Question about memory timings

[Flashes]

I read the guide about overclocking (excellent articel, by the way) on the 754/939 platform and I have a question about my results following that guide.

I have had my memories for some time and I can't get them above 250 MHz what ever I do so I followed the guide hoping that I had missed something and could do better this way.
I know that they don't like 2 2 2 5 timings so I started from what I use in my 24/7 setup 2,5 3 3 6 1T 2,6V

I increased the FSB in steps of 10 MHz up to 230 MHz since I know that the memories could take these speeds and continued in 5 MHz steps from that point.
I did run the memtest86+ test 5 and 8 for 15 min on each test before I increased the FSB again.
I can get my memories up to 245 MHz when I have to use 2,5 4 3 6 1T and can get the memory up to 250 MHz, after that it doesn't matter what I do, the memory sticks just don't go higher than that.
If I loosen the timings to 3 4 4 9 2T the computer doesn't even boot up and I have to set the timings back to 2,5 4 4 8 2T but I can't get any higher speeds than 250 MHz anyhow.
If I increas the vDIMM when I get faults in memtest86+ I just gets more faults during the tests not less, I'm just confused right now and don't know how to proceed.
Any sudgestion on what to do, because I'm stuck?
I'm not an expert overclocker this is only my second serious atempt to do something like this. My system is in my sign.

Additional to this I have a question on FSB since I can't boot into Windows if my FSB is above 260 MHz there must be a way to lock the FSB for the HDDs since a lot of people have FSB much higher than 260 MHz.
I have tried my HDDs on SATA 1 & 2, or SATA 3 & 4 but the result is the same. According to a forum in Sweden, I'm a Swede, the SATA 3 & 4 should be locked on a DFI board but it seams not to be or is there any other reason why I can't boot into win with high FSB? I have lowered the multiplier on the CPU and put in a stiff divider on the memories but still I can't boot into Windows and that only leavs the HHDs or?

[Flashes]

I have got an answer to my FSB question and it seams actually that all SATA ports on a DFI board is locked. My problem was that the CPU needs more VCore when the FSB gets higher for some reason, despite the fact that I lower the multiplier to keep the CPU speed below 2,8 GHz that I can run Prime stable at 11*255 and 12*235.
At 10*280 I need 0.05 Volts more at the same speeds, strange that is for me.

[muddocktor]

Glad to see you figured the problem out, Flashes. I imagine the reason you are having to bump the vcore up slightly with high HTT speeds above 260 is that the integrated memory controller might be needing a little more voltage to deal reliably with the stress of communication at the higher HTT speed. At least that's what I think. :)

[lemonlime]

Hi Flashes,

Glad you liked the overclocking article. I think mudd hit it right on. I noticed that you are also using 2x1024MB dimms, which adds a good deal of stress to the IMC when pushing beyond PC4000 speeds.

A few other things you can try:


Loosen your tRC timing (some of the DFI bioses default to 7.. try 9).. ensure tRFC is set to AUTO.

Experiment with DRAM drive strength. Try a 'Drive Strength' of 7 and a 'Data Drive Strength' of 2.

Increase your chipset voltage (vchip) one notch. (try 1.6V)

Try to cool the top right hand corner of the board. The PWMIC temp can get way out of hand when overclocking dual core chips. This may also get you some additional headroom. Just a ghetto-mounted 80mm fan decreased my PWMIC from 50'C at load to 34'C at load.

Experiment with other dividers, especially the 5/6 divider. I noticed in your sig that you are using the 9/10 divider. I know some chips choke a bit on that one. It really varies.. I had no problems using 9/10 on my Sempron/A64, but my Opteron looses a great deal of memory OC headroom for some reason with it. It's worth investigating.


These things are worth a try, but it is also quite possible that you are reaching the limitations of your memory. There are many DIMMs on the market that do not benefit whatsoever from extra vdimm. I know my samsung TCCC actually behaves better with lower than specified vdimm. 2.5V was better than 2.6.. Give that a try too :)

Nice rig you've got there btw :thumbsup:

[Flashes]

Thanks mudd and lemon

I will for sure test Your sudgestions and come back with the results.
It will take a fue weeks since I'm having a couple of work related trips to do, Japan and Austria.:type:



Nice rig you've got there btw :thumbsup:
Thanks again :)

[Flashes]

Cooling the PWMIC did wounders, I hit 3 GHz right away :)
I use the 5/6 divider on the memory and also the other settings recommended by lemonlime, I will try my 9/10 later.
Dual Prime95 (small FTT) stable for 4 hours so far, I can run dual super-pi (32MB test) at the same time as well so I think the setup is pretty solid.
http://valid.x86-secret.com/show_oc?id=74120

[lemonlime]

Cooling the PWMIC did wounders, I hit 3 GHz right away :)
I use the 5/6 divider on the memory and also the other settings recommended by lemonlime, I will try my 9/10 later.
Dual Prime95 (small FTT) stable for 4 hours so far, I can run dual super-pi (32MB test) at the same time as well so I think the setup is pretty solid.
http://valid.x86-secret.com/show_oc?id=74120

Fantastic! That PWMIC gets really toasty with dual cores. That must be a rockin' PC congrats.. :)

[Flashes]

I did pass almost 14 h of prime when one core, CPU0, failed :(
The temp on the PWMIC had gone up to 55 degC and the CPUs to 52.
I have to increase the volt but that will be another day, leave for Japan in 5h.

