gskill le .. driving me crazy

[madhatR]

no matter what i do .. i cant get my gskill to pass memtest @ 275 1:1 3-3-7-1HT .. and that is what it is rated at .. ive tried it at all voltages .. etc .. its on an msi k8n neo2 plat .. and i have a 3500+ .. all my rails seem fine ...
memtest has nothing to do with my cpu .. just checks the ram stability .. ive been trying and i cant hit 9x275 .. even 8x275 .. mem test keeps failing ..
im running 3x HTT

any ideas ?

[MrCooper]

no matter what i do .. i cant get my gskill to pass memtest @ 275 1:1 3-3-7-1HT .. and that is what it is rated at .. ive tried it at all voltages .. etc .. its on an msi k8n neo2 plat .. and i have a 3500+ .. all my rails seem fine ...
memtest has nothing to do with my cpu .. just checks the ram stability .. ive been trying and i cant hit 9x275 .. even 8x275 .. mem test keeps failing ..
im running 3x HTT

any ideas ?

Your mem-controller might have been maxed out already. What's the highest you can get the ram stable at?

[Sucka]

Try and test it in another system. That's reason to RMA if they won't do their rated specs.

[madhatR]

dont have another system to try it in .. this sucks

how long should i loop mem test for ?? to show they are stable ?? also how do i select to just run test 5 ??

maybe my 3500 sucks .. maybe i should switch to a 3000

[glock19owner]

What voltage are you running them at...if at 2.65v...then raise it up to 2.75 - 2.85...most need around those voltages for stability...

Notice it states 2.65- 2.85v for default voltage...and actually 2.75v is default voltage for TCCD skills...

Also...are you setting the timings manually or setting everything to auto?

[madhatR]

set everything manually ive tried all the voltages .. seems like anything above like 2.6 gets more errors ..

i mean even at 8x275 which is stock speed for cpu .. 2200 .. still no go

this really sucks

[deception``]

1. Because the memory controller of your Athlon 64 is on-die, your processor can be to blame in this situation.

2. What settings are you running on your memory? Have you tried running it at its advertised timings of 2.5-3-3-7? (Assuming the 3-3-3-7 was not a typo). Also, have you played with the voltage? My advice would be to keep it around 2.7, but a couple of deviations here and there can't hurt either.

3. As Sucka said, please try to test these in another system to see if an RMA is not in order.

Good Luck!

deception``

[Reefa_Madness]

To set it to test #5 - which version, as there are some small differences in the latest (I'm currently using Memtest86+ version 1.51):

Look across the bottom of the screen...from the choices select "(C) configuration" which brings up a "Configuration" screen with choices...select "(1) Test Selection", then "(3) Select Test" from the next screen, then enter the # of the test (5 in this case) you want to run and hit "enter", "enter" and it will start looping the selected test.

Some are slightly different, I believe the earlier version you selected "2" after configuration and then "5" instead of "3" for the Select Test screen but the steps should be similar, if not the same.

[madhatR]

well ive tried all the voltages ..

ive been running it on the advertised timings .. 2.5-3-3-7 sorry for the typo ..

i guess i can run out and maybe get a 3000+ and see if i could hit high fsb stable ..

the version of memtest i have is 3.2

now with mem test when testing the ram its also testing the cpu on die controller as well .. well that could definitely be a problem .. i heard the 3000+ is the cpu of choice right now for high fsb and bandwith ?

[madhatR]

well 10x245 4xHTT 1:1 2.5-3-3-7 passed

how long should i be looping mem test ??

also should i try out a64 tweaker ?`

[|3ourne]

try lowering your multiplier. If your cpu is maxed out memtest WILL error out. especially memtest #5. Try dropping your multiplier to 9x and then seeing whether they are memtest stable at those timings. If yes , then its not the ram's fault. Your winchester cannot do 10x275 simple. Unless its a week 40-50 chip then most prolly it wont do that high speeds. But 9x275 should be possible since your stable 10x245 . Also , take off the northbridge fan and apply a nice layer of arctic silver 5 or ceramique. MSI are notorious for not properly applying the nb paste which results in less stability due to nb getting too hot. Do NOT get a 3000+ right now , as venice is around the corner and current winchester steppings are the speed binned CBBID and LBBID ones which have a hard timing hitting 2.4 ghz or so.

[Reefa_Madness]

Wait for Venice to be released before you make any cpu purchase. Should be out first week of May and some places (Monarch for one) are taking pre-orders.

