Overclocking DDR2 RAM from 533 to 667 problems

[Derek12]

Hi

I currently have 2 RAM sticks, one (A) has 2GB in bank 1, the other (B) has 1GB in bank 2. "A" is DDR2 667 and "B" is DDR2 533. Both are Kingston.

If I set BIOS to DDR2-533 or Auto, none problems. But if I set RAM as DDR2-667 (overclocking module "B") Windows 7 continuosly restarts shorty after the "Starting Windows" screen appears. If I try to install it with the CD, after loading files, I get a BSOD that says sometimes error in acpi.sys and sometimes that my BIOS is not fully ACPI compliant.

What is weird: Memtest give none errors, and I can boot a liveCD of Windows XP and Ubuntu without problems, can this means that stick B isn't stable (I doubt that) or can be other issue?

Many thanks

 

[Bo$$]

put stick A in only, overclock to whatever you are trying, see if you still have the issues.
how many repeats (passes) are you doing the mem test?

 

[erocker]

At what timings to the sticks run at individually inserted in the mobo? Voltages?

 

[Deleted member 24505]

At what timings do the sticks run at individually inserted in the mobo? Voltages?

Let me correct that for you erocker Always wanted to correct you


OT, You might not have much joy trying to oc sticks with differant sizes and speeds.

 

[erocker]

Let me correct that for you erocker Always wanted to correct you


OT, You might not have much joy trying to oc sticks with differant sizes and speeds.

You didn't correct me, though you are correct in thinking that these sticks just might not work together. I've had success running two different sticks of RAM. What I was getting at is taking the stick with the highest timings and apply it to the other stick. It surely can be done but it's not going to work with AUTO settings on a motherboard.

 

[Bo$$]

You didn't correct me, though you are correct in thinking that these sticks just might not work together. I've had success running two different sticks of RAM. What I was getting at is taking the stick with the highest timings and apply it to the other stick. It surely can be done but it's not going to work with AUTO settings on a motherboard.

he replaced 'to' with 'do'
OP said he used auto settings

 

[Derek12]

put stick A in only, overclock to whatever you are trying, see if you still have the issues.
how many repeats (passes) are you doing the mem test?

Yeah If I remove stick B then all works fine. I can even put it on DDR2 800 without problems

Regarding memtest, I run already 1 pass and a half, I don't have time to run it more and overnight it makes noise

At what timings to the sticks run at individually inserted in the mobo? Voltages?

Stick A runs at 5-5-5-15, and B at 4-4-4-12
When I insert both, timings are the same as stick B, and when I overclock are as Stick A...
Voltage is stock (1.8), I'm afraid to overvolt them


OT, You might not have much joy trying to oc sticks with differant sizes and speeds.[/QUOTE]
Yeah they are totally different!
I want to run at DDR2-667 because with the "wonderful" stick B speeds of both sticks are downgraded to DDR2-533

Many thanks

 

[erocker]

At 667mhz try both sticks at 5 5-5-18 1.85 volts. Input these value manually. Perfectly safe (and stock) voltage for DDR2.

 

[Bo$$]

anything up to 2.0v (without heatspreaders) is safe for ddr2
you really should be trying to get 8 passes if possible

 

[cheesy999]

anything up to 2.0v (without heatspreaders) is safe for ddr2
you really should be trying to get 8 passes if possible

seriusly? all the way up to 2v without heatspreaders, i should try that on mine and see if i can reach 1066

 

[Bo$$]

yep, ive seen g.skills with 2.1v as the stock voltages

 

[Eric_On_Web]

Hi

I currently have 2 RAM sticks, one (A) has 2GB in bank 1, the other (B) has 1GB in bank 2. "A" is DDR2 533 and "B" is DDR2 667. Both are Kingston.

If I set BIOS to DDR2-533 or Auto, none problems. But if I set RAM as DDR2-667 (overclocking module "B") Windows 7 continuosly restarts shorty after the "Starting Windows" screen appears. If I try to install it with the CD, after loading files, I get a BSOD that says sometimes error in acpi.sys and sometimes that my BIOS is not fully ACPI compliant.

What is weird: Memtest give none errors, and I can boot a liveCD of Windows XP and Ubuntu without problems, can this means that stick B isn't stable (I doubt that) or can be other issue?

Many thanks

Currently you dont OVERCLOCK module B since its already at 667 but you are overclocking module A that is at 533.

 

[Eric_On_Web]

I was able to push a DDR2800 of Crucial Rendition at stock voltage stable at up to 1000 so i suppose the kingston should be able to do this...or not

 

[qubit]

Even though the 533 module does 667 or 800 on its own, it's still working out of spec and it looks like the signals are degraded enough that when you pair it up with the other module, you still get corruption and the PC falls over.

This is the case even though memtest reports no errors. Hardware on the edge is funny like that.

Try the settings that erocker suggested and see if it's any better.

 

[Derek12]

At 667mhz try both sticks at 5 5-5-18 1.85 volts. Input these value manually. Perfectly safe (and stock) voltage for DDR2.

Many thanks I tried that (though the default voltage is 1.8563 but still crashes, what I don't understand is the following:

1. Memtest pass fine.
2. Windows XP live and Ubuntu works.
3. The BSOD that got in the installer says that my BIOS is not fully ACPI compliant!
4. Windows 7 installed on HD restarts continuously but not BSOD (I disabled automatic reboot)

Are these symptoms of bad RAM??

