A64 OC Database & Overclock Guide

cornelious0_0 said:
200x9 - 1,800MHz - 0.992v
300x9 - 2,700MHz - 1.360v
305x9 - 2,745MHz - 1.392v
310x9 - 2,790MHz - 1.440v
315x9 - 2,835MHz - 1.440v
320x9 - 2,880MHz - 1.488v
325x9 - 2,925MHz - 1.536v
329x9 - 2,961MHz - 1.600v
Click to expand...
Looks very nice compared to my 3000+, which with 15C load temps still took around 1.6v to hit 2.8, and then topped out at 2.908 with 1.73v and load temps of 28C. I'd figure the lower temps would help out some, but this thing just stopped scaling at 2.8 without pretty decent vcore jumps.

I haven't heard of a burn in being mentioned in a while. What exactly constitutes a burn in to you?
 
I haven't heard of a burn in being mentioned in a while. What exactly constitutes a burn in to you?
Click to expand...

It's honestly not something that a lot (of non-hardc0re) people actually "believe in" but burning in these Venice/SD chips has seemed to help a fair bit. Some people will say, find the lowest voltage at your overclock that the chip is stable, and run CPU Burn-In for like 48 hours for as long as it takes. What I ended up doing was I started at 205x9 stock voltage, and started bumpin 'er up..raising the voltage as needed. Each jump of 5htt I would run SuperPi 32M (about a 30 minute test) before moving on. By the time I got up around 300x9 I had been running it over and over enough that one day I had left the computer sitting idle at its current 300x9 overclock, and when I came home, it had locked up. The only way I could get it stable was actually to lower the votlage...it had gotten to that point where it had "burnt in" and no longer needed that much extra voltage to remain stable. A lot of people have tried to tell me that this was just due to the AS5 setting in, but this was a good week after i had installed it...and I had NEVER seen a night and day difference like that after compound sets in. Some people indirectly gain a little on their OC, but at the very least you may be able to maintain your current overclock, but simply not require as much voltage...which is still a win/win situation.

First I'm going back through and running SuperPi 1M at each 5htt jump between 200 and 330 to get a nice database of what voltages it needed for each jump...just to reference to. After that's finished I'm going to go back to 200 at the 1M stable voltage and work at getting things 32M stable..raising the htt by 5 at a time again, and raising the voltage when needed.

All in all I'm excited to get started on teh 32M testing, because so far I'm up to 250x9. This chip is 1M stable at 200x9 at 0.992v...and so far it's fine at 245x9 at 1.136v...still doing pretty good.

I'm not gonna post up any more results of these sorts in this thread, I'll just wait, and I'll be starting a thread called "3000+ Venice...second time's a charm" in the OC and cooling forum...so stay tuned....I wanna finish the 1M testing tonight, and then I'll be starting the thread with the results I have..and will continue to update that thread.
 
out of curiosity, how do you think temperature might effect the burn in process? I might give this a go with my current chip, but I'm wondering which would be better (if it matters) a low temperature with the chiller (I can get as low as 9C load with 1.8 and low vcore), or a normal temperature without the chiller? I'm definately going to have to give this a try, I figure it couldn't hurt.
 
mikelz85 said:
out of curiosity, how do you think temperature might effect the burn in process? I might give this a go with my current chip, but I'm wondering which would be better (if it matters) a low temperature with the chiller (I can get as low as 9C load with 1.8 and low vcore), or a normal temperature without the chiller? I'm definately going to have to give this a try, I figure it couldn't hurt.
Click to expand...

From what I've seen it doesn't really matter a whole lot, essentially what you're doing is stressing the CPU to run at full load it a clock that it's JUST stable at...breaking it in at that clock...I have never seen temps play a detremental role in the process.
 
Eclipse, going to expand this thread now that Manchesters are shipping? I am not sure anyone has them yet, but I woulnd't argue with some OC results once people start getting them. I wont drop $850 on a processor anytime soon, but I knwo pleanty of people have been holding out for Dual Core.
 
Viper87227 said:
Eclipse, going to expand this thread now that Manchesters are shipping? I am not sure anyone has them yet, but I woulnd't argue with some OC results once people start getting them. I wont drop $850 on a processor anytime soon, but I knwo pleanty of people have been holding out for Dual Core.
Click to expand...
indeed i will! (go check the first post again ;))

and corn, i'm still waiting for the max prime95 stable :D
 
(cf)Eclipse said:
and corn, i'm still waiting for the max prime95 stable :D
Click to expand...

