Official GSkill 1gb-GH pc3200 2-2-2-5 info and results

I found out that if I use only one stick, I can OC up to 260@2-2-2-5, but the PC is not stable as SuperPI crapped out on me. I haven't tried the other stick and now I can't remember which one I tried by itself. One more finding is if I use both sticks, Memtest starts giving me errors (address) 0002a225ba0. It's always this same address every time. If I test only one stick, the error doesn't show up. I'll have to try the other stick now, too, and see if I can duplicate a no error test (with one stick only).

I've PMed OPB and we are in the process of info gathering.

What Vcore were you using at 260, 2-2-2-5? What multiplier were you using? Did you get any errors with that stick in memtest at the same FSB?

Try this, using the same stick you were able to run 260Mhz 2-2-2-5 at, run atleast 30 passes of memtest test 5. Set the multi to 260 x 9 = 2340Mhz at default 1.4v - 1.45v vCore Add the second stick at without changing the timings. See if you get errors and try noting the addresses if they appear to be similar.

If you get a lot of errors reduce the FSB to 230Mhz, which is just above your stable FSB with two sticks i believe, set the multi to 230 x 10 = 2300Mhz at default 1.4v - 1.45v vCore. Run memtest test 5 again to check for the errors with the "good" stick. Add the second stick again if you don't get any errors and run test 5 again.

The aim of this is to check and see if its just one of the sticks causing this or a mem controller issue (more to do with bios settings).

If adding the second stick gives you errors at the same address, swap the slots, and if you still get errors at the same address then its most likely to be more related to the CPU as there won't be any good or bad stick. If you are able to clearly identify the stick which runs well then its all due to the other stick holding you back

Hope that helps eliminate a few things

Quote:

---

Originally Posted by Agent-JCDenton

Check your Dram Drive strength, Dram Data Drive Strength and Tref. For me, those are the big three. I can't pass test 2,3 & 5 when I use a "bad" tref, regardless of how much voltage. I've also noticed that when using the wrong tref, increasing the voltage on these sticks gives me more errors.

With Dram Drive strength set to 5,6 & 7 I'm stable at 250 with 3.2v and i can set Data Drive Strength to 1,2 or 3. At 260 Mhz 3.5v only Drive Strength 5 with Data Drive Strength set to 2 or 3 will work. Dram Drive Strength set to 6 & 7 would also work but I need more voltage (3.6v) at that the same FSB to be stable.

---

I just have to give a big thanks for that info. Using Dram Drive strength set to 5 and Data Drive Strength to 2, I'm now 24/7 prime stable at 250x11 with my current setup. I just couldn't get it 100% stable above 240fsb. Using your suggested setting, I'm now a happy overclocker :woot:.

I'm having just a slightly better luck. Am now up to 246 semi (95%) stable. The reason for saying that is because now I can run SuperPI 32MB with no error. Memtest with no errors. But when I run a Demo of 3DMark2001SE, PC crashed in the middle of it. Still playing with more TREF and other timings.

What I really don't understand why I can't relax the timing at all. Anytime I relax the timing to something like 2.5-4-3-7, my PC would repeatedly reboots (before starting Windows). I want to go all the way to 260 and relax the timing a little to get the higher Ghz out of my Sandy 3700. But no, I can't!

For me, TREF is the key. I tried many of them, 2560, 3120, 3684 and many others , but none were not stable so far. I research around for any TREF threads and I got some hints to use 4672 (not stable) and I came down to 1168, which is where I'm at now.

Got my RMA setup with G.Skill (OPB). Not sure if I should continue playing around some more or send it back so OPB can send me two sticks that work better. Any opinions will greatly be appreciated (about the RMA or not to).

Thanks in advance...

Quote:

---

Originally Posted by B5I8

I just have to give a big thanks for that info. Using Dram Drive strength set to 5 and Data Drive Strength to 2, I'm now 24/7 prime stable at 250x11 with my current setup. I just couldn't get it 100% stable above 240fsb. Using your suggested setting, I'm now a happy overclocker :woot:.

---

;) Good to see you're whipped those stickes into shape. 250Mhz x 11 is my fastest setup too. Now squeeze everything out of them at that 250Mhz.

Try:

TrTw = 1 - Gives a nice boost in bandwidth
Max Async Latency = 6ns - This can shed off 1 or 2 seconds in Everest Latency times.
Trc = 7 & TrFc = 15 anything lower might not please your Mushkins (CH5)

jcniest5,

I was able to kind of recreate what you were expirencing there. I reduced the vDimm at 250Mhz and tried using a tref which i've notice a lot of people using, 3120, and I started getting lots of errors in memtest 2 & 3 - But only at addresses 00022977a04 -> 553.4MB & 000244291a8 -> 580.5MB. The cure was pretty simple, setting Tref = Auto got rid of all the errors in tests 2 & 3 straight away - increasing vDimm or running 2-3-2-5 didn't work. I checked the tref in memtest and it had defaulted to 7.8uS. Which means it could have been one of these settings:

0908= 100mhz(7.8us)
1032= 133mhz(7.8us)
1168= 166mhz(7.8us)
0016= 200mhz(7.8us)

Test 5 was still giving around 10-12 errors with Tref set at auto, increasing vDimm by .07 from 3.2 to 3.27 fixed that. Drive strengths were kept at 5 for DRAM Drive Strength & 2 for DRAM Data Drive Strength.

