Issues with running at 1:1

[imported_goku]

Bleh, this must be the third thread pertaining to this.. Ok, If I run my memory modules at 1:1 ratio and I get errors at 3.5ghz, or 250mhz memory/fsb, in memtest 86 but when I run at 5:4 ratio I don't get any errors, is it a safe bet that my memory is at fault? I went through a huge hassel of RMAing my motherboard twice because the first RMA board would error with these modules at 230mhz+, now it appears they'll error at 240mhz+. I'm using 2X1GB Mushkin Redline XP4000 memory modules which are RATED at 250mhz.

So, could this be a CPU issue or board issue or is it really just the memory modules? Oh and BTW, these memory modules are also the second pair I've recieved from an RMA, I asked the guy to test the modules before shipping them (mushkin) and they said they tested fine at 275mhz on their A64 platform.

:insert sigh emoticon:

 

[cmdrdredd]

Your board is crapping out. Also try more voltage on the memory.

 

[imported_goku]

Running at 2.8V which is the max for the IC7-G.... I'm finding this experience to be extremely frustrating...

 

[cmdrdredd]

Originally posted by: goku
Running at 2.8V which is the max for the IC7-G.... I'm finding this experience to be extremely frustrating...

You've found the limit for that particular chipset I bet. Nothing you can do about it.

 

[imported_goku]

Well what I can tell you is that there are others out there with my board that are running with my processor at 272mhz 1:1 ratio. Also with the P4 2.4C they're running at 300mhz FSB. Then there was this one guy who had a unlocked P4 and managed to get the board up to 400mhz!!! :shocked:

 

[Tweakin]

That's the real kicker with OC'ing...there are never two "like" systems...I figure if I can get anything above stock I'm ahead of the game...

 

[Lonyo]

You might need more volts, and if 2.8v is your limit, then the board may be the issue (not able to supply enough juice).
What's the timings of the RAM and what volate does it say it should have?
There are chips which need 3+v to hit higher speeds.


http://www.anandtech.com/showdoc.aspx?i=2416&p=2

Mushkin Redline XP4000 Memory Specifications
Rated Timings 2-2-2-6 at DDR500 at 3.3-3.5V
ACTIVE COOLING REQUIRED

3.3v and you are trying with 2.8v which is almost certainly where the problem is.

 

[cmdrdredd]

Originally posted by: Lonyo
You might need more volts, and if 2.8v is your limit, then the board may be the issue (not able to supply enough juice).
What's the timings of the RAM and what volate does it say it should have?
There are chips which need 3+v to hit higher speeds.


http://www.anandtech.com/showdoc.aspx?i=2416&p=2

Mushkin Redline XP4000 Memory Specifications
Rated Timings 2-2-2-6 at DDR500 at 3.3-3.5V
ACTIVE COOLING REQUIRED

3.3v and you are trying with 2.8v which is almost certainly where the problem is.

Good find...Like I said, limitation of the board and perhaps in this case, the memory.

 

[imported_goku]

Originally posted by: cmdrdredd

Originally posted by: Lonyo
You might need more volts, and if 2.8v is your limit, then the board may be the issue (not able to supply enough juice).
What's the timings of the RAM and what volate does it say it should have?
There are chips which need 3+v to hit higher speeds.


http://www.anandtech.com/showdoc.aspx?i=2416&p=2

Mushkin Redline XP4000 Memory Specifications
Rated Timings 2-2-2-6 at DDR500 at 3.3-3.5V
ACTIVE COOLING REQUIRED

3.3v and you are trying with 2.8v which is almost certainly where the problem is.

Good find...Like I said, limitation of the board and perhaps in this case, the memory.

This happens every time.. Sigh, these are 2X1GB memory modules, what lonyo found is for the 512MB memory modules. The 1GB memory modules voltage range is 2.6v-2.9v and don't require active cooling.

 

[cmdrdredd]

Originally posted by: goku

Originally posted by: cmdrdredd

Originally posted by: Lonyo
You might need more volts, and if 2.8v is your limit, then the board may be the issue (not able to supply enough juice).
What's the timings of the RAM and what volate does it say it should have?
There are chips which need 3+v to hit higher speeds.


http://www.anandtech.com/showdoc.aspx?i=2416&p=2

Mushkin Redline XP4000 Memory Specifications
Rated Timings 2-2-2-6 at DDR500 at 3.3-3.5V
ACTIVE COOLING REQUIRED

3.3v and you are trying with 2.8v which is almost certainly where the problem is.

Good find...Like I said, limitation of the board and perhaps in this case, the memory.

This happens every time.. Sigh, these are 2X1GB memory modules, what lonyo found is for the 512MB memory modules. The 1GB memory modules voltage range is 2.6v-2.9v and don't require active cooling.

Then figure it out yourself :roll:

Seriously...it's OBVIOUS you can't get more out of it based on your config.

 

[Duvie]

My IC7-G I did 292fsb...Actually had 2 different boards and they both did about270+....

You could carefuklly remopve the NB chipset and apply some AS and maybe a better fan. If the mobo is limiting you then it is likely the NB chipset and heat...

What timings have you tried?

 

[imported_goku]

Tried SPD and 3-4-4-8. Hmm.. well after 4 trials (run like 2-14 passes then fails in memtest 86) of different settings which includes: AGP voltage increase, relaxed AGP timings, relaxed chipset timings I'm still getting errors. I just increased the voltage of the CPU from 1.525/1.5500 to 1.6 and it errored after 4 passes which is unusally fast I guess... So I've increase the voltage again to 1.65v to see if it will error quicker... Man this is SERIOUSLY got to be goat rodeo...

 

[Duvie]

I remember my IC7 mobos undervolted everything....The vidmm set to 2.8v would not deliver 2.8v...It was less...

