[darklink]

I just order a 4gb kit (2x2) of g.skill PI 1066mhz memory and not sure what the problem is. I originally had 4x1gb of g.skill HZ 800mhz memory and was able to OC to 1081mhz at 2.3v. I was starting to get concerned at the fact that I had to over volt by .2 and read that a lot people's modules were dying quick because of the over voltage. I ordered the PI's because they ran pretty much what I had my HZ's at but at stock voltage. The problem is now my OC is no longer stable. The PI's are technically only overclocked by 15mhz to 1081 but cannot get orthos to run stable past 3 hours. When I check cpu-z, it shows that the part number for my the PI's is pc2-8500 but above that it shows as ddr2 800. Is this because the mem divider I run is ddr2 800?

Running memtest86+ now and about a half hour into it with no errors but even there is shows dimm 1 and 3 populated with ddr2 800. So frustrated right now and not sure what to do...

 

[ericeod]

Quote:

Originally Posted by darklink I just order a 4gb kit (2x2) of g.skill PI 1066mhz memory and not sure what the problem is. I originally had 4x1gb of g.skill HZ 800mhz memory and was able to OC to 1081mhz at 2.3v. I was starting to get concerned at the fact that I had to over volt by .2 and read that a lot people's modules were dying quick because of the over voltage. I ordered the PI's because they ran pretty much what I had my HZ's at but at stock voltage. The problem is now my OC is no longer stable. The PI's are technically only overclocked by 15mhz to 1081 but cannot get orthos to run stable past 3 hours. When I check cpu-z, it shows that the part number for my the PI's is pc2-8500 but above that it shows as ddr2 800. Is this because the mem divider I run is ddr2 800? Running memtest86+ now and about a half hour into it with no errors but even there is shows dimm 1 and 3 populated with ddr2 800. So frustrated right now and not sure what to do...

You will have to adjust your tRFC (try 55) and your Transaction Booster (Try an increase of 1). You might also have to adjust your Ai clck twister (lower the setting).

I run My 1066 with 1.8v at 1058 with Transaction Booster at 6 and a 423 FSB. But if I increase the FSB by 1MHz, I cant even run stable unless I relax the above timings.

BTW, mine wont even run at 1080 despite running 1058 at only 1.8v...

Hope this helps

 

[darklink]

Quote:

Originally Posted by ericeod You will have to adjust your tRFC (try 55) and your Transaction Booster (Try an increase of 1). You might also have to adjust your Ai clck twister (lower the setting). I run My 1066 with 1.8v at 1058 with Transaction Booster at 6 and a 423 FSB. But if I increase the FSB by 1MHz, I cant even run stable unless I relax the above timings. BTW, mine wont even run at 1080 despite running 1058 at only 1.8v... Hope this helps

Thanks ericeod... Rep+. That just seems so odd. I know they have a PI set binned at 1100mhz and I assumed that the 1066mhz kits were just down clocked slightly to make them cheaper. I guess they are both towards to the end of their spectrums so not much overhead room. Just disappointed. The only reason I got them is because I didnt want to kill my HZ's but almost thinking I'm just going to run them until they die and keep the PI's as a backup.

Do yours show as ddr2 800 in cpu-z in the DIMM information section?

 

[ericeod]

Quote:

Originally Posted by darklink Thanks ericeod... Rep+. That just seems so odd. I know they have a PI set binned at 1100mhz and I assumed that the 1066mhz kits were just down clocked slightly to make them cheaper. I guess they are both towards to the end of their spectrums so not much overhead room. Just disappointed. The only reason I got them is because I didnt want to kill my HZ's but almost thinking I'm just going to run them until they die and keep the PI's as a backup. Do yours show as ddr2 800 in cpu-z in the DIMM information section?

