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

Overclocking Escapades: socket 939 Athlon 64 4000+ San Diego (E6)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Tech Tweaker

Member
Joined
Dec 14, 2010
I'm really beginning to wonder if I've just got a lemon of a CPU when it comes to overclocking or if I just don't have sufficient cooling.

Reason I say that is I see all of these submissions by people on hwbot who are supposedly using air cooling and they're hitting 3-3.2GHz with 1.45-1.5V most of the time. I'm needing 1.65V just to get into the OS though, and this is on a chip with a VID of 1.35V.

The problem is not the memory because the sticks I usually use are easily stable at 250MHz with a 1:1 divider and I tried three different sets yesterday just in case it was the memory, ended up finding out that the problem was the processor not having enough voltage and that the memory had nothing to do with my issues. I also know it's not the board, because it's stable well past 250MHz FSB without any voltage increase to the chipset and when I'm "attempting" to get it to run at 3GHz I'm at 250FSB with a 12x multiplier and 1.65V. Which is weird, because I can run 2.95GHz with just 1.55V, so that seems like kind of big jump for just going up 50MHz.

My next thought was to try a higher FSB with a lower multiplier, but I really was hoping to get this to work with the default multiplier.

I did happen to notice though that my mosfet heatsinks were getting very hot after a few attempts at 3GHz, so I mounted a fan to blow air down on them, which did cool them down, but didn't seem to lend any stability to my attempts at reaching 3GHz and then completing some benchmarks.

Any suggestions or insights would be welcome.
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
Well, tried 261 FSB 11.5x multi, 273 FSB 11x multi, and 300 FSB 10x multi today. Still can't get it to boot at 3GHz. Really seems like a cooling issue to me, because at first it either makes it into the OS or almost makes it into the OS, but after it runs for a while it apparently heats up to the point where it overheats and then it won't even try to load the OS and will immediately blue screen and restart as soon as it gets to the point where it is supposed to start loading the OS up.

Tried raising CPU vCore and chipset voltage higher, loosening the memory timings and raising the voltage (it shouldn't need a voltage bump, because I can run them at DDR550 with timings of 2.5-4-4-8-2T and 2.7V totally stable, but I was grabbing at straws), but none of that helped.

Unless this is a cooling issue I'm stumped. Wouldn't surprise me if it was, as I'm using an older case design with only one intake and one exhaust fan (not counting the PSU, which also acts as a minimal exhaust), plus my CPU heatsink I'm currently using is just a Zalman CNPS9500 and is probably not designed for the kind of heat load I'm putting on it.
 

davedree

Member
Joined
Sep 20, 2010
Location
uk
long time since I played with socket 939.
With a overclock that works send us a full cpu-z printout of all your settings.
 

davedree

Member
Joined
Sep 20, 2010
Location
uk
Have you reduced your ldt so your ht link is around the max of 1000mhz

ie = 250 htt = ldt at 4x so ht link = 1000
250 htt = ldt at 5x so ht link = 1250 (probably unstable especially with overclocked memory)
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
Have you reduced your ldt so your ht link is around the max of 1000mhz

ie = 250 htt = ldt at 4x so ht link = 1000
250 htt = ldt at 5x so ht link = 1250 (probably unstable especially with overclocked memory)

Of course.

At 250 FSB I had the HT Link multi at 4x, at the FSB settings above that I had it at 3x.

These seem to be the only CPU-Z screenshots I have on hand, though I thought I had more.

Athlon 64 4000+ San Diego 2900MHz.JPG

Athlon 64 4000+ San Diego 2900MHz SuperPi 1M run.JPG

Athlon 64 4000+ San Diego at 2.96GHz.png

Athlon 64 4000+ 3GHz UCBench 2011.JPG -Just before my system crashed.
 
Last edited:

davedree

Member
Joined
Sep 20, 2010
Location
uk
My first was a winchester 3500+ on an abit av8 3rd eye and xms 4000.
I got that to about 2500. But then upgraded to a San e6 4000+ and that biatch clocked to 3ghz on 1.500v. Wow the memory bandwidths were good back in 2005.
I think it was around 8300 read and 7500 write..

