory divider that i plan on using still at 200mhz for the fsb, and boot to windows. Then i again test little by little, generaly i will start somewhere near the center of both overclocks. So if my memory does 260 and my cpu also does 260, its safe to start at 230. bump in 5mhz increments until it fails and back it off, then prime for 24hrs :) then 3dmark for 12 just to make sure its not going to fail in gaming.

Prime: When it comes to prime i would set it the same so both cores are stressed equaly. Its personal preference as to using large FFT or small FFT, i belive that small FFT stressed your cpu more and is better for this type of overclocking. Other feel differently, but i do not recomend blend. Also any priority that is high, but not 10 should do it, so anything 5+ is good.
 
g0dM@n said:
What priority level should I use in SP2004 running two instances?

I'm trying to run one small fft and one large fft on separate affinities, and with priority 10, the system is non respondant to mouse.

First, what priority should I set each to? Second, should I run large large fft, small small fft, or large small fft and then reverse it?

I think I should lower priority cuz I'm running smart guardian, and cpu-z on the desktop as well... priority 5? :/

Priority 10 is the best way to Prime for complete stability IMO.

Use small FFT to test your cores independently or at the same time. I prefer the same time since it does you absolutely no good to know if one clocks higher than the other. Then use large FFT or preferably Blend or make a run at both to test overall system stability. The reason I like Blend is b/c if set properly it will cause your system to page b/c of the high memory usage. So not only is it a great memory test but a great virtual memory test as well... again, overall system stability test.

Personally, I've never tried mixing two different tests small w/large vice-versa or any other combination. Please let us know if you find any value in testing that way.

I don't use Smart Guardian so I don't know, but is there any way to reduce the polling time? ...to like once every minute or 90 seconds? It would be a more effective test if SmartGuard was not stealing as many clocks from SP2004.

infinitevalence said:
Prime: When it comes to prime i would set it the same so both cores are stressed equaly. Its personal preference as to using large FFT or small FFT, i belive that small FFT stressed your cpu more and is better for this type of overclocking. Other feel differently, but i do not recomend blend. Also any priority that is high, but not 10 should do it, so anything 5+ is good.

Why not Blend and why not Priority 10? I'm just curious as to your experiences... I've used both the test and priority on my 170 with much success.
 
Hello Overclockers! (new to the forums)

I am having some problems with overclocking my system. I'm not sure this is the appropriate place to post but oh well.

Specs:

AMD Athlon 64 3200+
ASUS A8N5X Motherboard
1024 MB Corsair Value RAM <- I'm sure this is not the best for overclocking
GeForce 6800 GS GPU
460W SOLYTECH PSU

After trying to overclock on my own with some beginners guides I found, I have been getting a lot of instability with very low overclocks. I try to set the RAM low enough so that it is not being overclocked. I start by increasing the FSB(?) by 5MHz at at time with the multiplier at 10x untill it becomes unstable (220 FSB if i remember correctly). After this I try raising the cpu voltage a bit but it is still unstable. This is where I gave up on my own.

I then tried following this guide. I tried the first step, testing the limits of my HTT. I only got it up to 225 with the ram at 1:2 (half speed), my "Hyper Transport Frequency" (this is what my mobo calls it, this is the HTT multiplier correct?) to 5x, and my cpu multiplier set to 6x. It seems like this is pretty low.

I guess I'm asking, what am I doing wrong?

Edit: I do have it successfully oc'ed with the mobo's built in "AI Overclock" feature to 2160 MHz with standard voltage and RAM divider.
 
Last edited:
the HTT generaly goes from 1000mhz down or 200x5 so you want to test both with the high HTT like 5x and with a low HTT like 3x or 2x this will give you much more headroom for your overclock.
 
I'm kinda stuck on step 1.
when I set the CPU multi to 6x and the HT to 5x I get a HHT of 230x5=1150 which I guess is normal since you dont want to go over 100. But when I try to max out the FSB with CPU multi at 6x and HT at 3x I can only get to 230 again which in case im wrong is realy low. My mem is always set to 133 when testing this. Is there something else I need to set on my MB to allow the FSB to go higher or is my MB junk? Realy hoping that im overlooking on small thing in the BIOS that is keeping me from >230 FSB.


A64 3000+ DFI NF4 Lanparty ultra-D
520w ocz powerstream
1g value ram atm
74g raptor
eVGA 7800gt
 
Last edited:
you can try giving some more voltage to the LTD bus and the chipset see if that helps any, you should have no problem hitting 300fsb with an ultra-d
 
During this first step:
Begin your testing by raising the FSB with the FSB/HTT multiplier set to 5x. Push your HTT up in 5mhz increments...
I have the option of CPU, PCIe Sync or CPU, PCIe Async.

Right now I'm running an AGP card, not PCIe, so I figure it doesn't matter what the PCIe is set to, right? I'm in CPU, PCIe Sync and at 240 (CPU), 120 (PCIe) now, should I change it to Async and keep the PCIe at 100?
 
Last edited:
Hello!

Sorry if this is in the wrong place to post.

My setup:
3000+ Winchester
DFI Lanparty nf4 Ultra d
Corsair Twinx4400 2*512MB

After reading this post I thought it was time for me to properly OC my 3000+ Winchester. I followed the three steps to obtain the max for my Memory (285), FSB (295+, no extra volts), CPU (2430), adjusting volts to obtain the extra Mhz. I was very happy with these individual results which I tested on a individual basis using memtest86+, prime95(12 hours), superPI, 3dmark2001, 2005.

Now I knew what my hardware could do I wanted to test it all together. This is where it all went wrong. I was aiming for:

CPU 2338 (8.5 * 275)
Mem 275 (1:1)
FSB 275

Memtest86+ failed, Prime95 failed etc.

So I'm wondering what the problem is. I know individually the components can handle it but as a system is it expected that I should get issues?
 
That is not unusual, your going to need to back off your OC some where, most likly its your memory that your going to need to back down, try decreasing your memory overclock or your processor overclock by just a bit and then prime and memtest.
 
I calmed down and did as you said.

CPU * 9 @ 2160 (1.45V)
Memory 240 1:1, memory tighten up from settings from individual test (2.9V)
FSB 240 (chipset 1.5V)

Memtest passed
OS booted
3dMark2001 and 2005 passed
SuperPi 32M passed
Prime (1 Hour passed)

I then increased the FSB by 5Mhz and ran the above tests again.

I'm at:

CPU * 9 @ 2385 (1.62V)
Memory 265 1:1, memory tighten up from settings from individual test (2.9V)
FSB 265 (chipset 1.5V)

Memtest passed
OS booted
3dMark2001 and 2005 passed
SuperPi 32M passed
Prime (5 Hours passed so far....I will leave for 12+)

I will be as happy as Larry if this works, until I decide to push it a bit further :) My Arctic Freezer 64 Pro arrived this morning. I've got resist putting it on until I get this stable and then see the difference it makes.

Thanks for your help. I didn't realise I had to back off so far.
 
its honestly very hard to get any simpler than this... Overclocking is an advanced from of computer usage. If you dont feel comfortable using these instructions then you may want to wait a while and do more reading on what overclocking is about and how to do it before atempting it.
 
Back