problems with new gskill ram

[phrowzen]

Hey,

I have been running an a64 3500+ s939 90nm at 2.64GHz (240x11) ldt 4x for several weeks now, at stock voltage stable for 31 hours of prime95.

I just got my 2x512MB G.Skill PC4400 2.5-3-3-7 @ DDR550 (the LE model) and I set my clock to 9.5 cpu multiplier and 278HTT and ran my ram 1:1 = DDR556 and at 2.5-3-3-7 1T. After 4 hours of memtest86, I came come to find 74 errors.

What is the problem? I have not done any further testing as of yet, I'm hoping someone will quickly be able to point something out to me. I have not put any extra voltage through the modules, because as far as I know they should be doing DDR550 2.5-3-3-7 at default voltage. I can't see the extra 6MHz making it fail, so it must be something else. I have not found on g.skills website if this ram is 1T or 2T, should i be running it at 2T?

suggestions welcome,
phrowzen

 

[Mrvile]

No 2T ever.

Anyway did you try upping the ram voltage? I know 6mhz doesn't seem like a big OC but anything up to 2.85v is perfectly safe for any ram. And don't be afraid to up the timings, A64 performance doesn't really rely on ram timings at all.

 

[phrowzen]

sorry i dont quite understand "No 2T ever." -- you mean I shouldn't need to use 2T? But ok, ill try upping the ram voltage.

 

[Duvie]

I asked them and it is supposed to be 1t...fact is my did not due 550ddr at raed speed...try 2.5,4,4,8 and I bet you will be fine....

From talking to the main guys I can tell you most of their claims are actually based on prescott and intel systems.....I think the AMD ondie controller is far more sensitive so expect a slight raise of reported spec to run it....Try 2.5,3,3,8...2.5,4,4,8...2.5,4,4,10....

In the end RMA them if they do it for you....


Things they told me...

1) try using lower vdimm not higher ( didn't work for me)
2) power supply?? ( I tried 2 kick arse ones)
3)bios ...I tried 4 diff ones....
4) different slots....tried them....


Ultimately I had 1 bad stick and I am rmaing then...they allow up to 2.9v to stay within warranty...my board does 2.85v so no isse here...

I have 1 stick that is amonster though 580ddr 1:1 cas 2.5,4,4,10,1t and does 620ddr cas 3,4,4,10,1t...all memtested and passes...granted dual will drag it down a notch but 30 ddr and rasing the vdimm up a notch or 2 should have handled that...

 

[phrowzen]

k thanks guys. hmm it sort of upsets me that it may not run the spec timings -- but as you said, I suppose it doesnt make a difference. im testing now at the same speeds i mentioned above, but with 2.75V.

 

[phrowzen]

hmm just got home from friends and looked, after 48 minutes there was 10 memtest86 errors. this was with 2.75V. I moved timings to 2.5-4-4-8, still at DDR556.

ill keep you updated, but please keep suggestions coming in.

thanks,
phrowzen

 

[Duvie]

What test are the errors in??? Are they random or in the same address area generally??

 

[phrowzen]

hmm i didnt pay attention to any particulars, i will try and make note in the future.

 

[phrowzen]

just looked at it again, it seems to be failing at test 5 all the time. at least thats what the current buffer showed, i dont know how to scroll down the error list to view the rest, but all the ones on screen are test 5.

i am trying now just testing the ram at ddr400 2.5-3-3-7 just to ensure there is nothing actually wrong with the sticks

EDIT: Also, about address, i dont know how the addresses are assigned so i dont nkow if they are in the same "area" -- but i do know that the area is not consistant amongst all the errors recorded.

 

[Duvie]

test 5 and test 6 are waht I was told quite a while ago by an elite member are the most prone to error on an amd system...

I run 10 passes all test then run 100 passes on each test 5 and test 6 for full staility test.....

Mine errored at 2.5,3,3,7 in the 20th ,40th,60th and80th pass...LOL it was like clockwork....

My testing for bad stick was a bit more obvious...With the bad stick in slot 1, beep tones and no boot no matter the speed, settings, or vdimm...in slot 3 it post but no windows boot...says file missing...sounded like corruption but I switch the sticks around and put it in slot 2 ot slot4 depending if the otehr stick was in slot 1 or slot 3 and no error in windows loading...I didn't have to reformat or repair.....

Also the test failed in the 800-1000mb range all the time so it was pretty obvious.....

 

[phrowzen]

I see. So what should I be taking away from this? Are you meaning its ok that I am seeing errors in test 5 and 6?

I still have no idea what to do to make this stable at speeds that it should be stable at, by specification. It's real early to tell, but I dont think theres anything wrong with the moduels themseleves now, as memtest86 is still running without errors at DDR400 2.5-3-3-7.

EDIT: When I first started testing the ram, vdimm was on AUTO. As I said, that failed. I then tried 2.75V, which also failed. Now I set it specifically to 2.6V, as you mentioned that higher voltages actually do the opposite of what you might expect. I also selected it to run test #5 only.