What are your aida memory measurements ?

I then bought an a8n32 sli deluxe along with an x2 4400 toledo e6, that **** of turd refused to go over 2640 at 1.60v

Although you say you've ruled out the memory being the problem running ddr500 took me ages to find a perfect multi/cpu/ht.Theres a sure way to test if the memory settings are unstable just turn the multi down to the lowest with standard timings for the ram and focus on the cpu side of things and come back to the memory when your all good.

If the problem is still there then its heat and just the fact of an old chip and high v-core going through it. Its good to hear about socket 939, my grandads still running on one and my aunt is too.
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
My first was a winchester 3500+ on an abit av8 3rd eye and xms 4000.
I got that to about 2500. But then upgraded to a San e6 4000+ and that biatch clocked to 3ghz on 1.500v. Wow the memory bandwidths were good back in 2005.
I think it was around 8300 read and 7500 write..

What are your aida memory measurements ?

I then bought an a8n32 sli deluxe along with an x2 4400 toledo e6, that **** of turd refused to go over 2640 at 1.60v

Although you say you've ruled out the memory being the problem running ddr500 took me ages to find a perfect multi/cpu/ht.Theres a sure way to test if the memory settings are unstable just turn the multi down to the lowest with standard timings for the ram and focus on the cpu side of things and come back to the memory when your all good.

If the problem is still there then its heat and just the fact of an old chip and high v-core going through it. Its good to hear about socket 939, my grandads still running on one and my aunt is too.

I have no idea to be honest, I've never run AIDA. Edit: what speed would you like me to run it at though and/or what program should I run it on (got a link?)?

Yeah, I've tried that too. Ran them at the 166MHz (5:6) and the 133MHz (2:3) settings/dividers, but unfortunately didn't help. I wouldn't think it would be a problem though, because the memory I was running will go to DDR600 pretty easily, and DDR500 is no sweat for these sticks (love those TCCD sticks).
 
Last edited:
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
Well, I'm pretty sure it's a lemon.

Can't get it to run at 3GHz with any voltage (within reason for air cooling).

I fastened my Thermalright Ultima-90 down on it with two 92mm fans in push-pull and began cranking up the volts, starting at 1.5V and slowly raising it until I was clear up to 1.7V. Nothing worked though, it almost made it to the desktop at some settings, but would quickly blue screen and reboot. Reasons for my issues are unknown, the memory and HT Link are in check and running at or below rated specs, which leaves only the CPU, the board itself, or perhaps some component on the board which is overheating and causing these issues.
 

Roachgod

Member
Joined
Mar 1, 2009
Location
Texas
I had the SanDiego 3700 for my last rig and kept it running at 2.6ghz at stock volts the whole time. I was using some Geil platinum memory though I believe they were 2-2-2-6 if I remember right and an Asus a8vdeluxe moboard. Anyways I gave the old system to a friend of mine when I built my cuurent system and now it will only run at stock. Hope it gets better for your project. Did you try running only 2 sticks of ram at 1T, its been so long but I think that may help for the overclock?
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
I had the SanDiego 3700 for my last rig and kept it running at 2.6ghz at stock volts the whole time. I was using some Geil platinum memory though I believe they were 2-2-2-6 if I remember right and an Asus a8vdeluxe moboard. Anyways I gave the old system to a friend of mine when I built my cuurent system and now it will only run at stock. Hope it gets better for your project. Did you try running only 2 sticks of ram at 1T, its been so long but I think that may help for the overclock?

Interesting to hear that, may have to try out my 3700+ at some point and see what it can do.

