Athlon 64 3200+ & Asus A8v

[admilitaryman]

Here is the build i'm working with:

Mobo-Asus A8v socket 939 rev. 2.0
CPU-AMD Athlon 64 3200+ Venice
Ram-2x 1gb Kingston Value Ram 400mhz
HD-(7200 rpm IDE 80gb.. I think its a WD )
Case- CAF 932
Heatsink- Scythe Mugen 2
Vid Card-GeForce 6200A LE
PSU- Corsair 850TX

This is an old PC i've added a few parts of my i7 system to. My i7's in the RMA process atm.

Bios settings(0229):

3-3-3-8
Memclock Mode: Limit 1:1 DDR200
1T, 4 beats
FSB: 265
CPU Mult: 10
AGP/PCI Freq: 66.66/33.33
DDR Volt: 2.6v
AGP Volt: 1.5v
CPU Voltage: 1.7v
HT: 600 (3xLDT)
I can list more bios settings if needed.

Temps are 28°C Idle and 40°C Load.

1. The biggest problem i'm having right now is restarting the PC. For some reason, when i passed 260fsb, the PC will not reboot after restarting from Windows or Changin/Saving bios settings..I have to turn the PC off completely then turn it back on, and works fine. I've read it could be because of the AGP/PCI lock. I was hoping someone could help me determine if the higher AGP/PCI lock would be safe with the vid card I have? Has anyone seen this issue before?

2. I had another question about the mem controller. I've had my FSB up to 300 with a low multi and my ram OCs to around 220mhz but for some reason i cant hit higher clock speeds. I'm thinking the memory controller is the problem, are there any settings that would increase this?

3. Does anyone know of any VERY EASY vcore mods i could do to raise past my max bios setting of 1.7v? I'm running very cool for this setting (1.74v cpu-z)?

 

[admilitaryman]

Here's someting interesting..OCCT and Prime95 report errors within seconds of stress testing but I can run core damage for quite some time with no problems. This confuses me because i've trusted core dmg for my i7 as it gives the highest temps out of all the stress testing software.


The restart problem started @ 260x10 @ 1.7vcore. I'm at 270x10, 1.7vcore running core dmg just fine. No freezes or reboots yet.

 

[QuietIce]

220 MHz is where I would expect value RAM to start fading out so there's nothing wrong there. Normally I would suggest loosening the timings but with CAS 3 at 200 MHz you're already as far as you can go there.

The only option left is to turn down the RAM speed. With a 265 MHz clock and your RAM set for 5:6 (DDR-166) it'll be running 220 MHz already so I'd suggest dropping the RAM back to 2:3 (DDR-133) for the time being. That should give you plenty of RAM speed overhead with a 300 MHz clock bringing RAM speed back to 200 MHz. When you're done with your CPU OC you can go back and experiment with the RAM but, ultimately, CPU speed is King with RAM a distant second for performance ...

 

[admilitaryman]

well the thing is that while ocing i set the ram to the low 1:1 ratio where its running @ 100mhz..after i find the stable clock speed i then raise the ram to the highest i can get it.

 

[QuietIce]

That would be a 1:2 ratio if the RAM is running at 100 MHz with a 200 MHz clock, but OK, with that cleared up ...

1. If you're talking about not being able to use the reset button (having to shut down completely) but everything else being fine - many s939 boards are like that. I don't think anybody has been able to figure a way around it. Some boards do it and some boards don't.


2. If you're talking about higher RAM speeds then, like I said earlier, you're probably at your limit for value RAM. My Corsair XMS will run up to 225-7 MHz at stock timings (2-3-2-6), ~248 MHz if I loosed the timings a little (2.5-3-2-6), and I don't know how far but at least to 265 MHz if I loosed the timings another notch (3-3-2-6). On the other hand, the Corsair value RAM I had peaked at 221-2 MHz and, since it started at CAS 3, that was as high as I could go.

If you're talking about increasing the clock speed the board can run the only thing you can try there is adding a voltage to the NB - that might buy you another 10 MHz on the clock.


3. No clue but 1.7v for an s939 is pretty high already - just my opinion ...
http://www.overclockers.com/forums/showpost.php?p=2987106&postcount=4
(scroll down for Venice)

 

[admilitaryman]

First off, i really really appreciate the help, i've had a terrible time getting someone to respond to posts on this. I'm an electrician and not understanding stuff like this agitates me.

