XP-M packing it in?

manveruppd

manveruppd

manveruppd

Associate
Joined
24 Feb 2004
Posts
1,083
Location
Leeds/Cyprus
I've been running my XP-M 2500 (IQYHA) at 2.2GHz since summer because the heat gave me stability problems. Yesterday it was snowing so I decided maybe it's safe to venture back up to higher speeds:p so I tried to run Prime 95 to see if it was stable at its current speed before upping it. To my surprise, it wasn't: it stopped after only 2'. I've now upped the voltage and it's been priming stably for over an hour (temps only 45 degrees), but I remember it could run stably enough at the voltage I used to have it up until a few months ago. Is the processor slowly but surely dying from wear and tear, and is this normal? Not hugely bothered, but I want to know so I can plan my upgrade path! :D
 
I've put 2volts through one of these on air though its a bit pointless at that point usually. Ive got mine at 1.8v and 2.5k mhz , it'll go higher but I cant get it stable :(
Even now it fails on s+m and occt but does prime fairly well

What motherboard are you on (dfi here), I suspect my psu to be the weakest link. What % variation in volts qualifies as good ?

Its a near silent rig so I cant complain really
 
yuh, mine's dying slowly!

ive got a 2400 (35w) and it had been running at 2350Mhz with 1.75V on air for a good year or just over...

i came to do a complete format so decided to clock down a little to my "100% stable P95 48hr" setting which was 2200Mhz with 1.55 v (oh yes, this thing didnt take many volts)

and lo and behold, windows install would crash after a minute or so, so i had to up the voltage to 1.65 volts to get it stable!

i'm gonna milk this for all its worth till Vista and all its new hardware and spangly bits come out. it should last that long!
 
Yep, shame they don't last longer really :( They're good chips, I was thinking of retiring this to a silent media server or download box when I upgraded, but at this rate it'll need 2V just to run stable at stock frequency by then!

However, I now suspect it's my RAM, not the chip, as it was Priming for an hour without errors on Small FFTs but errored out immediately on any of the other two tests. It's a pair of old Geil 512MB sticks (I forget the model name... not the Value, not the Ultra, the ones in between which they stopped making but people could get 220FSB out of?). I've got them on 2.6V on an NF7-S 2.0. I've slackened the timings to 12-7-7-3 but still getting errors at 210: Memtest would always generate 2 errors before reaching 10%, but on one run it got to 50% without generating any more than those 2 errors, which is strange... Could be just one stick is borked, I might test them individually if I get the time tonight. Might they be overheating? I've got one of those Zalman brackets, it might be worth setting it up to gently fan them and feed them peeled grapes.

PS. MBM5 shows CPU voltage at only 1.66 - I've forgotten how much the NF7 undervolts it! Fairly steady though. Also, the other strange thing is I have it set to 210x11 but clock speed is reported as only 2269 which means either my computer can't multiply or something strange is afoot...

2nd EDIT: Memtest is rock solid at stock speeds, starts giving errors at 208FSB. Will test Prime by upping the multiplier to make sure it's not the CPU. Anybody tried overvolting their RAM?
 
Last edited:
Update: Yep, it's the RAM... running Prime95 Blend test at 200x11.5 and it's been stable for 45' (which isn't "stable" yet but it's better than when it was failing within 2').

I also ran Memtest86 earlier (the bootable kind), and got some very strange results: when set to loop Test #5, it would be fine up to 99% of a pass and then generate about 6,000 errors! This was 100% repetable, it would do the same thing every single pass! VERY VERY strange!

Shame though, as it'll cost me more to replace the RAM than it would to replace the CPU, and I was hoping to carry the RAM over when I went A64 too... :(
 
Nope, I'm a bit nervous about overvolting RAM sticks cause I have no way to measure their temp. I know that people with BH5 RAM and the like can push up to 3V through it, but I have no idea how much voltage my chips would like. I might give it a go, could set up the Zalman bracket as well just to be safe. How far have people pushed theirs to then?

Incidentally, I'm at 200x12 and Priming nicely atm (tightened RAM timings back to 11-5-5-3 as well), so it's definitely the RAM, especially since I used to get 220FSB without a problem before summer.
 
Well I used to have bh5 then had to sell and use sometihng else except I didnt reset the bios settings, opps Anyway it didnt blow the ram up just didnt work.
If you keep it cool, your ok to edge up the volts a bit. No need for more then a few goes to see if it helps or not
 
What mobo are you on? Couple of things that might help - try the mem at 2.5-4-4-11 (nf2 didn't like cas 3 for some reason and I found it more stable at 2.5, and 11 seems to give best stability, without causing problem with your hdd's!)

2.7 or even 2.8 would be fine for the ram (should be rated to at least 2.7 - but google if you're unsure)

Not all NF7 mobo's are stable much above 200FSB - let us know what you've got and there may be a work around (remember you've got a variable multi - so just keep the FSB at 200 and you can tighten up the timings ;) )

Make sure you're got enough juice going to the chipset - helps!

Don't worry too much about the prime 'memory' test - If it's Small FFT stable, and Memtest stable - that's good enough - I never had much luck with the blend tests - not sure why... :rolleyes:
 
MY NF7 was stable ish upto 220 but the same components on a DFI board can reach 243 stable.

I agree on the chipset volts and make sure the north and south bridge heatsinks are properly seated and cooled. I had many problems due to using the northbridge Heatsink to pick up the motherboard when moving it
 
Yep, I'm aware that 220 seems to be limit for most of the NF7s (though a lucky few can get them up to 235). Mine is the NF7-S 2.0 (I was warned not to confuse it with the NF7-S2 or the NF7-S 1.0 when I was buying it ;) ), so I should be able to get 220-230 with it. I actually bought it 2nd hand from MM from someone who'd been running it at 227. The only thing I changed about it was flashing it the the D26 Black Mantaray BIOS.

