loss of stability

Woden

i got the new memory for my computer yesterday but now my overclock is unstable, at 200mhz FSB the computer crashes out of prime95 in a few seconds, it also crashes out if i lower the multi from say 11 to 9 but its ok if i drop the FSB to 166*11 so it seems to be a combination of the high FSB and the memory, the system at 200 FSB is not just prime unstable but generally unstable.

i ran memtest on the ram last night though at 2-3-3-7 @200mhz nothing fancy i've never used memtest before just booted into it and let it run all night i got

39 passes, 0 errors and 0 ECC errors.

anyone any idea whats going on?

i'm already putting 2.8vdimm through the chip the abit motherboard slightly overvolts so i don't really want to go up to 2.9 as it will void the warranty, i'm gonna test the other ram but the system was fine with that ram previously, im also gonna test the new ram in a different slot.

would a new bios help, or upping the chipset voltage?

cheers

data

Woden

update, regardless of the memtest passing, i still think the memory is borked, put in the old one now in slot 1, it used to be in slot 3 and again its fine, running prime95 no bother now and i'll memtest it later, looks like rma time for me

data

Col_Loki

Did you try the other Ram seperatly? Ie not in Dual DDR format ? If you did, and its unstable then yea it has to be the ram as your computer is fine with the older one.

Sir Random

What make is the ram? It might just be incompatible with the mobo, since it passes memtest fine.
Abit's can be fussy about ram, my IS7 wouldn't accept some Crucial ram that worked fine in other mobos.

EDIT: If it is a ram/mobo conflict, then the latest BIOS might help. You can see what's been fixed in BIOS's newer than yours, by checking Abit's website.

Woden

yeah i've tried just that ram on is own in two slots and is unstable in prime and windows

its corsair 3200LL ram is should be able for 2-3-2-6 but its only stable at roughly 185mhz at vdimm 2.9. so yeah i reckon its got to be the ram

the other stick is corsair 3500 i got the 3200LL though as i only run 200mhz FSB and i don't think you can get the 3500 anymore really

i may look at the new bios' but i don't like the look at the flashing and i don't think there is a conflict

unkel

if memtest runs fully stable with your timings, the memory is fine

Like Sir Random said, look into BIOS upgrade and also explore abit forums for compatibility with your mobo.

Just out of interest, why did you change your memory?

Woden

thats interesting then,

wasn't changing but more upgrading to a gig of ram, i have a hard time thinking its compatibility though but i'll take a gander

OfflerCrocGod

What your saying Data is that you stuck in an extra stick of RAM and your old stick is still there?. Umm well it could be the extra RAM increases latency - the more banks are filled the longer it takes for a signal to propagate through your RAM. The first thing you should do is run the new RAM on it's own, check if it's stable, it sounds as if it should be OK. The second thing would then be to loosen your timings, if they are too tight the signal will have problems propagating. I dont think there is anything wrong with the RAM - it's more likely you are asking too much from your system:dunno:.

Woden

no i've run it as a single stick in two seperate slots and its still not stable and the ram is rated for 2-3-2-6 and i've just been running it at 2-3-3-7 and i still can't get it stable at 200mhz on its own

Woden

right update, i've flashed my bios to the latest version (not for the faint hearted imo, i used the flashmenu windows utility that you can get from abit)

anywho its still unstable with just the one stick of new memory and i don't see what i should have to put 2.9vdimm through it to try an achieve stablity on a stick thats meant to easily do 200mhz, rma time i'm thinking

data

unkel

The ram should be running @default voltage and @default timings (by SPD) @200MHz. Did you actually try this before changing timings and voltage?

If you didn't already, try this with both your old and your new separately. Did you try the new memory in a different rig?

If memtest is stable / errorless @default settings, there is nothing wrong with the memory afaik - anybody any views on this?

Woden

the timings by spd are even tighter then the ones i use i think, i'm not sure what the normal voltage is i think its 2.6volt but i lashed 2.8vdimm through it straight away

yeah i've tried both old and new seperately only the new stuff gives me issues, i don't have another rig though to try it in.

data

unkel

Originally posted by Dataisgod
the timings by spd are even tighter then the ones i use i think, i'm not sure what the normal voltage is i think its 2.6volt but i lashed 2.8vdimm through it straight away

Not all timings can be used just like that, loosening is not more stable by default! Also you're not overclocking the memory, so why voltage increase?

Try running @spd @2.6 volts @200 and report if stable

Overclocking, upping vdimm and other timings come later

Woden

my apologies for doubting you, its at spd now at 2.6vdimm and its running prime95 more fine at the moment, lest see how it gets on.

if it is stable the question is can i get the xms3500 stable at such tight timings also?

data

Adeptus Titanicus

Is adjusting timings a case of "suck 'em and see"? I've also tried loosening timings ever so slightly, and occasionally, the system may fail to POST.