Secondly..yeah the ram isnt the issue because i'm running it at 2:1 while i OC the clock. Like i might have said, i've lowerd the FSB multi to 4x and reached a 300FSB which is the max on this board. I automatically assumed that mean i should be able to hit 3ghz because all i'm doing now is increasing the multiplier and vcore...i guess i was wrong to assume. My theory isnt totally wrong, though, because while using the FSB @ 300 to OC and raising the multi i can only hit 2.4ghz (300x 8) however I am able to hit 2.7-2.8ghz when using a lower FSB and the default (10x) multi. While ocing the two different ways (raising fsb vs raising multi) i should achieve around the same total clock speed, shouldnt i?

The restart issue is weird, when overclocking past 2.6ghz i have to turn the pc totally off then back on for it to post. If i change a bios setting and save and restart it will not post, if i restart from windows it will not post, if i use the restart button it will not post...I'm wondering if its a amperage/voltage thing.

There is so much to overclocking its impossible not to ramble and impossibly to type everything i've done. Hopefully we can figure this out b/c its annoying me..I appologize for rambling but i just cant figure out what my total core clock max is and y prime95 and occt fail at 2.6ghz+ but i dont have any freezes, reboots, bsods and core damage will run fine.

 

[QuietIce]

The restart issue is weird, when overclocking past 2.6ghz i have to turn the pc totally off then back on for it to post. If i change a bios setting and save and restart it will not post, if i restart from windows it will not post, if i use the restart button it will not post...I'm wondering if its a amperage/voltage thing.

I'll address this one first because it's the easiest. I don't know what causes that but some systems do it - including my sig rig - some don't. I've got four of the A8N32's running right now. My sig rig and one other machine does this. Of those four, TWO are running a higher clock and CPU speed than my sig rig - in fact those two are almost exactly alike in all their settings (one at 9x322 one at 9x323) but only one won't reboot, the other works fine. Go figure! I sometimes think it has more to do with the PSU than the board. Both the rigs that have this issue are running newer PSU designs, the two that don't have "old-style" PSUs.

Secondly..yeah the ram isnt the issue because i'm running it at 2:1 while i OC the clock. Like i might have said, i've lowerd the FSB multi to 4x and reached a 300FSB which is the max on this board. I automatically assumed that mean i should be able to hit 3ghz because all i'm doing now is increasing the multiplier and vcore...i guess i was wrong to assume. My theory isnt totally wrong, though, because while using the FSB @ 300 to OC and raising the multi i can only hit 2.4ghz (300x 8) however I am able to hit 2.7-2.8ghz when using a lower FSB and the default (10x) multi. While ocing the two different ways (raising fsb vs raising multi) i should achieve around the same total clock speed, shouldnt i?

I'm not sure what you're talking about when you say "FSB multi to 4x". Are you refering to the HT Link speed? If that's the case then it should be set to 3X to avoid exceeding 1000 MHz. Some boards can run 1200 MHz but I never count on it.


Top end CPU speed is very CPU dependent, as that OC'ing results list shows, but you should achieve about the same CPU top end either way you go if nothing else is getting in the way of the overclock. Assuming the CPU will run 2700 MHz you should be able to use both of these settings:
10x270
DDR-133 (= 180 MHz)
HT Link at 3X (= 810, probably even 4X = 1080)

9x300
DDR-133 (~193 MHz )
HT Link at 3X (= 900 but probably not 4X = 1200)


If nothing else it could be that your board will just barely run 300 MHz on the clock and the extra voltage draw from the faster CPU decreases that by a small amount. That would account for your slightly lower OC when trying to use the faster clock speed ...