What vCore are you using for 240Mhz+ FSB? I'm assuming you're using x10 Multi.

About the RMA, I'd try the settings above and see if they are atleast able to reduce errors at 250Mhz+. But i'd let OPB give more advice on whether RMA would be the better route to take. If you decide to stay with the sticks i'll try and see what other settings I can use to recreate your errors and come up with a few more solutions.

Hope that helps.

Quote:

---

Originally Posted by Agent-JCDenton

;) Good to see you're whipped those stickes into shape. 250Mhz x 11 is my fastest setup too. Now squeeze everything out of them at that 250Mhz.

Try:

TrTw = 1 - Gives a nice boost in bandwidth
Max Async Latency = 6ns - This can shed off 1 or 2 seconds in Everest Latency times.
Trc = 7 & TrFc = 15 anything lower might not please your Mushkins (CH5)

jcniest5,

I was able to kind of recreate what you were expirencing there. I reduced the vDimm at 250Mhz and tried using a tref which i've notice a lot of people using, 3120, and I started getting lots of errors in memtest 2 & 3 - But only at addresses 00022977a04 -> 553.4MB & 000244291a8 -> 580.5MB. The cure was pretty simple, setting Tref = Auto got rid of all the errors in tests 2 & 3 straight away - increasing vDimm or running 2-3-2-5 didn't work. I checked the tref in memtest and it had defaulted to 7.8uS. Which means it could have been one of these settings:

0908= 100mhz(7.8us)
1032= 133mhz(7.8us)
1168= 166mhz(7.8us)
0016= 200mhz(7.8us)

Test 5 was still giving around 10-12 errors with Tref set at auto, increasing vDimm by .07 from 3.2 to 3.27 fixed that. Drive strengths were kept at 5 for DRAM Drive Strength & 2 for DRAM Data Drive Strength.

What vCore are you using for 240Mhz+ FSB? I'm assuming you're using x10 Multi.

About the RMA, I'd try the settings above and see if they are atleast able to reduce errors at 250Mhz+. But i'd let OPB give more advice on whether RMA would be the better route to take. If you decide to stay with the sticks i'll try and see what other settings I can use to recreate your errors and come up with a few more solutions.

Hope that helps.

---

great tips... I managed to get my GBGH sticks up to 269.

Quote:

---

Originally Posted by charlie

great tips... I managed to get my GBGH sticks up to 269.

---

Thanks c. I've been able to get to 265Mhz but i need 3.75v, a bit too much vDimm for me. :D

I should be getting my GH sticks sometime next week! But since I only have a 20pin power supply I can only use the 3.2v rail for now. How high are these things getting on 3.2v? I need 255-260 so how much vdimm should I be looking at?

Quote:

---

Originally Posted by TheMeatFrog

I should be getting my GH sticks sometime next week! But since I only have a 20pin power supply I can only use the 3.2v rail for now. How high are these things getting on 3.2v? I need 255-260 so how much vdimm should I be looking at?

---

I've been able to get 250Mhz @ 3.2v 1.5-2-2-5, and 255Mhz @ 3.3v 1.5-2-2-5. Those extra 4 power supply pins would really come in handy here for your BH5 :) .

But you can also try this to get 3.3v on your board. Its said to work with any PSU ;)

Do you have any idea how this memory performs on Nforce2 board (Leadtek K7ncr18d proII DL or NF7s2 in particular), especially the 2-2-2-5 Timings at a low voltage, up to 2.8v with a FSB at 200 or a little more? :stick:
I just went to the G.Skill homepage forum and there were people having big trouble with running that memory on a NF7s2!

Quote:

---

Originally Posted by Agent-JCDenton

I've been able to get 250Mhz @ 3.2v 1.5-2-2-5, and 255Mhz @ 3.3v 1.5-2-2-5. Those extra 4 power supply pins would really come in handy here for your BH5 :) .

But you can also try this to get 3.3v on your board. Its said to work with any PSU ;)

---

I will probably just break down and get a Powerstream 520 from the egg :slapass: That should do the trick right?

that'll work well... OCZ PSU's are some of the best!

C

Would just a 20 to 24pin adapter do for my Antec PSU? Because looking at the OCZ it's just a 20pin with an adapter...