I have finished my SuperPi 1M testing...found the voltages needed for every clock from 200x9 up to 329x9...and now I'm gonna be stepping back to 200x9 and working my way up, finding my required voltages for every 5htt increment up to 329x9 again...for 32M this time.

After that's done, I'll take my max 32M stable clock and start fine tuning it for your list...so it'll be awhile, but work is in progress.

:cool:
 
wow, talk about thorough... i didn't even go into that much detail with my clawhammer when i got bored :p
 
(cf)Eclipse said:
wow, talk about thorough... i didn't even go into that much detail with my clawhammer when i got bored :p
Click to expand...

Well, I finished the 1M testing....but I decided to start at 250x9 and just start with 1.400v (in BIOS) for the 32M testing. I'm going to wait until I have my 32M stable overclock maxed out before I post the thread, and I'll stop at 300x9 to see if I can get it 32M stable at anything lower then 1.360v...but I've still got a fair bit of work to do....I'm still going from 250x9 up to 330x9 in increments of 5...
 
ADA3000DAA4BP
LBBLE 0517APJW

AMD64 3000+ VENICE
STOCK AMD HS/FAN
MSI NEO2 PLATINUM
2X512 CORSAIR 3200 XMS
WD 250Gb SATA HD on Channel 3
Everything else is irrevelant @ the moment

Here's 2.491 @ stock volts and cooling, running a 166 divider, ram @ 1T 2-3-3-6 226MHz HTT x 3


This is the highest I got until I figured out my mobo doesn't like anything less than a 7x multi. Stable & Folding 24/7 for 7 days. I might go back and find higher stock settings once I get final OC.

Here's the current set-up, Prime 95 stable for about 8 hrs. I wanted to at least hit 2.6 GHz, so this is where I'm at for the moment. Ram @ 1T 2-3-3-6 236Mhz 166 divider




On a side note...I had it stable @ 2727, 133 divider, mem @ 192 MHz, but the benchmarks were slower than the 2491 mhz , due to the slower memory speeds. I'll probably throw on a bigger 80mm fan until I decide on what type of cooling I want to get.

This is a work in progress , so stay tuned.
 
Sorry for the offtopic Eclipse, but SpoogeMonkey .. if you want me to look at your screens for overclocking, don't show your wallpaper next time because that's all I saw. :D
 
StarWolf said:
Sorry for the offtopic Eclipse, but SpoogeMonkey .. if you want me to look at your screens for overclocking, don't show your wallpaper next time because that's all I saw. :D
Click to expand...
agreed :eek:

and i would update this now, but the data is all on a hard drive that's attached to a computer that is currently under testing for various purposes ;)
 
Just an update:

2.628 Ghz (292 x 9) 1.425 vcore, ram 239mhz 1t 2-3-3-6 2.7vdimm 166 divider



24 hrs prime stable, still on stock cooling, going to let it fold for a few days until my scythe ninja gets here, then I'll press on. Stock hsf is screaming loud, I can hear it on the other end of the house when I go to bed :(

(And no distracting wallpaper this time) :D
 
cornelious0_0 said:
find the lowest voltage at your overclock that the chip is stable, and run CPU Burn-In for like 48 hours for as long as it takes.
Click to expand...
Wouldn't this just be your stock voltage, I would imagine people start from there and work up?
 
mikelz85 said:
he's talking about a "burn in" if you read what he did in his earlier posts.
Click to expand...
yes exactly isn't that what you do?
 
I haven't tried it yet (well I ran 8 hours at varying settings), but the theory here is to start at the lowest vcore, running the chip as fast as it will go (and remain stable) under a stress program such a super pi or prime95 for 6 hours or so, then after a while bumping up the frequency more (but not much), and then voltage as needed, in the hopes of gaining more headroom.
 
booted up my venice 3200 at 2.7ghz strait up, had 5 mins before i had to sleep. thats it so far, so busy! :(
 
My First OC, Thanks to Eclipse for that guide, or this wouldn't be possible My OC

ADA3000DAA4BP
LBBLE0519EPDW

Cooling: Air
Voltage: 1.4

EDIT: New Photo
 
Glow said:
yes exactly isn't that what you do?
Click to expand...

