• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Love your socket 754?

GreenMonkey: I didn't know there were any boards that could do half times multipliers. Learn something new everyday.

If your computer is Prime stable at about 2700 MHz then you could use 225X12 or 245X11 or 270X10 or 300X9. Lots of options to choose from without needing to use a half multiplier.

I've benchmarked the last two and there is no difference as long as the memory is running at the same speed in both tests. At 2700 MHz I like 300X9 so I can use a 3X HT multiplier.

My Gigabyte K8NS board has a nice feature when you get carried away overclocking. Turn off the power supply switch, wait a couple of seconds, turn it back on, hold down the DEL key on the keyboard and then push the ON/OFF switch on your computer and keep holding the DEL key as it boots up. Your computer will boot up and skip your previous overclocked settings and allow you to go directly into the BIOS to make adjustments and try something new. I never have to open up the case and reset the CMOS since learning this trick, even when I get really stupid!

My Gigabyte also used to lock up once in a while when rebooting until I upped the chipset voltage. I set it at +0.2 volts in the BIOS and this cured everything. Any problems now are rare at this setting. My chipset came with a fan on it but I actually pulled it off and threw it in the closet. Too noisy for me. I was going to replace it but never got around to it. The chipset is warmer but it doesn't seem to be causing any problems.

My Athlon 3000 was running Prime stable at 2700 MHz and full voltage today. CPU-z 1.34 reports that as about 1.76 volts under full load. It was only a short test but my son was stable in NFS Most Wanted at 2750 MHz for over half an hour without any problems. I upped the CPU fan to 3000 rpm to keep things stable. Good chips once you find what they like.

Don't be afraid to give 'em some juice. By the time you burn one out they'll be giving them away in cereal boxes. My E6 core Sempron ran fine at 1.75 volts full load for a long time without ever having a nuclear meltdown.
 
If you keep it cool there should be nothing to worry about. My overclocking was primarily limited by the high heat problem I was having. 1.7v @ 3.01 ghz was giving me 55-57C startup temps, not a safe place to be.
 
AMD rates most of their Athlon desktop chips at 65C, 69C or 70C. Head to Fab51 to check your particular model:

http://fab51.com/cpu/guide/opn-s754-e.html

Prime testing will heat your CPU to its maximum. Keep it below 60C while running Prime and you'll probably be OK. Most normal applications or games won't heat your CPU up nearly as much as Prime does so your computer should run fine for many years to come.
 
unclewebb said:
GreenMonkey: I didn't know there were any boards that could do half times multipliers. Learn something new everyday.

If your computer is Prime stable at about 2700 MHz then you could use 225X12 or 245X11 or 270X10 or 300X9. Lots of options to choose from without needing to use a half multiplier.

I've benchmarked the last two and there is no difference as long as the memory is running at the same speed in both tests. At 2700 MHz I like 300X9 so I can use a 3X HT multiplier.

My Gigabyte K8NS board has a nice feature when you get carried away overclocking. Turn off the power supply switch, wait a couple of seconds, turn it back on, hold down the DEL key on the keyboard and then push the ON/OFF switch on your computer and keep holding the DEL key as it boots up. Your computer will boot up and skip your previous overclocked settings and allow you to go directly into the BIOS to make adjustments and try something new. I never have to open up the case and reset the CMOS since learning this trick, even when I get really stupid!

My Gigabyte also used to lock up once in a while when rebooting until I upped the chipset voltage. I set it at +0.2 volts in the BIOS and this cured everything. Any problems now are rare at this setting. My chipset came with a fan on it but I actually pulled it off and threw it in the closet. Too noisy for me. I was going to replace it but never got around to it. The chipset is warmer but it doesn't seem to be causing any problems.

My Athlon 3000 was running Prime stable at 2700 MHz and full voltage today. CPU-z 1.34 reports that as about 1.76 volts under full load. It was only a short test but my son was stable in NFS Most Wanted at 2750 MHz for over half an hour without any problems. I upped the CPU fan to 3000 rpm to keep things stable. Good chips once you find what they like.