Quote:

---

Originally Posted by Agent-JCDenton

;) Good to see you're whipped those stickes into shape. 250Mhz x 11 is my fastest setup too. Now squeeze everything out of them at that 250Mhz.

Try:

TrTw = 1 - Gives a nice boost in bandwidth
Max Async Latency = 6ns - This can shed off 1 or 2 seconds in Everest Latency times.
Trc = 7 & TrFc = 15 anything lower might not please your Mushkins (CH5)

jcniest5,

I was able to kind of recreate what you were expirencing there. I reduced the vDimm at 250Mhz and tried using a tref which i've notice a lot of people using, 3120, and I started getting lots of errors in memtest 2 & 3 - But only at addresses 00022977a04 -> 553.4MB & 000244291a8 -> 580.5MB. The cure was pretty simple, setting Tref = Auto got rid of all the errors in tests 2 & 3 straight away - increasing vDimm or running 2-3-2-5 didn't work. I checked the tref in memtest and it had defaulted to 7.8uS. Which means it could have been one of these settings:

0908= 100mhz(7.8us)
1032= 133mhz(7.8us)
1168= 166mhz(7.8us)
0016= 200mhz(7.8us)

Test 5 was still giving around 10-12 errors with Tref set at auto, increasing vDimm by .07 from 3.2 to 3.27 fixed that. Drive strengths were kept at 5 for DRAM Drive Strength & 2 for DRAM Data Drive Strength.

What vCore are you using for 240Mhz+ FSB? I'm assuming you're using x10 Multi.

About the RMA, I'd try the settings above and see if they are atleast able to reduce errors at 250Mhz+. But i'd let OPB give more advice on whether RMA would be the better route to take. If you decide to stay with the sticks i'll try and see what other settings I can use to recreate your errors and come up with a few more solutions.

Hope that helps.

---

Hey buddy, great info..I think someone 's gonna doubt you're one of Gskill..lol

Quote:

---

Originally Posted by Onepagebook

Hey buddy, great info..I think someone 's gonna doubt you're one of Gskill..lol

---

Ok, you got me Kev, my previous job applications were unsuccessful - thought i may as well try a different approach, lol :p:

I think i've given these sticks just about all the settings I can think of and they are holding up really well. My first pair of G-Skill and a damn good experience so far!

Quote:

---

Originally Posted by Rhymers

Do you have any idea how this memory performs on Nforce2 board (Leadtek K7ncr18d proII DL or NF7s2 in particular), especially the 2-2-2-5 Timings at a low voltage, up to 2.8v with a FSB at 200 or a little more?
I just went to the G.Skill homepage forum and there were people having big trouble with running that memory on a NF7s2!

---

I haven't tried them on any NF2 board, tho i can't see why they'd not work as they are rated at 2.7v - 2.9v @ 200MHz

I got these babies yesterday and the first impression was good. 230MHz a 3v tight timings right out of the box.

I seem to be having a problem with the dividers :( Until I get my new PSU I can't run the 5v rail so I limited to about 3v since anything above my rails aren't solid on this PSU. I can boot into windows just fine with the 180 divider at say something like 10x250 that's puts my ram at just under 230MHz. This will be ideal once I get my new PSU and can give these things the juice the need for 255-260MHz. So when I put the 166 divider as soon as windows loads BSOD/reboot. Guess these are the divider problems I've been hearing about but never encountered since I was running high end TCCD before this. :(

Also what should I change here...
http://img223.echo.cx/img223/5919/ram1dr.jpg

You can also try this to get 3.3v on your board. Its said to work with any PSU ;). 3.2-3.3v should be good for 250 which i was able to reach with these settings

Btw, nice wallpaper, where did you get that?

Thanks for the reply! I got the wallpaper from http://www.parkoz.com/zboard/skin/parkoz_wallpaper.php lots of really cool high resolution wallpapers :)

Edit: Also what drive and data strength were you using with those settings?

Quote:

---

Originally Posted by Agent-JCDenton

;) Good to see you're whipped those stickes into shape. 250Mhz x 11 is my fastest setup too. Now squeeze everything out of them at that 250Mhz.

Try:

TrTw = 1 - Gives a nice boost in bandwidth
Max Async Latency = 6ns - This can shed off 1 or 2 seconds in Everest Latency times.
Trc = 7 & TrFc = 15 anything lower might not please your Mushkins (CH5)

jcniest5,

I was able to kind of recreate what you were expirencing there. I reduced the vDimm at 250Mhz and tried using a tref which i've notice a lot of people using, 3120, and I started getting lots of errors in memtest 2 & 3 - But only at addresses 00022977a04 -> 553.4MB & 000244291a8 -> 580.5MB. The cure was pretty simple, setting Tref = Auto got rid of all the errors in tests 2 & 3 straight away - increasing vDimm or running 2-3-2-5 didn't work. I checked the tref in memtest and it had defaulted to 7.8uS. Which means it could have been one of these settings:

0908= 100mhz(7.8us)
1032= 133mhz(7.8us)
1168= 166mhz(7.8us)
0016= 200mhz(7.8us)

Test 5 was still giving around 10-12 errors with Tref set at auto, increasing vDimm by .07 from 3.2 to 3.27 fixed that. Drive strengths were kept at 5 for DRAM Drive Strength & 2 for DRAM Data Drive Strength.