Starting at stock frequencies, lower your voltage bit by bit, and run Prime95 for 6-12 hours each time.....keep lowering it until the cpu is no longer stable. Then, start to raise the HTT by about 5 at a time, (as slowly as possible) raising the vcore as needed.....

For instance.....my 3000+ Venice will complete SuperPi 1M at 0.992v, but I'm starting at stock voltage and lowering it until it's no longer stable...because it'll most likely need more then 1v to remain stable for a full 6+ hours. The idea behind starting with the ridiculously low voltages is that you're stressing the chip and getting it "used to" running at these lower voltages.....with the intent that once you've been stressing it for long enough, it'll be broken in at the lower voltages, and not simple stock 1.4v.

I just started this, wanted to run 200x9 at stock voltage just for a technicality, but I'm gonna drop from 1.4v to 1.3v now and re-run Prime95 overnight.....then drop it from 1.3v to 1.2v and so on...until I find the 0.025v increment where it'll remain stable for 6-10 hours.

Yes this does take a lot of time, but if you truely want to push your cpu to its max, at the lowest voltage possible...this is the best method I've come up with.
 
well, I'm doing a shorthand version of the burn in for now, letting it run for about 30min and then bumping up the frequency until failure, then vcore, repeat. I started with longer cycles around 1.2v, right now I'm at 2.7 with 1.41v or so, 2.733 resulted in a failure, so I'll let it sit here for an hour or two, then proceed upward. Hopefully this will end up well, but past 2.5 or 2.6 the voltage stops scaling the CPU as well as the other increases seemed to, though that could be a function of what the vcore is doing to my temps as well. Right around 11C load now, hopefully letting it run @ 2.7 1.41 for a while helps it some more. Though I am tempted just to go for broke right now, patience... Then hopefully I'll have a 3ghz update, or at least a vcore that is 1/20th to 1/10th of a volt lower @ 2.9

Also I have a question. I have a DFI nf4ut, my bios vcore monitor reads about .25 lower than what I set it to, and in windows it reads a good .35-.45 lower from sandra (ntune doesn't read it correctly at all). Which reading should I believe? Should I get smart guardian?
 
mikelz85 said:
Also I have a question. I have a DFI nf4ut, my bios vcore monitor reads about .25 lower than what I set it to, and in windows it reads a good .35-.45 lower from sandra (ntune doesn't read it correctly at all). Which reading should I believe? Should I get smart guardian?
Click to expand...

Don't believe any of them just use them as very rough estimates. I just pick one and stay with it they all read from the same sensors anyway. Every reading I've ever tested reads lower than my DMM some as much as 0 .25V
 
can you point out the contacts in a enlarged pic to touch the leads to with a DMM? Just curious to try it.
 
Autochthon said:
Don't believe any of them just use them as very rough estimates. I just pick one and stay with it they all read from the same sensors anyway. Every reading I've ever tested reads lower than my DMM some as much as 0 .25V
Click to expand...

Just what I was gonna say, I go by what I see from CPU-Z in windows and that's it. You're gonna get different results from a few different sources...so pick one and stick with it.
 
mikelz85 said:
can you point out the contacts in a enlarged pic to touch the leads to with a DMM? Just curious to try it.
Click to expand...

I've only tried probing the 12V, 5V and 3.3V lines as they are easily accessible at various places. I use the 24 pin header generally. The memory can be probed but it can be awkward to get to that area with large HS and fans in the way. Cpu can be checked from the back of the board but I'm uncertain exactly where. I'm certain there are people here who know exactly where to probe for both, hopefully someone will chime in. If not next time I have everything torn apart I'll see if I can track down some easily accessible probe points
 
crap, I forgot that the contacts are probably covered with dow corning silicone anyways.
Took the chiller down for the night, still couldn't seem to get 2.8 with 1.55 set in the bios (10min prime error on small fft), sandra reading 1.51 bios reading 1.525. Hopefully I'll get to 3 ghz, but I'm getting close to 1.7v, which I'm guessing is going to be close to the cutoff point for super diminishing returns this time around (unlike last time @ 1.8v)
 
Update on mine,

Running:

HT x 3
FSB x 285
CPU x 9.5
Timings 2.5-3-3-11
Core voltage 1.62
Mem voltage 2.85
 
SGTFury said:
Update on mine,

Running:

HT x 3
FSB x 285
CPU x 9.5
Timings 2.5-3-3-11
Core voltage 1.62
Mem voltage 2.85
Click to expand...