 

[admilitaryman]

When i say "FSB multi to 4x" i mean the clock multiplier. I had this big post typed out about how i hit 2.6ghz blah blah, then i BSOD'd.. i'll try to retype what i stated. Basically i got to 2.64ghz with ram at 5:3 (219mhz) @ 1.7vcore before the BSOD. I just dont understand how i'm now at 255x10; ram at 5:3 (212mhz) HTT is 1020 and vcore is on auto @ 1.45-1.475v but i cant get above 2.6ghz core speed. Oh, when i said i hit 300 FSB/clock speed i did that @ 1.45v..so i dont understand how i cant stay stable at a measly 2.6ghz with 1.7v vcore. I really dont understand how i cant hit 2.7-2.9ghz clock speed with this board, especially if my clock can run at 300x4 @ the stock voltage (1.2ghz clock speed).

I'm an electrician and this voltage stuff annoys me when i cant figure it out. :s

*edit* it seems like the voltage increase does nothing to help the OC.

 

[col_sanders]

I've got a 3000+ with 2x1GB G.Skill pc3200 memory and the highest i've ever got the clock was 2.5 semi-stable and 2.45 GHz rock solid. Although mine is the Newcastle core. Although I've seen at PcPerspective they did a review where they were able to get their 3200+ Venice to 2.8GHz (http://www.pcper.com/article.php?aid=127&type=expert)
though they bumped up their NB frequency and had their memory running at 280 (560 rated)

 

[QuietIce]

When i say "FSB multi to 4x" i mean the clock multiplier. I had this big post typed out about how i hit 2.6ghz blah blah, then i BSOD'd.. i'll try to retype what i stated. Basically i got to 2.64ghz with ram at 5:3 (219mhz) @ 1.7vcore before the BSOD. I just dont understand how i'm now at 255x10; ram at 5:3 (212mhz) HTT is 1020 and vcore is on auto @ 1.45-1.475v

OK, so you're calling the HT Link multiplier the "FSB"? OK. The "clock" to me is the reference clock. I've been around computers a long time, "clock" always means "reference clock" unless otherwise noted. (Although, strictly speaking, the reference clock isn't what we adjust.)

reference clock (aka "clock", aka HTT): default 200 MHz
CPU multiplier: max 10X
HT Link multiplier: default 5X
RAM ratio default 1:1 (= 200 MHz)
(The "RAM ratio" in BIOS is really just a variable used to calculate the RAM divider, which is always a whole number. I'll get to that later if we have to go that far. )

So you're at (using the order above):
clock: 255
CPU: 10X (= 2550)
HT Link: 4X (1020)
RAM: 5:3 (this is what I would call 5:6 = 166.67 @ 200 MHz clock)


From there I would turn the RAM down to 2:3 (= 133.33 @ 200 MHz clock - I don't know what your BIOS will call this) and, this is important, set the stock (200 MHz) RAM timings manually if you haven't already done so. Also, lower the HT Link down to 3X. For now leave the CPU at 10X and the clock at 255 then reboot and check your RAM timings and speeds - you should be at 170 MHz on the RAM and 765 MHz for the HT Link.

If those speeds are all correct and your timings are right, then reboot and set the vCore at 1.45v and raise the clock to 260 MHz. Test with Prime95 and if all is good reboot and raise the clock another 5 MHz. Repeat until failure then increase the vCore to 1.50v and go back to increasing your clock by 5 MHz.

You can continue doing that as long as the CPU and your load core temp holds out. Keep in mind as you raise the vCore your CPU temp increases and the it's potential top end (highest overclock) will decrease. It's a vicious circle and once you're caught in it you're at what we call the Heat Wall. That's why you have to kind of sneak up on the overclock one voltage step at a time and why, sometimes, less vCore is better. Maybe that's what you're running into ...

 

[admilitaryman]

I have been doing the steps you've stated but i didnt know about the "heat wall"..this would explain y i can reach a nice OC at 1.45v @ 2550ghz but i cant reach 2.6ghz @ 1.7v. The "heat wall" would also explain how i was getting varied results with my overclocking. When i first started OCing this rig 2 days ago, i started with a low vcore and raised the clock until i had to add more voltage, like you stated above. However yesterday i decided to try a different approach and set the voltage to the max i was going to use (1.7v vcore) and just started raising the clock until failure. Tonight i will re-do my overclocking and post any problems i have. Oh, on the restart issue, it seems the reason y my pc wont restart after changing bios settings is because its actually the bios resetting itself, AKA i've pushed it too far. I dont ever get the message "Overclocking failed" like i've seen on other posts but my rig just wont restart. When i power down then back up, my clock speed is at default until i enter bios again. <--This is hard to explain, i hope you can understand.