Don't be afraid to give 'em some juice. By the time you burn one out they'll be giving them away in cereal boxes. My E6 core Sempron ran fine at 1.75 volts full load for a long time without ever having a nuclear meltdown.
Click to expand...

Hm, I guess that's how the delete key thing works. Thanks. It does seem to have to be a complete power cycle of the box to get it to work.

Good point on the cereal boxes thing. I set the voltage to 1.65. I noticed the board is an awful amount off, though...according to easytune the voltage is fluctuating between 1.700 and 1.740 under load. Even though it's set in the BIOS to 1.65. A double check with CPU-Z says 1.648 :confused:

ARGH! I just found out my ram is running at 230 again. Sandra and cpu-z agree. I'm running at 460mhz. Somehow the memory divider isn't working right. No wonder it's not working :mad:
 
Here's what I'm settling on. 2.8ghz seems to be out of reach for stability. I get hard reboots under prime95 at anything over 2.8. Think it's the board at work as the CPU never goes over 50C. 2.77ghz. Good enough. Using 11.5x multiplier. Can't overclock my valueRAM much. IIRC this is 241x11.5.


 
GreenMonkey: Your memory divider may be acting screwy because of the half times multis that you're using.

If you have this CPU running between 2750 MHz and 2800 MHz it's not the board that's holding you back. Nothing's holding you back! :cool:

That's pretty much the reliable outer limits for Venice core CPUs with air cooling. I had my Athlon 3000 up in the 2950 MHz stratosphere but that's just a one shot deal. It's a ticking time bomb at that speed and would likely struggle to calculate 2+2 without returning 5.

Could you try uploading another screen shot to image shack? It's very difficult to make out the details of your combo. Don't be afraid to use up some storage space there, it's free. A png pic might be a little clearer and if you crop off a couple of inches from the top and a few off the side it shouldn't be too big a file.
 
GreenMonkey: Your memory divider may be acting screwy because of the half times multis that you're using.
Click to expand...
Yes, 1/2 multipliers will do goofy things to ram speed readings. I found a chart that lists ram dividers @ 11.5 cpu multi as follows:
200 (1:1) = 11.5/12
166 (5:6) = 11.5/14
133 (2:3) = 11.5/18
So if you are running 241x11.5 on the cpu and a DDR333 (166) divider your ram will be running at (241x11.5/14) = 198 mhz (DDR396).
BTW - you posted previously that you had your Hypertransport mult set to 1x. Try going to 3x as 1x is way too low. This will give you a Hypertransport speed of 723 mhz which should work nicely. If not, back off to 2x.
If you have this CPU running between 2750 MHz and 2800 MHz it's not the board that's holding you back. Nothing's holding you back!
Click to expand...
Agreed. My max has been 2.73 to 2.76 ghz with my 3200+ and 3400+ respectively. As unclewebb observed 2.7 to 2.8 on air is about as high as you'll get. I can get 235x12 with the 3400+ but anything above that doesn't work out. I'll have to try the 11x and 10x dividers but it runs so well right now, especially the ram, that I just don't feel like messing with it much. I may have to splurge for a water setup next. :p
 
If you made it to 2700 MHz with two sticks of RAM with a K8NS-Pro board then you've done real good. These boards are great with one stick but usually really suck with 2 or more sticks. Good work. Your Sempron is at its practical limit so be happy. :)
 
I decided to do some experimenting to try and determine just how hot is too hot.

My Athlon 3000 was set to 2653 MHz ( 294.8 X 9 ) and under full load CPU-z was reporting 1.776 volts. I let Prime run for 5 hours to make sure everything was nice and stable. I use a 3 position CPU fan speed adjuster which was set to medium ( 2250 rpm ).

athlon26532al.png


It was very stable at 58C to 59C and in previous tests has been quite happy running at this speed and temp.

Next up I decided to reduce the CPU fan speed down to 1500 rpm and I let the heat gradually come up. With Prime still running it didn't take long for it to hit 71C . The temperature seemed to stabilize around here for about a minute and that's when Prime reported an error and quit running.

toohot2po.png


AMD officially rates the Athlon 3400 and my Athlon 3000 at 65C and the Athlon 3200 is rated at 69C. Keeping these chips under 60C at full load seems like a good idea to me.

