Lessons learned from Overclocking my X2 3800+

[govtcheez75]

1) I used to think that prime95 or Stress Prime combined with Memtest was the "be all" of stress testing. After having my processor pass dual prime for 20 hours and memtest overnight...and then crashing in 3DMark05 or COD2 demo in about 4-5 minutes, I learned my lesson.

2) I used to think that looser memory timings = higher stability. 2.5-4-3-7 was crashing out even thought it would memtest beyond 270mhz. 2-3-3-6 works flawlessly without crashing and still passes memtest at 250mhz (which it didn't do with my previous processors).

3) I used to think that higher voltage always solved instability problems (as long as you didn't hit "the wall"). 1.4v-1.5v all gave me similar results.

4) I used to think that a stable overclock would come from about 50-100mhz below where your windows would boot to windows from. This thing boots up to Windows at 2.75ghz. Seems to only be stable from 2.4-2.5ghz.

5) I used to think that raising the VGA voltage helped to stabilize the video card. Doesn't seem to have any effect.

6) I used to think that fast writes had to be off. Turning it off seems to give slightly worse results in benchmarks while offering no more stability.


any other lessons that people have learned while trying to OC their X2? Anyone agree or disagree with any of the statements above?

*update* 10/5/05: ...I tried something last night. I've never set my VGA voltage to anything more than 1.6 before, but thought that I'd step it up. 1.7v seems to have really helped stabilize things!

 

[imported_michaelpatrick33]

My X2 won't boot into Windows at 2750 at 1.47voltage but will run stable (Prime95, Superpi, 3dmark05, gaming for hours) at 1.55 (watercooled). I can get into Windows at nearly 2900 but Prime95 drops out in about 1 minute.

I have a 7800GTX at 521core and 1405memory on stock voltage and don't feel the need for anymore than that myself.

My memory stinks. I have Patriot 1gig dimms and they don't work well with memory dividers on my DFI NF4 board. I end up with 196 at 1.5-3-2-7 timings. If I don't use dividers I can get up to 220-230 with looser timings but I would rather have more cpu speed at this point.

 

[Duvie]

sounds like power issues whether it is a product of the mobo or your power supply...

1) This still works for me....as I can pass games and 3dmark at 2.7ghz yet I cant get prime or superpi to pass....

2) looser timings are stil often better...sounds like memory compatability as I once had geil sticks that would not run cas 3 timings, ever!!! even when cas 2 and cas 2.5 worked

3) often times too high of voltage causes obvious heat issues...My ram liked to 2.75v or less...anything higher could often lead to instability...A well placed case fan later on solved that....they were possibly heating up...

4) I agree here, but nothing new to me....I could boot at 2.8ghz with my winchester yet couldn't get 2.7ghz fully stable...2.66ghz was the best....remember that we are dealing with 2 cores and it is very conceivable the reason you are seeing this is a product of one of the cores being far worse then the other. The main core or cpu 0 may be the factor as it would likely be the only core taxed at boot up, yet once apps stared the cpu 1 would show its shorcomings...

5) Works for me only when ocing...The highest core I could get stable was 385 at 1.55v...raised it to 1.6v and 395 worked....

6) That one is often a questionable one...


Most of the quirks with the X2 are not that much different from the quirks I observed since using a A64 chip with an ondie memory controller...I think that is where ocing has changed the greatest....

I cannot think of any others right now, but I dont agree with many of yours and belive others will not either. I think there is other factors in your case.

Finally no stress test is complete without running some graphics apps....video cards are the second biggest power draw in the system and need to be tested to fully conclude system is well powereed and stable...I know ppl who could pass prime95 and F@H yet freeze in 3dmark almost immediately...It was a power issue solved by a differenet or more powerful power supply unit...

 

[govtcheez75]

Thanks Duvie...I want more opinions or observations. I wonder if my PSU is bad? I have a OCZ PowerStream 520w but the vcore does tend to fluctuate quite a bit. I always assumed that it was the MoBo that caused such fluctuations.