See how the Venice does when there are about 10 or more published reviews out there, plus the comments from the guys that order right away and then you can decide which way to go. Early indications are that if your going to buy a new socket 939 the Venice is probably your better choice. From what I remember, the pricing at Monarch didn't seem to way out of line.

[Reefa_Madness]

well 10x245 4xHTT 1:1 2.5-3-3-7 passed

how long should i be looping mem test ??

also should i try out a64 tweaker ?`


For testing new ram to find max speed you can run just a few loops then move up the ladder with your ram but if you want "for sure" comfort that the ram won't error, you should consider running it overnight. For example, test #4 needs no less than 32 loops to cover all of the various possible permutations according to the information found in the link below. By running Memtest overnight you should have enough runs to test all the memory.

http://www.bleedinedge.com/forum/showthread.php?t=442&page=1&pp=15

[madhatR]

thanks for all the help guys ..

well i just looped 10x250 4xHTT in test 5 .. ran it for about 1hr or so .. 81 passes no errors .. but 10x250 would freeze as soon as it entered windows ..

i tried dropping the multiplier down to 8x .. and tried 275 .. no go ..

i mean 8x 300 would be 2400 .. i wouldnt mind running that ..


Bourne .. ur running 9x300 .. 1T or 2T ?? wut are ur timings ?

[glock19owner]

well i just looped 10x250 4xHTT in test 5 .. ran it for about 1hr or so .. 81 passes no errors .. but 10x250 would freeze as soon as it entered windows ..

This is caused from the on die memory controller...odds are it is the memory controller messing up and not the ram...

What voltage are you trying to get into windows running 10x250?

Also try lowering the HTT multiplier to 3x...

[madhatR]

ok i just looped 9x270 .. 2.6vmem .. no problems .. so im in windows right now at 9x270 .. 2.6vmem and 1.5vcore .. i guess we will see how stable it is ..

the only reason i would get a 3000+ is cuz i could sell my 3500 + and still have some money left over ..

but i guess if i can stay stable at 9x270 ill be happy .. 275 is a no go ..

i would like to run 8x300 .. but mem test errors right away .. even at 2T

[glock19owner]

i would like to run 8x300 .. but mem test errors right away .. even at 2T

That could be from your NEO2...mine errors out at 40% after 289...yet the LC's could run at 300 on a DFI...

Very few NEO2's will run at 300HTT 1:1 anymore...if you are wanting 300HTT+ at 1:1/1T...look at getting a DFI NF4 board...

[madhatR]

im waiting for the dfi NF3 ultra S939

[|3ourne]

thanks for all the help guys ..

well i just looped 10x250 4xHTT in test 5 .. ran it for about 1hr or so .. 81 passes no errors .. but 10x250 would freeze as soon as it entered windows ..

i tried dropping the multiplier down to 8x .. and tried 275 .. no go ..

i mean 8x 300 would be 2400 .. i wouldnt mind running that ..


Bourne .. ur running 9x300 .. 1T or 2T ?? wut are ur timings ?

My timings are as following on the A64 tweaker. 1T ofcourse :rolleyes:

http://img206.echo.cx/img206/4533/a6404409xl.jpg

[glock19owner]

Those are pretty loose timings for LE's...try these A64 tweaks...these are from my old LC's running at 2.85v...

http://home.insightbb.com/~myglock19/5-3bench4.JPG

[madhatR]

wow 1.68 vcore ?? wut cooling

what bios are u running ?

also .. for a64 tweaker .. i keep my current bios settings .. then in windows change them thru the a64 tweaker ?

[itchy5]

Wait for Venice to be released before you make any cpu purchase. Should be out first week of May and some places (Monarch for one) are taking pre-orders.

See how the Venice does when there are about 10 or more published reviews out there, plus the comments from the guys that order right away and then you can decide which way to go. Early indications are that if your going to buy a new socket 939 the Venice is probably your better choice. From what I remember, the pricing at Monarch didn't seem to way out of line.

yea, wait for venice, the mem comtroller is much improved :cool:

[glock19owner]

wow 1.68 vcore ?? wut cooling

what bios are u running ?

also .. for a64 tweaker .. i keep my current bios settings .. then in windows change them thru the a64 tweaker ?

XP-90 with 92mm tornado

1.36MOD BIOS (1.45v with 13% OV)

And I let windows boot then tweak...