What board again bro? Just asking, because typically people try to stay away from the .5x multipliers.
 
I just got a 3000+ 0517 APFW retail from Monarch. Ordered Sunday and received today. It's doing 2.6ghz at 1.4 volts with a Zalman 7700-cu. I haven't had time to try anything higher yet, and if I did, the 4400+ is more fun to play with right now!

What an awesome HTPC cpu. This is so much better than my Intel spaceheater. With the same cooler and at stock speeds, this venice is running at HALF the load temperatures of the P4 530.

EDIT: 2.7ghz at 1.45v now.
 
working on updating! sorry i've been so busy, between the a64 memory thread, my 1gb crucial value, work and girl issues :rolleyes: :D

Romir, you forgot the stepping :p
 
(cf)Eclipse said:
working on updating! sorry i've been so busy, between the a64 memory thread, my 1gb crucial value, work and girl issues :rolleyes: :D

Romir, you forgot the stepping :p
Click to expand...

Haha, at least you CAN updated....my PSU is dead and I'm waiting on a final reply from OCZ support so that we can get going on the RMA. This is why I'm glad I got my Powerstream PSU.....5 year warranty, first 3 years qualify for a cross-ship RMA...which I'm taking advantage of. As soon as I get ahold of them this last time it shoudln't be long before I'm up and running....but then once I have the money, I'm selling the new 520W Powerstream to a friend and spending $100 to buy a 510W PCP&C for my up and coming project. :)
 
heh nice. sucks to hear that it died though.. i love ocz's warrantee

also, notice what i put in for your overclock :p :D
 
I keep seeing those results and man I'm tempted to remove the IHS this sunday/monday when my ceramique gets here, the neoprene I have installed should act as a good cushion, if it isn't too tall already once the IHS is removed. The key would be keeping it from getting crushed, but once installed I bet the chiller and RBX would make short work of 3ghz, finally.
 
I can hardly wait to just setup a regular h2O loop on mine....2.96GHz on air at 1.600v.....I'll be benchin' at 3GHz+ on water for sure...very nice. :cool:
 
mikelz85 said:
The key would be keeping it from getting crushed
Click to expand...
shouldn't be a problem unless you're careless with it. my newark is under water, and it definitly didn't come with a heatspreader :p
 
(cf)Eclipse said:
shouldn't be a problem unless you're careless with it. my newark is under water, and it definitly didn't come with a heatspreader :p
Click to expand...

Water and direct die cooling are two different things...a lot of guys running chillers and phase change setups usually end up applying a great deal more preasureto the core...and it's not always the safest thing.
 
SpoogeMonkey said:
Just an update:

2.628 Ghz (292 x 9) 1.425 vcore, ram 239mhz 1t 2-3-3-6 2.7vdimm 166 divider



24 hrs prime stable, still on stock cooling, going to let it fold for a few days until my scythe ninja gets here, then I'll press on. Stock hsf is screaming loud, I can hear it on the other end of the house when I go to bed :(

(And no distracting wallpaper this time) :D
Click to expand...



Final OC :
ADA3000DAA4BP LBBLE 0517APJW
2.661mhz , 1.45 vcore, 296 htt x 3 , ram 242mhz 1t 2-3-3-6 , 166 divider 2.7 vdimm
CPU Temp: 40c load 32c idle
Super Pi 1.4mod: 1M = 31.53sec


The cpu has more headroom left, but this is the max speed of my memory. Even with looser timings and more voltage, I couldn't push it any more. 2.73ghz w/ram @ 200mhz, just isn't as fast as current ram settings.
 
SpoogeMonkey said:
2.73ghz w/ram @ 200mhz, just isn't as fast as current ram settings.
Click to expand...
really? that's strange.. as it goes against basically all of the testing i have done with memory and my single channel newark that is definitly more memory limited than the venice :confused:
 
yeah, when I read that it didn't make sense to me. Benchmarks show that scaling the A64 gives way more performance than scaling the RAM. Also, if you have to drop the frequency down because of dividers, couldn't you tighten up the timings from 242mhz 1t 2-3-3-6, to 2-2-2-6 1t at 200mhz?
 
You must log in or register to reply here.
Back
Top