The good news is that my computer did not lock up or catch on fire during the test. Under light load and with the CPU fan at medium the temp has dropped back to 33C. Hope this test re-assures users that these chips can safely take some volts and some heat.
 
unclewebb said:
The good news is that my computer did not lock up or catch on fire during the test. Under light load and with the CPU fan at medium the temp has dropped back to 33C. Hope this test re-assures users that these chips can safely take some volts and some heat.
Click to expand...

holy mother of god, 1.77 volts on air! even if heat isn't an issue, electromigration can occur. I'd strongly reconsider running that vcore...
 
quadnad said:
holy mother of god, 1.77 volts on air! ...
Click to expand...

There's always a fine line between genius and idiot and I think I just might have crossed that line today. :rolleyes:

I just wanted users to feel a little more comfortable when exploring the outer limits of their processor. It won't be too hard to use more common sense than I did today. I've tortured my E6 Sempron 2600 like this for the last year and it still works fine. If my Athlon melts down, I have the Sempron as back up and if it ever melts down I can go buy another s754 chip for $50 these days.

If this was a $1000 X2 chip I obviously wouldn't be treating it like that. It takes a lot of extra voltage to get the last 50 MHz out of a CPU and it's really not worth it. I agree that something a little less crazy maybe around 1.60 volts would be a more sensible long term setting.
 
unclewebb said:
GreenMonkey: Your memory divider may be acting screwy because of the half times multis that you're using.

If you have this CPU running between 2750 MHz and 2800 MHz it's not the board that's holding you back. Nothing's holding you back! :cool:

Could you try uploading another screen shot to image shack? It's very difficult to make out the details of your combo. Don't be afraid to use up some storage space there, it's free. A png pic might be a little clearer and if you crop off a couple of inches from the top and a few off the side it shouldn't be too big a file.
Click to expand...

I settled on the 11.5 multi because I got the best overclock and the closest to 200mhz on the RAM. I might back off 1-2 as although it passed prime, I got a "system recovered from serious error" once with it at that speed. I got it working at 3x HTT now. Not that HTT speed makes any difference (there's guy out there somewhere that did the benchies, pretty much 0 dif between 200 and 1000).

The FSB multiplier kept getting reset to auto while leaving the other settings alone. I think it has something to do with the "press delete key to skip overclocking" Engrish on bootup. I kept a very careful eye on it after that, sometimes the del key reset the RAM divider but not the other settings. :mad:

I found out during all this that one of my sticks of Corsair Value Select was problematic ..errors in memtest...which explains the occasional Guild Wars archive rebuild, hanging explorer.exe during shutdown, other small issues. So I'm not sure how stable my Windows install was. This Windows install went through a few nforce motherboards too :eek: . For example, I had to repair my command.com and related files in the Windows directory to run a DOS app to make a SATA driver disk today- apparently mine were corrupt :mad: Stupid RAM.

New install of windows on my new 160GB SATA is in progress, I should have some better stability results then. And some new screenies. It said 800x600 but that looks tiny. It's hard to tell what's too big and too small for folks with smaller screens on my FW900...1920x1200...

Still I think mine tops out at 2.75ghz or so. 2.8+ is almost but never quite stable.
 
Nice going unclewebb! That's alotta voltage there! I've topped mine out at 1.7v (which, according to CPU-z = 1.73v). I ran mine on up to 244x11.5 l@ 1.68v ast night but it wouldn't run Sisandra (kept crashing). At that speed I backed my ram down to the DDR400 divider with the same timings as in my sig. It ran Superpi in the low 33 sec range.
Since it wasn't stable I just put it back to the sig settings and it seems quite happy. You've confirmed what I always said - the Venice cpu's need a lot of voltage to get to the upper limits. The 3400+ I have now has been the best at getting to 2.7 with merely 1.54v, whereas my 3000+ needed 1.66v to go to 2.65 ghz and the 3200+ needed the same to get to 2.73 ghz. Like GreenMonkey, I don't see the 3400+ getting much beyond 2.76 stable; 2.8 seems out of reach with this cpu no matter how much voltage I throw at it.
I'm amazed by the 71C reading you got - I've never once had mine beyond 59C even at 1.73v. You sir are a madman :p
Thanks for putting your equipment on the line, though. Hopefuilly it will be an encouragment for the rest of us that the Venice can take some volts and not be too badly fazed.
 