Woden

generally tighter timings are more difficult to obtain, so if you loosen the timings you shouldn't loose stability.

any exception i have seen to this though is that my xms3500 memory doesn't like a CAS about 2, its stable at 2 but not 2.5 or 3.0.

latest development, was running prime95 stable there for about 15mins at 2-3-2-6 at 2.6vdimm, upped the voltage to 2.8vdimm and failed immediately have to say i find that weird that upping the voltage actually looses stability.

anywho at 2.7vdimm its just after crashing out with this error

62C0C895, expected: 83A9F8CB

the errors i get seem to fluctuate form something like this to around error where it expects whatever but gets something higher

i shall continue to tinker and keep ya's updated

Adeptus Titanicus

Originally posted by Dataisgod
...latest development, was running prime95 stable there for about 15mins at 2-3-2-6 at 2.6vdimm, upped the voltage to 2.8vdimm and failed immediately have to say i find that weird that upping the voltage actually looses stability.
...

It is weird allright. I read on some OCZ foum before that my memory had a sweet spot at 2.65, even though it can go to 3v. So I've kept it at that when trying for larger O/Cs, and thought it seemed more stable than 2.85. It let me benchmark the hell out of it for an hour or so, but was not Prime stable...

Woden

just put it back to 2.6vdimm and loosened to 2-3-3-7, the question is now can i get stability with the other stick at this voltage, we shall see.

unkel

And is it stable?

Might have been better to have bought another stick of what you had if the default timings are different :dunno:

Woden

yeah see its a bit messed up because afaik the original xms3500 i bought had exactly the same chips as the xms3200LL at the time, the 3500 was just rated for 2-3-3-7 at 217mhz, while the 3200LL was rated for 2-3-2-6 at 200mhz

however i think the 3500 and 3200 at the time where based on BH chips and now you can't get them and you can't get the 3500, so i was left with buying what i though would be very similar the current 3200LL and whatever that is based on perhaps CH chips

anywho at 2.6vdimm using both sticks at both 2-3-2-6 and 2-3-3-7 it crashes out of prime after 6 minutes.

i'm just gonna check one more think, i know some versions of prime have issues with a gig of ram so i'll see if i can fix that

data

Woden

ok have updated to the 23.8 version of prime95 which as different options to the one i used to have.

whats the best test to be running that will completely test the ram and cpu combination, i've been running a 768mb ram, 100% cpu and FFT up to 1024 test for the last 20minutes and it is going well

but i want to set a test up for an overnight run

cheers

data

Adeptus Titanicus

I believe the Blend Test under Torture Tests is the best all rounder. As it says, tests some of everything, Lots of RAM tested.

Woden

ok thanks folks ran the blend test last night and found out this morning it crashed out after an hour and a half, which is better then before alright, but i don't think i'll be able to get these sticks stable together, as one seems to need high voltage while the other is unstable at high voltage.

i've been talking to the ram guy on the offical corsair forum we'll see what he suggests

data

Col_Loki

God i must say thats very ****ed up!! I know the newer OCZ is having problems stability wise with over 2.8v (which prob isint far off what the NF7 is giving it - overvolts)........ but this was generally with the Higher speed stuff.

Up until lately ive always been under the impression that more voltage = more stability and this has been correct up until recently......... You shouldnt have much of a problem selling the BH5 stuff anyway, as for the other any chance of returning it mabye?

[As a BTW, the CH5 stuff never had problems with high voltage (that ive seen) and i would presume CH6 would be simular.......... might not be winbond at all]

Woden

i think i'll try to return the ram to komplett using the cooling off period, then i was looking at a gig of ram from ocz 2-2-3-5 for 310. if could then sell my xms3500 to make up the difference.

depends on komplett though

data

Sir Random

Originally posted by COL_LOKI
I know the newer OCZ is having problems stability wise with over 2.8v (which prob isint far off what the NF7 is giving it - overvolts)........ but this was generally with the Higher speed stuff.

Where did you see this? :eek:
I recently spent €340 on 1GB of OCZ's new Platinum EB PC3500 and I'm running it at 2.85v (rated for 2.8v). I checked all the reviews I could find, and feedback from other users in OcUK has been very positive.
Was it a ram review? or guys posting on a thread somewhere?

I've been meaning to try a lower vdimm, just to see how much juice they really need. I just left it at 2.85v when I put them in.

Dman_15

i may be able to get you 512 (2x256) winbond bh-5 ram if your interested

Col_Loki

Where did you see this?

http://www.nordichardware.com/reviews/memory/2003/PC4000_roundup/index.php?ez=6

This was one review i have seen a couple of things about this where the OCZ or Corsair has refused to work at higher voltages, or had produced memtest errors.
Very strange and might just be a once off bunch of chips.......... i know they are generally using Hynix at the min and it has no problems whatso ever with high voltage or high FSB.

Think your fine with the EB range, i havent come accross anything bad about it yet. But the increased voltage thing came up of late, mainly in some threads and a couple of reviews (it has been rare and i have been lookin at alot of reviews/threads of late).

[edit] PS didnt mean to scare you BTW, i ment the slightly newer OCZ versions........ not the newest EB range (my mistake).

Sir Random

Originally posted by COL_LOKI
PS didnt mean to scare you BTW, i ment the slightly newer OCZ versions........ not the newest EB range (my mistake).

No worries, I just didn't want to be pushing unnessecary juice through this stuff, it's rock stable at 2.5,2,2,5 2.85v 460Mhz. My mobo vdimm jumps in 0.1v steps, so I've no 2.8v setting. I tried 2.75v and it crashed loading windows.

Woden

i've read a few weird things about voltage regulators on CH5 chips and possiblely someone saying that that they don't like high voltage.

the mushkin special thats around now isn't BH-5 either but BH-6 apparently