[madhatR]

damn i tried to tweak with ur settings and windows froze lol

[|3ourne]

Those are pretty loose timings for LE's...try these A64 tweaks...these are from my old LC's running at 2.85v...

http://home.insightbb.com/~myglock19/5-3bench4.JPG


How are my timings loose for DDR600 ? Trc and Trfc do not affect your bandwidth , just stability . Idle cycle limit of 256 helps stability. Max Async of 8ns is recommended for Gskills , so is read preamble of 5 ns . I dont like using dynamic idle counter , I have a more responsive system with it disabled. For the timings which matter , you sir have looser timings than I do.
That being said , I am not competing or anything , these are the tightest , most stable timings I could run at 300 mhz. I might be able to drop the Trcd to 3 from 4 . Its memtest #5 stable right now but errors out on memtest #8. Still fine tuning it . And I can definitely run tighter timings than your screenie when I am at 275 mhz.

*Edit : Here are the timings of my older with week 37 chips . My newer ones run a tad tighter but the difference is not noticable anywhere other than benchies.
http://img94.echo.cx/img94/3205/a649cy.jpg

[glock19owner]

Trc and Trfc do not affect your bandwidth , just stability . Idle cycle limit of 256 helps stability. Max Async of 8ns is recommended for Gskills , so is read preamble of 5 ns . I dont like using dynamic idle counter , I have a more responsive system with it disabled. For the timings which matter , you sir have looser timings than I do.[/IMG]

These do not effect memory bandwidth but does effect the memory latency which a A64 will run better then on bandwidth alone...

If you would tighten those setting up a little then your memory latency would be greatly improved...

At those settings I was hitting around 34ns on latency...and I bet yours is in the 40's or so...and with the lower latency...my system at those settings would actually be "faster"...

This is why people lower the trc and trfc and also lower the preamble to around 3.5 or so...Also if you would raise the read write QB to 16x and the bypass to 7x...this will impove calculation times along with latency...

As for bandwidth...I was still at almost 8G at 275...

Now when it comes to latency...this is why I traded my Skills for VX's...the system is actually faster and a better performer now at 2.6 - 2.7 then it was at 2.75...

http://home.insightbb.com/~myglock19/32-latency.JPG

[tyson-chris]

It might be a good idea to find out whether what's holding you back is your memory or your board's HTT. Try running a divider on your memory (instead of 1:1) like the 166/200 divider (5/6) or 133/200 (4/6) divider. Then try upping the HTT above 275. If it can't go, it's almost certainly your board/cpu's HTT. If it can, it's probably the memory.

[madhatR]

ok .. i switched the cpu:fsb ratio to 166 instead of 200

and i did 8x275 .. no problem .. then 8x310 .. no errors .. stopped there ..
so the ram was only running at 258 i think on that 310 test

so this means what ?

[madhatR]

update: well it looks like my system cant overclock at all .. even 10x230 crashes .. even 10x220 .. its insane .. im back at stock speeds .. this is truely disappointing

[madhatR]

do u guys think i could just not have enough power .. im running an antec neopower 480

my system specs are:

msi k8n neo2 plat
3500+ winnie
bfg geforce 6800 ultra oc
2x512 gskill le

[Adrayic]

What are your rails like? Are they within spec?

-Adrayic

[madhatR]

according to the bios they are .. but it was weird .. when i was folding . just for stability testing .. nvidia box pops up saying my 6800 ultra is throttling back due to lack of power then my pc reboots

[Adrayic]

Hmm, if your 6800 Ultra is running out of power then there is a good chance that the power supply is responsible for your instability. Do you have another video card that you could throw in to reduce the load on your PSU? If so, throw that in and then run prime 95 (or anything to stress the CPU) and watch the rails (use motherboard monitor or something like it). If all is well, try OCing it again and see if it makes a difference. I OC'd a 3.0c and the minute I gave it a small voltage increase my rails dropped out of spec and the system would reboot. This was w/ a cheap 350w power supply but nevertheless, it shows that a good PSU is essential.

-Adrayic

[madhatR]

i never had a problem with my true control 550 .. but it died on me .. i needed one asap and all i could get was a neopower 480 .. i see i made a mistake .. when im running at stock speeds right now .. sucks so bad .. until i can put some money together and get a 600 watt ocz one

[austinbmxnig]

i just switched to a 3200 venice and my g.skill won't do 275 2.5-3-3-7 and the cpu and mobo (133 divider) will do 2.9 on air, my PSU is undervolting like MAD according to core center, i still think i need to check with a multimeter to be sure though.....are the pots on the blue storm 500w adjustable?