 

[QuietIce]

As an electrician you should understand the heat wall as good or better then any of us. Shove too much current down a line that's not built for it and the line heats up, which increases it's resistance, which decreases the power it's delivering, which requires more power for the same amount delivered, which heats the line more ...........


Not hard to understand at all - just wish you had caught it sooner.

Post back when you have some new results - or if you hit another snag ...

 

[admilitaryman]

I'm on lunch but i'm running prime95 while i'm at 255 clock, cpu 10x, ht link 4x(1020) and ram at 5:3..your 5:6. I'm still testing to find 100% stability at this. Vcore is 1.5v atm. I'll try pushing it further later tonight.

 

[admilitaryman]

From there I would turn the RAM down to 2:3 (= 133.33 @ 200 MHz clock - I don't know what your BIOS will call this)

My mobo calls this 4:3..its (clock x A) / ( B x2) to figure out the ram speed. Unfortunately, i bumped the core to 255 and failed to POST, instead of just increasing the vcore, considering my ram was at 212.5mhz, i decided to try the 4:3 ram divider but it didnt post, so i tried 1:1 and now although im running at a low ram speed, i'm at 255 and testing. Will keep you updated when i have more time tonight.

 

[admilitaryman]

1.5v vcore failed on Prime95...Btw thanks for the link col sanders!

Heres an update..i got the 4:3 memory divider to work by adjusting my timings although i still had to increase vcore to 1.55v. I'm still trying to find a stable OC. I'm still at:
clock: 255
CPU: 10X (= 2550)
HT Link: 4X (1020)
RAM: 4:3

 

[Catscratch]

As you know, no chip is the same and they may not take 1 OC that others achieved. I had 3200+ Orleans chip and it could take 500mhz out of the box at default voltage and to reach 2.6ghz it needed like just above the voltage limit of AMD and I never reached 2.7. Yes it was an AM2 rig but what i'm trying to say, you may already found your chip's limit.

 

[admilitaryman]

Seems like it, I got 255 clock to work at the above speeds then tried 260 and I booted at:

clock: 260
CPU: 10X (= 2600)
HT Link: 4X (1040)
RAM: 4:3 (173mhz)

I ran prime95 which caught errors and then i restarted. For some reason the restart didnt work and when i tried to reboot again it wouldnt POST. I lowered to 255 with the above settings and guess what? NO POST..i dont understand it, i ran prime95 for 8 hrs @ 255 but now it wont work. Any ideas of whats going on? Did i just reach the max OC? I dont understand how i can up the vcore to 1.7v with great temps but i cant get a better OC than 1.55v vcore.

Althought i reached 300 clock at a low CPU multiplier (4x) it seems like my best oc may well be:

clock: 250
CPU: 10X (= 2500)
HT Link: 4X (1000)
RAM: 5:3 (208mhz)
Vcore: 1.55ghz

 

[QuietIce]

It's possible you've reached the limit and it may not all be in the CPU. The motherboard and it's power circuitry have a lot to do with that as well. Not being a fan of VIA chipsets I don't know much about your board I'm just saying it can play a roll in OC'ing as well as the CPU and RAM choices ...

 

[admilitaryman]

I think i'll stick with 2.4ghz as i dont have to raise any vcore to reach it, i BSOD'd on 2.5ghz today and have no clue why. Now my last question then this thread is finished..Would it be worth buying the best agp card available instead of having to replace the entire machine? Keep in mind its my wife's pc, she only plays wow or aion and i know my i7 is quad core but 2.4-2.5ghz single core isnt too much difference is it? I'm not sure on how many multi threaded apps there are out now.

 

[QuietIce]

I'll have to let one of the old-time gamers answer that question. I mean, a single is plenty to run Office, browsing, e-mail, and all that but when it comes to gaming I just don't know. I do know there are a LOT of multi-threaded games out there now. Whether your wife's games are part of those I have no idea and a younger gamer will tell you your machine is junk because they don't have a clue about the performance of the older machines, which is why I said you'd need an old-time gamer to tell you.

You might try posting in the gaming section and you should probably be careful of people who have no clue what AGP is ...