 

[Avalon]

Originally posted by: govtcheez75
Thanks Duvie...I want more opinions or observations. I wonder if my PSU is bad? I have a OCZ PowerStream 520w but the vcore does tend to fluctuate quite a bit. I always assumed that it was the MoBo that caused such fluctuations.

Large fluctuations are bad. That would make me a little suspicious.

I've never had a problem passing 3dmark loops if my system was prime and memtest stable.
BTW, what cooler are you using on that X2? If you're using something that's not much better than stock, that may explain why you can boot so high, yet only get stable at much lower speeds, even if your temps "appear" to be fine.

 

[nealh]

Originally posted by: govtcheez75
Thanks Duvie...I want more opinions or observations. I wonder if my PSU is bad? I have a OCZ PowerStream 520w but the vcore does tend to fluctuate quite a bit. I always assumed that it was the MoBo that caused such fluctuations.

While I do not have a dual core ..I do have SLI -DR and OCZ Powerstream 520

How much fluctuation are you getting...

I am running a Venice 3200@ 2750 with 1.57v and get very little vcore fluctuations....
I have a X800XL with artic cooler...435/540(not much of an OC but really appears to be the limit on it)
2 x 1gb crucial valuram @250 with 2.8v
2 x 200 Seagate 7200.8 hdd
1640 dvdrw
audigy 2
4x 120 mm fans

do you have all 4 connecotrs plugged in on the DFI mobo?

 

[govtcheez75]

Avalon: I'm using a XP-90 with a 92X38mm fan. It idles around 38 degrees. I'm getting water cooling which should get here tomorrow.

nealh: My Vcore seems to fluctuate about .05-.06volts. If I set it to 1.42 in the bios, I get about 1.4 in windows but fluctuates from about 1.36 during load to 1.42 when nothing is going on. Is that out of the ordinary?

 

[govtcheez75]

Originally posted by: nealh
do you have all 4 connecotrs plugged in on the DFI mobo?

4 connectors? I only have 2 plugged in, the 20 pin connector, and the 4 pin connector.

if you look at This Picture, I have the power supply connectors plugged into the white ports. I don't think there are other connectors that I need to connect to the power supply...are there?

 

[nealh]

Originally posted by: govtcheez75

Originally posted by: nealh
do you have all 4 connecotrs plugged in on the DFI mobo?

4 connectors? I only have 2 plugged in, the 20 pin connector, and the 4 pin connector.

if you look at This Picture, I have the power supply connectors plugged into the white ports. I don't think there are other connectors that I need to connect to the power supply...are there?

Hmm..I thought you were using a DFI NF4 Ultra..sorry..I missed read the mobo....

that fluctuation does not seem bad...when I had my Neo2 Plat...at stock speed and vcore...there was more fluctuation, higher vcore I got less..wonder if the same is true here

also does this mobo support X2's properly....I am not sure about NF3 mobos, I see Duvie is using a Neo2 Plat without issues

 

[Gronich]

My lessons learnt:

1) While still running W2K - my old uniprocessor installation - which only uses one core, I find that the CPU runs around 5oC cooler underload at the same clock speed i.e. about 46-47oC.

This allows me to push the speed up to 2.9Ghz@1.5+ volts at the same 52oC load figure I get running in W2K with dual core installation.

I haven't bothered priming at this speed as what's the point of having a dual core but use only one core? But it has been 100% stable while running games...which are still single core.

2) The memory controller on the CPU is a lot better and more flexible than on my Winnie 3000+

3) "The wall" for overclocking is a lot more abrupt than on previous processors I have had
i.e. Primestable at 2.8Ghz take 1.4v but getting stability at 2850Mhz requires 1.52v+ which gives me an 8oC leap in temperatures under load.

 

[govtcheez75]

Originally posted by: nealh

Originally posted by: govtcheez75

Originally posted by: nealh
do you have all 4 connecotrs plugged in on the DFI mobo?