BigMacAttack said:
You sir are a madman :p
Thanks for putting your equipment on the line, though.
Click to expand...

Thank you, I think! :) Anything in the name of science.

Using the Gigabyte EasyTune5 utility I've got my CPU fan down to 1350 rpm under light load which is earily quiet at 2700 MHz and 1.7+ volts but as the temp goes up it automatically increases to as high as 3000 rpm to keep it from crashing. Even 3000 rpm isn't really that loud but the quieter the better I say.

Here's a quote from the Gigabyte EasyTune5 manual:
"It is highly recommended to set CPU cooling fan at full speed at 60C."

I normally try to keep things under 60C and I've never had a problem. I don't use any real world apps that heat up my CPU like Prime does. For light duty, general internet use with the fan at its slowest speed it is sitting at 34C. I'm not sure what it hits during gaming but I know it's less than 60C. With the fan now allowed to reach full speed it will probably barely hit 50C.
 
unclewebb said:
There's always a fine line between genius and idiot and I think I just might have crossed that line today. :rolleyes:

I just wanted users to feel a little more comfortable when exploring the outer limits of their processor. It won't be too hard to use more common sense than I did today. I've tortured my E6 Sempron 2600 like this for the last year and it still works fine. If my Athlon melts down, I have the Sempron as back up and if it ever melts down I can go buy another s754 chip for $50 these days.

If this was a $1000 X2 chip I obviously wouldn't be treating it like that. It takes a lot of extra voltage to get the last 50 MHz out of a CPU and it's really not worth it. I agree that something a little less crazy maybe around 1.60 volts would be a more sensible long term setting.
Click to expand...

My A64 3000+ Would take 1.8V on air like a man, err, CPU :D , even though, I stopped trying my luck at 1.8v and went back down to the voltage I needed :)
 
I'm having some issues testing dividers above DDR433. When I go to try a divider of DDR466 or higher, my BIOS shows I'm running as such, but CPUz isn’t showing the correct memory speed for my settings.

Just to be sure, I ran some benchmarks and they DO change when I up the divider...I'm guessing this is a CPUz issue, I'll try Sisoft Sandra and see if it reads correctly.
 
Use CPU-z 1.35. I had the same issue. There is a problem with CPU-z 1.35 in showing improper voltage above say 1.6v but it will show you the proper mem speed.
 
BigMac: Did you check out CPU-z 1.34.1? It shows both the voltage and memory dividers properly on my board and I was just wondering how it worked on your board?

Unknown-One: Everest by Lavalys is another program that might show you what MHz your memory is really operating at.

What benchmarks do you run? Not all benchmarks are accurate and repeatable enough to determine small changes in memory timings. I used 3DMark2001 for years but even with the exact same set-up, I've seen differences of 500 points from one run to the next. The latest versions of SuperPi with 3 digits after the decimal point also show a lot of variation on the same computer from run to run.
 
I used PCMark05 to test the changes, and the margins on its scores are normaly very small (within 10 points) where the jumps i was seeing were over 100 points.

I'm going to head over and download the latest verson of CPUz and Everest...
 
One simple benchmark that I use is the latency program that is included with the CPU-z download.

Here's version 1.34.1
http://www.cpuid.com/download/cpu-z-134.zip

It only takes a couple of seconds to run and is very consistent. Reducing your memory latency is equivalent to increasing your memory bandwidth so lower is better.

The first run was with the memory divider set to CPU / 12 ( mem = 225.5 MHz )
The second run was with the memory divider at CPU / 14 ( mem = 193.3 MHz )

latency1av.png


The rows up to and including the 512 Kb row for both runs are identical since they are totally dependant on the size of the on board cache which is 512 Kb in the Athlon 3000.