Yeah, they are tightly binned. I was hoping to at least get 1080 from mine since ram at that speed would not make te 1100 binning, but they didnt. If I were you, I would just send them back and get a refund if possible. The DDR2 1100 modules have been dropping in price lately, so maybe you could pick up a set of 1100 when you need them, and just run the HZs. They do have a lifetime warranty, so you should be able to get them replaced when/if they do die.

And yes mine show up as 6400, but also show SPD profiles for 1066 (if I remember correctly.. on a laptop at school ATM).

 

[darklink]

Quote:

Originally Posted by ericeod Yeah, they are tightly binned. I was hoping to at least get 1080 from mine since ram at that speed would not make te 1100 binning, but they didnt. If I were you, I would just send them back and get a refund if possible. The DDR2 1100 modules have been dropping in price lately, so maybe you could pick up a set of 1100 when you need them, and just run the HZs. They do have a lifetime warranty, so you should be able to get them replaced when/if they do die. And yes mine show up as 6400, but also show SPD profiles for 1066 (if I remember correctly.. on a laptop at school ATM).

I just changed by BIOS settings to all stock and ran the divider for ddr2 1066 and even then when I check cpu-z the max bandwidth line in cpu-z for the memory shows as pc2-6400 so I guess it's not a divider issue.

EDIT: you are right though... its shows SPD table for ECC at 1066 and JEDEC for 800 and 533

 

[darklink]

Okay.... I restored my OC settings back to 450FSB with 5:6 mem divider to put memory (the PI's) back at 1081mhz. Running 5-5-5-15 with 55 for the tRC at 2.1 volts. Left AI clock twister at auto and set the transaction booster to manual with a performance setting of 11. Booted into Windows and now I'm running orthos blend test again. I'll post the results.

 

[ericeod]

Do your best to tighten up the Transaction Booster, it really affects ram performance.

 

[darklink]

I have it set at 11 right now. You mean go lower? I read an article on anandtech about it and 10 is base 333 mem strap. I raised it for stability

 

[darklink]

Quote:

Originally Posted by darklink I have it set at 11 right now. You mean go lower? I read an article on anandtech about it and 10 is base 333 mem strap. I raised it for stability

Which turned out to do nothing as it still failed orthos after about 2 hours. Right now have it at 1:1 ration for ddr2 900 at 4-4-4-12-40-2T at 2.1 volts. I reduced the tRD to 9. Orthos good for a half an hour now so we'll see...

 

[ericeod]

Quote:

Originally Posted by darklink Which turned out to do nothing as it still failed orthos after about 2 hours. Right now have it at 1:1 ration for ddr2 900 at 4-4-4-12-40-2T at 2.1 volts. I reduced the tRD to 9. Orthos good for a half an hour now so we'll see...

I have my transaction booster at 6 with my ram at 1058 and a 423 FSB. The lower (tighter) the better.

 

[darklink]

I didnt know I could go that low. The article I read on anandtech showed a reference table of base tRD timings based on the memory reset strap selected. When using the 333 strap and leaving transaction booster at auto the tRD shows as 10 in the bios (which is what was showing on the reference table). I'll try to lower and see what happens. The OC at 1:1 (ddr2 900 @ 4-4-4-12-40-2T 2.1v) failed quicker than any of the other OC attempts. A few more tests to go and if still nothing, think I'm gonna take your advise and return the mem and see if it can be applied toward the ddr2-1100 kit.

 

[ericeod]

Quote:

Originally Posted by darklink I didnt know I could go that low. The article I read on anandtech showed a reference table of base tRD timings based on the memory reset strap selected. When using the 333 strap and leaving transaction booster at auto the tRD shows as 10 in the bios (which is what was showing on the reference table). I'll try to lower and see what happens. The OC at 1:1 (ddr2 900 @ 4-4-4-12-40-2T 2.1v) failed quicker than any of the other OC attempts. A few more tests to go and if still nothing, think I'm gonna take your advise and return the mem and see if it can be applied toward the ddr2-1100 kit.