What vCore are you using for 240Mhz+ FSB? I'm assuming you're using x10 Multi.

About the RMA, I'd try the settings above and see if they are atleast able to reduce errors at 250Mhz+. But i'd let OPB give more advice on whether RMA would be the better route to take. If you decide to stay with the sticks i'll try and see what other settings I can use to recreate your errors and come up with a few more solutions.

Hope that helps.

---

My RMA went through and OPB is sending me two new sticks (so he said). I will definitely play with what you recommended (did) and see how the new sticks will do. I haven't receive them just yet.

How do you raise the voltage from 3.2V to 3.27V? There isn't really a .7V increase, is there? All I see is something about .2V option (or is it .4V?), so basically a 3.2V becomes 3.22V (or 3.24V), whichever is the correct format as I have forgotten.

Also, regarding the 7.8us, all the memory modules I have, Everest is reporting that all of the are 7.8us, how reliable is this info anyway?

I'm waiting patiently for my RMA, which hopefully will be here on Thur or Fri.

OPB, can you confirm that (my RMA's ETA)?

TIA...

last time i checked these were no longer on newegg :(


will they come back?

Quote:

---

Originally Posted by Playful_Buffalo

last time i checked these were no longer on newegg :(


will they come back?

---

GH memory, you mean? Probably not as BH-5 UTT is being phased out as well. They already starting making the CH-5, and Samsung TCC5, pretty soon UTT BH-5 and TCCD all will just be nothing but faded memory of the past.

So why does only the 180 divider work for me? Is this the case with everyone else or just me? I can rn 10x260 with a 180 divider just fine but when I try to use the anything below 180 I get a BSOD/reboot as soon as windows trys to start. What's upw with this?

Quote:

---

Originally Posted by TheMeatFrog

Also what should I change here...
http://img223.echo.cx/img223/5919/ram1dr.jpg

---

You can try:
Max Async Latency - 6.0ns
Read Preamble - 4.0ns
Idle Cycle Limit - 0 Clocks
Trfc - 13 or even 12
Trrd - 0

You might need to loosen Max Async Latency back to 7.0ns @ >250Mhz, the same thing might apply to Trfc (12 - 14) and Trrd (2 Clocks).

My other findings:
DRAM Drive Strength : DRAM DATA Strength
The 5 : 2 combination was the most stable
5 : 3 & 6 : 2 also worked but stable only to arround 255Mhz
View this post for a few other tips on Drive Strength.

Quote:

---

Originally Posted by TheMeatFrog

I got the wallpaper from http://www.parkoz.com/zboard/skin/parkoz_wallpaper.php lots of really cool high resolution wallpapers :)

---

That link has some great wallpaper :cool: thanks!

Quote:

---

Originally Posted by TheMeatFrog

So why does only the 180 divider work for me? Is this the case with everyone else or just me? I can rn 10x260 with a 180 divider just fine but when I try to use the anything below 180 I get a BSOD/reboot as soon as windows trys to start. What's upw with this?

---

Several people are experiencing the same thing with dividers on DFI nF4 boards hence i tend not to work with dividers myself, but there's a thread on here somewhere about it.

Quote:

---

Originally Posted by jcniest5

How do you raise the voltage from 3.2V to 3.27V? There isn't really a .7V increase, is there? All I see is something about .2V option (or is it .4V?), so basically a 3.2V becomes 3.22V (or 3.24V), whichever is the correct format as I have forgotten.

---

My board seems to overvolt when it comes to vDimm, e.g. 3.27 vDimm setting in bios is actually 3.2 with +0.3 enabled and JP17 set to use the 5V rail. 3.2 was achieved by setting the JP17 set back to normal (3.3v Rail). I've modded my PowerStream to give 3.37V on the 3.3V rail tho so i shouldn't be complaining.

Quote:

---

Originally Posted by jcniest5

Also, regarding the 7.8us, all the memory modules I have, Everest is reporting that all of the are 7.8us, how reliable is this info anyway?

---

I've never used everest to check tref, i've always done it either in memtest or by using A64 Tweaker.

what is the highest FSB without void the warranty? :) (so 3.1V i think)
latency: 2.5-4-3-8 or something else? but 2.5..

Thank you!

Hmmmm....now that's a good question....no idea how high these modules will go. I guess that just leaves one thing - i'll be testing that 2moro to find out :)