Actually it's the reverse, 2T is recommended for increasing stability, and 1T is recommended for greater performance (though sometimes at the cost of stability). But yes, I had tried that too.
 

trents

Senior Member
Joined
Dec 27, 2008
Tech Tweaker, one of the most important considerations in any overclock effort is "what are the core temps under load?" Have you done any temp monitoring during your stress testing to check for core temps? I see no mention of it in this thread other than your fear that it could be a heat-related instability problem and the concern that the Zalman 9500 may be inadequate for the challenge. We know that AMD's tend to become unstable when the core temp rises above 55 C. and if it doesn't have a core temp sensor then we could look at CPU socket temps which would give us some idea of how hot the processor is running. Could you run about 15-20 minutes of Prime95 blend with HWMonitor open on the desktop while you are doing it? Even if you have to back off on the CPU frequency a tad to get it stable long enough to run the temp check it would be helpful.
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
Tech Tweaker, one of the most important considerations in any overclock effort is "what are the core temps under load?" Have you done any temp monitoring during your stress testing to check for core temps? I see no mention of it in this thread other than your fear that it could be a heat-related instability problem and the concern that the Zalman 9500 may be inadequate for the challenge. We know that AMD's tend to become unstable when the core temp rises above 55 C. and if it doesn't have a core temp sensor then we could look at CPU socket temps which would give us some idea of how hot the processor is running. Could you run about 15-20 minutes of Prime95 blend with HWMonitor open on the desktop while you are doing it? Even if you have to back off on the CPU frequency a tad to get it stable long enough to run the temp check it would be helpful.

Yes, I had Hardware Monitor and/or Core Temp running each time I ran blend and/or during benchmark runs, which was several times actually.

The only time I managed to get it to load into the OS at 3GHz it was reporting 66-68°C under load, which I somewhat doubt the accuracy of quite frankly given that the heat sink only felt slightly warm at the base at that point, and no warmer than it had in the previous overclocks. That kind of gave me a scare, and even though I wasn't sure of the accuracy of the temp readout I shut the computer down after it did that. It was idling in the mid-forties, so around 44-47°C. Unfortunately I took no screen shots of the temp display though and didn't actually write anything down, so this is all going from memory.

If memory serves at 2.9GHz it was idling somewhere in the upper-thirties to low-forties, and I think load temps were somewhere in the upper-forties to low-fifties Celsius.

These results were all obtained under the Zalman 9500 heatsink, I unfortunately can't remember what the temps were with my Thermalright Ultima 90 heat sink or maybe I just didn't check that since I couldn't get it to load the OS under my overclock settings.

I currently have that chip out and have swapped in a different one to bench though, so I'd have to switch it out again before I could get any definitive results and/or images of the load temps.
 
OP
Tech Tweaker

Tech Tweaker

Member
Joined
Dec 14, 2010
My first was a winchester 3500+ on an abit av8 3rd eye and xms 4000.
I got that to about 2500. But then upgraded to a San e6 4000+ and that biatch clocked to 3ghz on 1.500v. Wow the memory bandwidths were good back in 2005.
I think it was around 8300 read and 7500 write..

What are your aida memory measurements ?

Okay, out of curiosity I started testing this.

I'll be using an Athlon 64 4000+ ClawHammer for the processor, since I already had it in the system for benching.

Testing done at DDR468 (2.5-3-3-5-1T) with two sticks in dual channel with the CPU at stock speed or as close as I could get without going over.

Read: 7148MB/s
Write: 6789MB/s
Copy: 5904MB/s
Latency: 48.0ns


Now, these settings are with the processor overclocked to 2.8GHz (234FSBx12) and the memory at DDR468 (2.5-3-3-5-1T) with those same two sticks of memory in dual channel.
Read: 7154MB/s
Write: 7150MB/s
Copy: 6168MB/s
Latency: 46.3ns

Now DDR500 (2.5-3-3-6-1T), processor as close to stock as I could get it without going over.
Read: 7501MB/s
Write: 6550MB/s
Copy: 6271MB/s
Latency: 46.5ns

CPU at 2.75GHz (250FSBx11) (highest I can get the CPU at this FSB), memory at DDR500 (2.5-3-3-6-1T).
Read:7647MB/s
Write: 7648MB/s
Copy: 6302MB/s
Latency: 43.9ns