I just rebuilt a friend's PC using the G.Skill 1100 in a Rampage Formula and they are clocking to about 1120 stable with only 1.9v. I have his E8400 OCed to 4.1GHz and the FSB at 460 with a 5/6 divider which puts the ram at 1104. I set the transaction booster to 8 and the tRFC to 55 with 5-5-5-15 and all is stable.

I know the tRD table you are refering to (Anandtech's article "Why we were wrong about the X48" pages 6-9).

 

[darklink]

Well.... I've changed my mind about a million times this morning. I setup a return through the egg to return my current kit. I then placed an order for the ddr2-1100 PI's. As I was sitting here at work I started really thinking and then cancelled the order for the 1100's and decided that for right now, I'm just gonna run my HZ's until they die and make a decision then. It's been a long few days since I received the PI's and at this point I'm fine with the HZ being over volted by .2. Thanks for all your help ericeod.

 

[darklink]

Dipped home on my lunch break and pulled out the PI's. Got the pc2-6400 HZ's back in running at 5-5-5-15-45-2T with a tRD of 8 at 2.3 volts. Letting orthos run until I get home again to see if hopefully I can still run what I was running before the PI's.

EDIT: mem running at 1080MHz

 

[ericeod]

I think you made the right decision. I gave up my 4x1Gb Ballistix "Newegg lanfest edition" which ran at 2.1v at DDR2 1200 on my eVGA 680i board, and DDR2 1080 on my P5E board because I wanted to try the 2x2Gb Ballistix. Well those wouldnt even run at 924 without failing memtest so I switched to 2x2Gb G.Skill DDR2 1000 which wouldnt run higher then 980, and finally settled on these G.Skill that can run at their rated speed, but it is less then my original 4x1 set.

 

[darklink]

I got caught up in the wanting a 2x2 set as well. All the stuff you read about how it's less stress on the MB and requires less voltage and blah blah blah. It sounds nice in theory but when you cant get RAM to run at its rated speed or close to without your OC failing, then I'll take my chance with what I know works.

 

[darklink]

ericeod.... what voltage do u have ur NB at to get tRD of 6?

 

[ericeod]

Quote:

Originally Posted by darklink ericeod.... what voltage do u have ur NB at to get tRD of 6?

I am running 1.45v to the NB. But remember I have an X48. The P45 chipset has undergone a Die shrink and cant handle as much voltage. For example, the X48 has an operating range of roughly 1.25v - 1.65v. The equivelant P45 is 1.10v - 1.45v (IMO). So I would think the equivelant for you would be 1.30v to the NB.

What are you currently running to yours?

 

[darklink]

Currently I'm running 1.2 but I had a hunch that I would need about 1.3 to hit tRD 6. I followed the equation on anandtech with my current CAS of 5 and CR of 2 and essentially I could set it as low as 4 but I don't think the stock cooling is adequate enough to bump the NB voltage beyond 1.3 to acheive the lower tRD. Been back at work so if orthos is still running when I get home (by then it will be a good 9 hours stable) I'm gonna bump the NB to 1.3 and shoot for tRD of 6. If I can hit that with my current timings and FSB I think I will be happy.

 

[darklink]

I checked Intel's site for the voltage specs on the P45 chipset and it seems they state the max voltage is 1.21. I guess because of the die shrink the voltage threshold isn't that much. I tried booting at tRD 6 with 1.24 volts on the MCH but it wouldn't POST so right now I have the tRD at 8 with 1.24 volts. My earlier OC failed at tRD 8 with 1.2 on the MCH as when I got home from work my computer was on the login screen and I got a message that Windows recovered from a serious error. It's been about 2 hours on orthos running large in place FFT's. If I can at least get this current OC stable then I think I'll stop.

Maybe if I get a better chipset cooling solution I can push it higher but at least the tRD 8 would be tighter than when it was at 10 on the auto setting. Here's hoping.....