EDIT#2: Hmm, again, failed. I'm pretty I'm noticing every time that the fail region seems to be between 900ishMB and 1000MB. Would this indicate a specific DIMM thats failing?

 

[Duvie]

No...error are not acceptable....

What I can tell you is most ram just plain cant do the rated speed on oc'd A64's in DC...Is this a porblem with AMD and early Wichesters?? IO dont know. But I dont see most FX owners ocing getting rated speed out of their ram as well at rated vdimm...

I really think you will need 2.5,4,4,8 to 2.5,4,4,10

Howver I got few errors in a 100 test and I needed 2.5,4,4,10 to get stable for 100 passes....

Lower speeds I could get ratings for cas 2,3,3,8 to work so I dont see why you will have much issue there. Even bad stick I was somehow passing that....


If it doens't run rated speed email them and see if you get the same guy I dealt with...

 

[phrowzen]

Ok well I am testing 2.5-4-4-10 right now. By the way, I noticed I posted my second edit right aroudn the time you made your last post, so just make sure to check that again, because it may be important.

As well, I am running 278HTT, at LDT 3x. This falls below the 1000MHz bus limit that many talk about, but do you believe 2.5x may be required?

EDIT: God! 2.60V memory and 2.5-4-4-10 doing test 5, it failed has like 6 errors in under 5 minutes. I noticed this time the range is between 481 and 977MB

EDIT2: 2.75V 2.5-4-4-10 ... 16 errors in under 5 minutes

 

[Duvie]

I never said the lower vdimm worked...they said try it..I did!! It failed faster...

try 2.8v....

I use 3x with 275 in my testing and it was fine..I heard at least on my board the .5's didnt work...you could alwyas try it...

have you tried different slots?? different bioses???


I think the 2.6v is too low...i would say above 2.7v and go up from there...

 

[phrowzen]

I have not tried different slots, I suppose I can attempt that next. Can you recommend a BIOS to try for me?

 

[Duvie]

What mobo?? MSI?? I tried 1.36mod, 1.37mod, 1.41mod and 1.51 beta and they made nodifference for me in terms of the ram testing....

 

[phrowzen]

yes, msi k8n neo2, same as yours.. ill try these new bioses, and switching slots

EDIT: By the way, 2.5-4-4-10 @ 2.8V failed with 80+ errors running test5

 

[Duvie]

Also how is your 3.3V rail??? Just asking cause they will!!!

 

[phrowzen]

core center says 3.18-3.23V on the 3.3 rail, using enermax 435w. -- by the way, again, read above edit because i made at same time u made new post

 

[Duvie]

Are you getting mor errors with vdimm raising?? If so maybe heat is your factor??? Aklso could be power...Does the 3.3v differ when rasing the vidmm up or down....

That 3.18v if reliable throught the monitoring app is low...5% would say you wouldn't want to be below 3.14v.....

My enermax 431 watt was 3.36v and my OCZ came at 3.26v and I notched it up to 3.36v

 

[phrowzen]

Hmm well I don't have another power supply aroudn to try. I just tested the ram in the other 2 dual channel memory slots, and same problem.

I'm really running out of ideas, I guess the last thing I can try is an alternate BIOS, which I have limited faith in.

Anyways, I'm doubting heat is a major issue. I have 2 exhaust fans, 1 side mount fan, and my cpu has a thermalright xp-90 with a tornado 92mm currently running at full rpm.

Last suggestions? -- Do you figure its the ram thats bad, or something else in my system?

 

[Duvie]

Ulitmately start clocking down until it does pass at rated timings and rated vdimm...then if you can't live with the money you paid and it not delivering them rma them...I dont think you will ever get a refund from GSkill so make them keep sending you new sticks until they finally get a pair that actually runs as advertised....That is what I am going to do!!!

 

[phrowzen]

Yes I cannot live with premium modules not running at their rated speed. I set my system to LDT5x, 200HTT, DDR400 2.5-4-4-10 (basically stock) and I got a blue screen upon loading windows, don't know what that is about. Possibly again a memory problem, so I'm back into memtest86.

Anyways, I will hopefully get a refund/exchange from the people I bought it from, pcsupersite on ebay. They have multiple quantity of this item for sale, so I may get an exchange.

 

[Duvie]

I had no issues at stock or pc3200 so maybe you have some bad sticks there....Here is to hoping it isn't the 3.3v rail...do you have a way to test it from the power supply to see what is being delivered....

Have you tried 2t or even just one stick???

 

[phrowzen]

I just reseated the dimms in slots 1 and 2, and now windows is booting. But I don't know what to think about bad sticks, as I ran DDR400 at slack timings and it went through i think 6 passes no problems (i know thats obviously not long enough, but DDR550 etc was showing problems nearly immediately)

And no unfortunately I cannot test the power supply.