Currently you dont OVERCLOCK module B since its already at 667 but you are overclocking module A that is at 533.

My fault, module A is 667 and B is 533.

Even though the 533 module does 667 or 800 on its own, it's still working out of spec and it looks like the signals are degraded enough that when you pair it up with the other module, you still get corruption and the PC falls over.

This is the case even though memtest reports no errors. Hardware on the edge is funny like that.

Try the settings that erocker suggested and see if it's any better.

Yeah it works out of spec but with latencies increased due to nature of DDR2-667 and the module A which have those. I overclocked module A in DDR2800 and worked fine

Then does this mean that though Memtest passes without errors doesn't guarantee that the RAM is working properly?

Many thanks

 

[qubit]

Then does this mean that though Memtest passes without errors doesn't guarantee that the RAM is working properly?

No, there's no automated test that can guarantee RAM works properly with 100% certainty, unfortunately. At least not on a consumer level.

Sometimes it can be a combination of factors that makes things fall over, while most of the time it will be solid as a rock you just can't tell. That's why, when electronic circuits are designed (analog and digital) they leave a lot of headroom in them, to minimise these unpredictable effects. This is especially important as the components age and drift out of spec, the whole device lasts much longer; it takes just one little component to be too far out to kill the whole device otherwise. By us enthusiasts running them close or closer to their margins, they can behave unpredictably, as you've seen. You've just got to be lucky.

I buy top end graphics cards for my rigs. These run hot and stressed due to their high performance. Therefore, I don't overclock them, even though I know they'll do it well. Adding more stress simply shortens their lifetime, quite possibly before I get to upgrade them. My GTX 580 just cost me £400 a few weeks ago and I'm not risking it for a few percent more performance. The CPU on the other hand... (See specs)

I never overclocked my previous GTX 285 and it still runs like a champ to this day, two years later. It might still have done if overclocked, but I can't be sure.

Tell me, did you run both modules together at the slower speed? If they both work at that speed, I recommend leaving them like that, or getting another higher speed module to match.

I had a situation like this years ago: DDR333 & DDR400 modules.

I got a DDR333 256MB module first and ran it at stock. This is when the 400s were quite expensive, around 2003.

A few months later I got another 256MB module, this time DDR400 as they had dropped significantly in price. Now here's the weird bit.

Run the modules together in dual channel and they would both run fine at DDR400, even more.

But run that DDR333 module at 400 speed on its own and the f*cker would fall over. Not straight away, but the PC would crash or show other instabilities within a few minutes. I even checked that the timings were set identically to the 400 module, but no dice. Relaxed them a bit, still no dice. I think memtest showed errors too, I can't remember now. The 400 module ran fine on its own at 400, so it wasn't the mobo having a fit.

Put them in bed together again and all was well once again.

Odd, huh?

 

[Derek12]

No, there's no automated test that can guarantee RAM works properly with 100% certainty, unfortunately. At least not on a consumer level.

Sometimes it can be a combination of factors that makes things fall over, while most of the time it will be solid as a rock you just can't tell. That's why, when electronic circuits are designed (analog and digital) they leave a lot of headroom in them, to minimise these unpredictable effects. This is especially important as the components age and drift out of spec, the whole device lasts much longer; it takes just one little component to be too far out to kill the whole device otherwise. By us enthusiasts running them close or closer to their margins, they can behave unpredictably, as you've seen. You've just got to be lucky.

I buy top end graphics cards for my rigs. These run hot and stressed due to their high performance. Therefore, I don't overclock them, even though I know they'll do it well. Adding more stress simply shortens their lifetime, quite possibly before I get to upgrade them. My GTX 580 just cost me £400 a few weeks ago and I'm not risking it for a few percent more performance. The CPU on the other hand... (See specs)

I never overclocked my previous GTX 285 and it still runs like a champ to this day, two years later. It might still have done if overclocked, but I can't be sure.

Tell me, did you run both modules together at the slower speed? If they both work at that speed, I recommend leaving them like that, or getting another higher speed module to match.

I had a situation like this years ago: DDR333 & DDR400 modules.

I got a DDR333 256MB module first and ran it at stock. This is when the 400s were quite expensive, around 2003.

A few months later I got another 256MB module, this time DDR400 as they had dropped significantly in price. Now here's the weird bit.

Run the modules together in dual channel and they would both run fine at DDR400, even more.

But run that DDR333 module at 400 speed on its own and the f*cker would fall over. Not straight away, but the PC would crash or show other instabilities within a few minutes. I even checked that the timings were set identically to the 400 module, but no dice. Relaxed them a bit, still no dice. I think memtest showed errors too, I can't remember now. The 400 module ran fine on its own at 400, so it wasn't the mobo having a fit.

Put them in bed together again and all was well once again.

Odd, huh?

Many thanks now all is clear BTW yes, the motherboard clock them at DDR2 533 for both sticks, then I will leave it default

 

[BarbaricSoul]

yep, ive seen g.skills with 2.1v as the stock voltages

My first set of Crucial Ballistix were 2.2v recommended.

 

[Steevo]

Things like clock jitter, clock drift, and other latencies that you cannot control have a direct effect on the stability where the OS will stress the system in different ways than memtest will.

64bit Operating systems are more sensitive to memory errors. I could run all the live CD's, XP, and Vista at 32 bit I wanted while overclocking my last pair of 1066, but Vista or 7 at 64 bit was a no go until I tweaked the timings and dropped speed.