It came with the NB cooler replaced with a passive Zalman,and I've also been guilty of using it as a carrying handle... :D It may well be overheating because of this, especially as I'm overvolting it (Baddass clued me in to the bug some of these boards have where if you drop chipset voltage to 1.4 they actually give it more current than the BIOS actually lets you select, and this was how I got it to 220 before the summer in fact). I seriously can't be bothered to take the PC apart though, so I'll only do that as a last resort.

Don't worry too much about the prime 'memory' test - If it's Small FFT stable, and Memtest stable - that's good enough - I never had much luck with the blend tests - not sure why...
Click to expand...
I think this must be because Prime wants to test more memory than Windows allocates to it, and so either fails or keeps reading from the swap file and therefore not testing anything since CPU usage is at 30%. When I'm Priming I run a custom test using the blend settings except for testing a lot less RAM (I use 555, just enough to test both chips).

At 200x12 it tested fine for 8 hours last night, but when I tried 210x10 with slackened RAM timings and VDimm upped to 2.7V I still got errors in Memtest (to which I also allocate less RAM than it wants to test to prevent page faults). I've had a sniff about Geil's website and this seems to be the closest match for the RAM I have. It used to be sold on OCUK but they only sell the Ultra-X and Ultra Platinum ones now, haven't seen this particular one for months. Mine doesn't have the advertised thermometre thingy so it's probably not exactly the same, but it's guaranteed to the same timings. According to that though it's rated to 2.95V, so I'm going to try and give it some more juice, since I'm SURE the RAM's what holding me back. I'd completely forgotten about Geil's lifetime warranty, so I'm no longer bothered about borking it completely! It definitely IS deteriorating since I could get 220 in the past, so might as well speed up the process and get a new set! ;)
try the mem at 2.5-4-4-11 (nf2 didn't like cas 3 for some reason and I found it more stable at 2.5, and 11 seems to give best stability, without causing problem with your hdd's!)
Click to expand...
I didn't know it doesn't like CAS 3, nor that 11 is stabler than 12... I assumed that the slacker the timings the more stable it'd be, so I've been testing at 3-5-5-12 (normally I run at 2-3-3-11 and that gives me no problems at stock). I'll try your timings and more voltage for my next test, but what did you mean about problems with my HDD?

UPDATE: Tested at 210x10 with CavemanOC's timings and VDIMM at 2.8 and was a lot stabler, but still produced errors at 250% coverage in Memtest. Don't know if it was the timings or the voltage that made the difference, will experiment more though, maybe risk 2.9, use the bracket, and even reseat the NB HS if I'm REALLY bored!:p
 
Last edited:
I'd try placing a nice 120mm fan over the NB first just to see if that makes a difference - less effort than reseating the whole thing. The tRas setting at 11 is an NF2 thing - only explanation I ever heard was that at tighter timings it places more stress on the ide ports - seems a stretch, but testing has shown that 11 is the fastest setting, so 'meh...' :)

Only thing I can say about the RAM really is be thankful Geil are overconfident about their ability to produce indestructible RAM! Test it at stock (200) and see what it does: Errors=RMA ;) About to send one of my Value sticks back if I can be bothered (bearing in mind it's worth about 2p!) - you got any other sticks to test with?

What voltage are you now running through the XP-M? Try between 1.75 and 1.8 to see if that helps - should be enough power without producing too much heat - then if it works you can back it off to preserve the chip life - those IQYHA's are outstanding (I still have 2 and they'll be clawed from my cold dead clammy hands :p )

You using the Windows version of 'memtest' or the proper Boot disk?
 
"Stress on the IDE ports"? :confused: Sounds made-up to me...

I'm going to take a different approach: got it at 200x12 at tightened RAM timings, and I'm going to eke the FSB upwards from here, however little speed I gain it'll still be more than 2.4 which was my pre-summer oc. When the RAM starts giving errors I'll fit the Zalman bracket over it (with a big enough fan it should catch the NB as well), and see if it makes a difference.

I'm using the Windows Memtest but I limit the memory it can test to avoid it getting mucked about by Windows's memory management too much :)
 
Yeah - sounded like it - use the bootable one - far more reliable, and takes the cpu (mostly) out of the equation!)

(actually the explanation was plausible at the time, something about the way the memory was handled and the i/o rates meant that the drives were more prone to errors... oh well, one day I'll track it down again!)

What voltage are you using on the XP-M?
 
Well, I found that when the windows-based one gives errors the bootable floppy will invariably give errors too, so I thought meh. If I get it to the point where Memtest finds no errors and I want to bulletproof-test it then I'll use Memtest86 for the extra reliability. CPU is at 1.75V atm, but MBM5 reports only 1.68, so either it's being undereported or undervolted. Will bump it up further if needed, doesn't get hotter than 50 degrees atm.
 
Stick with the readings in the Bios - never entirely happy with CPU-z etc for accuracy - if you've got a multimeter handy...

Have a look Here

Very useful stuff - though you've definitely got my favourite bios in there already - could have links on there that will help tweak your bios settings to squeeze that little bit extra out of it.

If you're still unsure about the memory, you can always go back to basics, set the chip to something like 200*10 and do some testing, seperate out the components - nothing worse than a duff stick - I nearly went mad when my Geil gave up the ghost, only spotted it when I lent it to a mate and it fell over in basic testing :mad:

Also - I found that one of the Ram slots gave up on my board - pretty insidious, just a 'bit' unstable, so worth maybe doing 1 stick at a time and trying all the slots - might help.

(Oh, and I wouldn't be afraid to squeeze that voltage up to about 1.8 as long as the temp doesn't get much above 55C)
 
You must log in or register to reply here.
Back
Top Bottom