[tmh88]

I never really got into overclocking. My pc has been fast enough for what i need it to do. I overclocked my amd 64 3200 from 2.2 to 2.55 and thats about it. I would like to get watercooling and see how far i can go with it.

[Flashes]

I have probably killed CPU0 it can hardly do 2800 now, don't know what has happened.
I haven't gone above 1,58V on the CPU at any time, one day when I run Prime CPU0 failed as normal before CPU1 and after that it just don't want to oc anymore :(
CPU1 is stable at 3102 (11*282) at 1,55V but it doesn't help when the other core fail,
well **** happens.

[lemonlime]

I have probably killed CPU0 it can hardly do 2800 now, don't know what has happened.
I haven't gone above 1,58V on the CPU at any time, one day when I run Prime CPU0 failed as normal before CPU1 and after that it just don't want to oc anymore :(
CPU1 is stable at 3102 (11*282) at 1,55V but it doesn't help when the other core fail,
well **** happens.

Wow, Core1 clocks really high, but with dual cores, there is almost always one core weaker than the other. I don't think the overclocking you did previously would 'wear down' core 0 to 2800MHz. I think it is more likely changes in ambient room temperature, or other variables of that sort. You mentioned 3GHz was stable for 14 hours, but how quickly does it fail now? have you changed anything in the BIOS since then?

[Flashes]

I didn't change anything in the bios at that time, I have tried to tighten my memory after that, I'm running them at 2,5 3 2 5 1T with no problem at 2,6V

Prime fails within 3h and that's to short for my liking, some times after less when 1 min. so something isn't right
When I back down on the MHz I get core 0 stable at between 2805 and 2820 depending on mulitiplier and so on.
The CPU works just fine and I can play the games with no problems but when the CPU was running 3GHz it's to bad this happens.

[lemonlime]

I didn't change anything in the bios at that time, I have tried to tighten my memory after that, I'm running them at 2,5 3 2 5 1T with no problem at 2,6V

Prime fails within 3h and that's to short for my liking, some times after less when 1 min. so something isn't right
When I back down on the MHz I get core 0 stable at between 2805 and 2820 depending on mulitiplier and so on.
The CPU works just fine and I can play the games with no problems but when the CPU was running 3GHz it's to bad this happens.

Just a suggestion, give slightly higher vdimm a try (memory voltage) try 2.7-2.8V just to see if it makes any difference. I'm just curious to see if it has any effect.

2.8 is still an awesome overclock for a Toledo based core on air-cooling, so don't feel too discouraged if that is the best Prime95 stable config you can do :)

[Flashes]

Well I'm not on air anymore I'm using water.
Increasing the Vmem doesn't do anything for the memories that I know, will try and see if it does anything for the CPU, I will come back.

[Flashes]

Just a suggestion, give slightly higher vdimm a try (memory voltage) try 2.7-2.8V just to see if it makes any difference. I'm just curious to see if it has any effect.

2.8 is still an awesome overclock for a Toledo based core on air-cooling, so don't feel too discouraged if that is the best Prime95 stable config you can do :)

Just to test I increased the MHz to 2904 (242*12) and 1,55V
I increased the VDimm to 2,7V and are running dual Prime, they didn't fail right away so maybe You have something here lemonlime :) :Rocker:

EDIT
Nope both 2,7 and 2,8 gives me error on core 0

[lemonlime]

Just to test I increased the MHz to 2904 (242*12) and 1,55V
I increased the VDimm to 2,7V and are running dual Prime, they didn't fail right away so maybe You have something here lemonlime :) :Rocker:

EDIT
Nope both 2,7 and 2,8 gives me error on core 0

Yep, looks like its just close to the maximum the core can handle. 2.8 is an awesome clock though, thats for sure :) How much voltage do you need for 2.8 prime stable?

[Flashes]

1,45V at 2805 MHz so I'm pleased with that.
I changed my memory timing to 2,5 3 3 6 1T just to test, the memory are 24h memtest86+ stable at 2970 MHz (270*11) at 2,5 3 2 5 1T, and the CPU has just passed 11h of dual Prime95 at 2904MHz :) 11h is good enought for me on this PC since it's not a 24/7 mashine. I run at 1,55V will se if I can increase the MHz more or decrease the VCore.

Thanks for Your input lemonlime.

[Flashes]

Run some more tests during the night it seams that 2915 11*265 is stable at 1,55V but anything higher causes core 0 to fail.
I will try with other mulipler but I guess that 2915 is the roof for core 0 at this stage.
I have tried 11*270 at upto 1.65V but it doesn't help a singel bit, the core 0 fails even quicker at that VCore.

I am realy not disapointed of my CPU but it would have been fun if both cores there a little more the same, and especially if they where more like core 1 :)

[lemonlime]

Definitely a sweet OC, Flashes! :thumbsup:

Glad you are pleased with the results.