4 connectors? I only have 2 plugged in, the 20 pin connector, and the 4 pin connector.

if you look at This Picture, I have the power supply connectors plugged into the white ports. I don't think there are other connectors that I need to connect to the power supply...are there?

Hmm..I thought you were using a DFI NF4 Ultra..sorry..I missed read the mobo....

that fluctuation does not seem bad...when I had my Neo2 Plat...at stock speed and vcore...there was more fluctuation, higher vcore I got less..wonder if the same is true here

also does this mobo support X2's properly....I am not sure about NF3 mobos, I see Duvie is using a Neo2 Plat without issues

Yes. This one has supported X2 since it's release. It even works with the original bios.

 

[TGS]

Originally posted by: michaelpatrick33
My memory stinks. I have Patriot 1gig dimms and they don't work well with memory dividers on my DFI NF4 board. I end up with 196 at 1.5-3-2-7 timings. If I don't use dividers I can get up to 220-230 with looser timings but I would rather have more cpu speed at this point.

Another ding against patriot dimms. I have still yet to find out if the memory chips are TCCD or something else. I know the XBLK's are reported as TCCD on BP, which is why they are a great deal. Though the non-XBLK's and the 1GB dimms appear to be of another type. I know my non-XBLK 512MB dimms are extremely sensitive to dividers. Though I could be an exception as of now they are generating errors in memtest, and I had to scale back to CAS 2.5 for stability.

Any other folks using the patriot non-XBLK or known TCCD/BP modules?

 

[2kfire]

Originally posted by: govtcheez75
Avalon: I'm using a XP-90 with a 92X38mm fan. It idles around 38 degrees. I'm getting water cooling which should get here tomorrow.

nealh: My Vcore seems to fluctuate about .05-.06volts. If I set it to 1.42 in the bios, I get about 1.4 in windows but fluctuates from about 1.36 during load to 1.42 when nothing is going on. Is that out of the ordinary?

Just wondering, do you have Cool n' Quiet enabled?

 

[chinkgai]

you never enable cool and quiet when overclocking, so i bet govt doesnt

 

[govtcheez75]

Yes. Cool and Quiet is disabled. I'm not a complete noob when it comes to overclocking.

 

[Bull Dog]

DFI Nforce4 mobos have 4 places were you can connect power connectors. 24pin connector, 4pin P4 power connector, Molex connector, floppy connector.

 

[Bull Dog]

Originally posted by: govtcheez75
1) I used to think that prime95 or Stress Prime combined with Memtest was the "be all" of stress testing. After having my processor pass dual prime for 20 hours and memtest overnight...and then crashing in 3DMark05 or COD2 demo in about 4-5 minutes, I learned my lesson.

As someone else said here, it sounds like power stability problems. Mobo isn't getting enough power to remain stable when the vid card starts sucking on the juice.

 

[Thor86]

Are using LARGE FFT stress tests in Prime95, and two instances?

 

[govtcheez75]

Bulldog: I'm using DFI NF3 Ultra-D and not the NF4 baord. Only two connectors in this one.

Thor86: Yep. Large FFT with priority 10 overnight.


...I tried something last night. I've never set my VGA voltage to anything more than 1.6 before, but thought that I'd step it up. 1.7v seems to have really helped stabilize things!

 

[Thor86]

Originally posted by: govtcheez75
Bulldog: I'm using DFI NF3 Ultra-D and not the NF4 baord. Only two connectors in this one.

Thor86: Yep. Large FFT with priority 10 overnight.


...I tried something last night. I've never set my VGA voltage to anything more than 1.6 before, but thought that I'd step it up. 1.7v seems to have really helped stabilize things!

Good man. I can't believe how many people I see using small FFT tests for cpu stability.

Sounds like a motherboard issue. Upping vagp volts definately could improve stability with your graphics card/game performance when overclocking your system.