The rows after this are dependent on the latency of your memory. Compare the 4 numbers in the bottom 4 rows and in the last 512 ( stride ) column at the far right. You can see latency goes from an average of 121 to 104.

Latency: 121 / 104 = 1.163
Memory Speed: 225.5 / 193.3 = 1.166

The direct relationship between latency and memory speed is obvious. Maximize your MHz, minimize your latency with tighter timings and your computer will be faster.
 
unclewebb said:
My Crucial Ballistix used to enjoy 2.5-2-2-8 at 250 MHz to 260 MHz but only at 2T. If you can run it at 1T up to 245 MHz then you are definitely getting the most out of that RAM.

Good work!
Click to expand...

I'm running mine at 2.5-3-3-8 244mhz 1T right now. Not because of problems but just a lack of time to do proper stability testing since this is my only computer. Someday when I have the time I'll see if I can lower those 3s and get to 250mhz. When I started out I could go to 210mhz at 2-2-2-6 but jumping to 215 was giving me memory errors. At that point I decided to loosen all the timings and just focus on the CPU.
 
You shouldn't have any problems at 2.5-2-2-8 1T as long as your cpu's memory controller is up to the task. My Ballistix is running at 247 mhz at those timings with no issues and can handle the same timings at 230x12.
 
When I first started fiddling around with the timings I got frustrated and felt like leaving that part for later on. :p I'll see what I can do with these sticks eventually.
 
DedEmbryonicCe11 said:
At that point I decided to loosen all the timings and just focus on the CPU.
Click to expand...

Tight memory timings are just as important as a fast processor. Once you get your processor near its max head back and play with those timings.

Fast memory can be equivalent to an extra 50 to 100 MHz in performance.
 
Ok BigMac, I didn't know what Multi you wanted me to test the dividers at so I did it at 10x. here are the results:
.......................................................................
|.Divider..................|..CPU / x..|..Actual Speed..|
| DDR400 (200MHz) | 2000 / 10 | 200MHz .........|
| DDR433 (216MHz) | 2000 / 10 | 200MHz .........|
| DDR466 (233MHz) | 2000 / 9 ..| 222MHz .........|
| DDR500 (250MHz) | 2000 / 8 ..| 250MHz .........|

BTW, you have no idea how hard it is to make an ASCII table 0_o
 
BigMacAttack said:
I am using the DDR466 divider.
Click to expand...
Yes, but at what CPU multi? Your CPU multi has an effect on how dividers work.

The table in my last post is for a CPU multi of 10x and a 200MHz FSB, I'm working on one for 11x and 12x now...
 
uzum4ki said:
http://hardforum.com/showpost.php?p=1029681065&postcount=1

Interesting...
Click to expand...

"0.13 Micron, SOI"
Yeah I saw the ad myself but it's not a Venice so I'm not real interested even at that awesome price. If you get unlucky and are shafted on the rebate (my luck is bad in this regard) you'll end up paying the same as you would for the 3400+ Venice OEM. Good luck if you spring for it though.
 
Unknown-One said:
Yes, but at what CPU multi? Your CPU multi has an effect on how dividers work.

The table in my last post is for a CPU multi of 10x and a 200MHz FSB, I'm working on one for 11x and 12x now...
Click to expand...
Ah - sorry. All specs are listed in my sig. I have a 3400+ Venice at 225x12. 235x12 will also work with the same DDR466 divider and ram timings.
When I ram memtest (back when I had my 3200+ Venice w/11x multi installed) my Ballistix went to 265 mhz @ 3.1v. I stopped testing it at that point.
 
ep1taph said:
http://www.newegg.com/Product/Product.asp?Item=N82E16819103010

Looks like newegg is sporting some OEM 3400+ Venice's.

I think they messed up a bit on the statistic input, but that part # ( ADA3400AIK4BO ) is the 3400+ Venice, right?
Click to expand...

Same part number as my 3400+ venice from starmicro. $21 cheaper there. Then again starmicro's reseller ratings suck pretty badly, it might be worth the $20 to go with the egg. However, I got mine with no problems.
 
You must log in or register to reply here.
Back
Top