What PSU do you have? How do the 12v and 3.3v rails look?

 

[imported_goku]

After the increase to 1.65v, it errored after a first pass, k.... This is probably the first IC7-G board that I've seen that actually overvolts the memory. My memory is at 2.83v, 5v is 4.97 but multimeter says slightly over 5v... 12V is like 12.05, 3.3v is 3.28v. My first IC7-G used to be 2.75v for the 2.8v selection.

The one thing that I've been split on is the issue with the mosfets...


The mosfets above the CPU and below the AGP card are of concern to me. I know for sure that when I increase the voltage of my board, a few of the mosfets (transistors) get very VERY hot like 80C iirc and I've heard that the mosfets below the AGP port should have some sort of sink on them. Now I'm not sure if I should go and get some sinks for them or not. I mean I planned on doing so anyways but what if it's just the board?



Also the first RMA board I had would limit overclock to 225mhz memory, error at 230 and above, this one seems to limit at 240 and above. If I put sinks on the mosfets and it doesn't fix anything, don't know if I can still RMA the board.

 

[Duvie]

Originally posted by: goku
After the increase to 1.65v, it errored after a first pass, k.... This is probably the first IC7-G board that I've seen that actually overvolts the memory. My memory is at 2.83v, 5v is 4.97 but multimeter says slightly over 5v... 12V is like 12.05, 3.3v is 3.28v. My first IC7-G used to be 2.75v for the 2.8v selection.

Yeah that was like mine....I think it did like 2.77-2.78v...My 3.3v was always low like that as well 3.28v....

I did put an aftermarket HSF on the NB chipset and I was able to OC a bit further but at stock it did do over 270 no problem...

OK lets look at this!!!
increased the vcore and it failed quicker....Why would it do this?

1) the added vcore increased the heat? How are the temps?

or

2) the power supply cannot deliver the power....


Have you tried lowering the vcore 1.5v and running it???

Also memtest does check the cache so the issue could be with the cpu....Isn't there an options to not test the cache? Are you running the Dos bootup version?


It was a good idea to raise the vagp...In these older boards there wasn't a chipset voltage controller and many speculated the vagp voltage controlled the NB chipset as well.

 

[imported_goku]

Well the CPU temp is another thing I've been speculating about. PWM temps are hovering around 90F-100F, before I tell you the CPU temps I have to let you know that one thing I noticed that when I got the RMA boards back is that the CPU temps seem to be significantly more acccurate than before..

I'm running my CPU fan at max speed and I'm getting 50C temps, thing is though, I can only find out the temps until AFTER I reboot and by then, the temps have cooled a bit. As soon as I enter the bios the temps seemingly settling from a higher number as I see it go from 51C to 50 and then down to 47 depending on the voltage.

ATM I'm testing at 272mhz FSB with 5:4 divider, IIRC it's 1.575, 3-4-4-8, 2.8v and it's on pass 11. (I'm testing with test #5). One problem with all of my testing though is that sometimes it seems like my system is stable (I thought it was at 250mhz until I was doing my final test to make sure it was) and then bam I saw errors.

One positive I guess is that I see usually a pair of errors this time, with my first RMA board I would see like 30 errors at I time. If I want to see lots of errors at a time, all I have to do is go to 272mhz 1:1, increase the voltage and I'll get a lot of errors opposed to the situtation I'm in right now where I get just a pair of errors.


Man I wish there was a sigh emoticon.

I was wondering if this was important or not but the 'walltime' in memtes 86 the time seems a bit irregular with it 'speeding up' quite often. I'm running memtest 86 v3.2 from the ultimate bootcd.

 

[Duvie]

I think the temps may be a tad high....You could try to run it using the windows version and use speedfan to monitor temps....mY idles temps for my p4 werein the high 30's and often running 2 instances of prime95 (much more taxing then memtest) would only hit 52c....however the northwoods were usually quite good up to mid 60's...

One thing to remember is the PWM temp is not the chipset temp...It is a sensor by the power caps....The NB chipset could still be the culprit here...I would suggest removing (carefully) the chipset hsf...clean off stock goo and apply some AS and reapply HSF....that may help if the temp issue is related to the NB chipset. the NB chipset does control the interface between the memory and the CPU.

 

[imported_goku]

So you think the chipset is getting too hot? One problem though is that if I let the system cool, I may effect the variables that I've been trying to look atm. Just tried the memtest 86+ version at the same settings/speed that I was just testing in my last post, well now it errors immediately entering the test so i guess thats a good thing... There is ONE more thing I can think of, every time I try setting the HSF on the processor, IIRC I accidentally 'pop' it up on one side after setting the HSF on the thermal paste before getting both clips to set the sink down.

 

[Duvie]

Originally posted by: goku
So you think the chipset is getting too hot? One problem though is that if I let the system cool, I may effect the variables that I've been trying to look atm. Just tried the memtest 86+ version at the same settings/speed that I was just testing in my last post, well now it errors immediately entering the test so i guess thats a good thing... There is ONE more thing I can think of, every time I try setting the HSF on the processor, IIRC I accidentally 'pop' it up after setting the HSF on the thermal paste before getting both clips to set the sink down.

Shouldn't be an issue.....

With the P4s and their integrated heat spreader your main concern is more about not overtightening the heatsink (usually with aftermarkt bolt down types) causing the board to warp and not sit flush on the IHS...or...over or under application of thermal paste...too much is just as bad as too little....


Trust me try the simple NB chipset reapplication and see if that helps...then work on cpu temps....heck just aim a big fan in the case at the cpu/NB and see if that helps...

 

[imported_goku]

I've got a fan blow at the system ATM, not sure if it's goanna help but not sure how that would help either as that wouldn't be practical. What would it mean if this does help? I don't think the chipset gets very hot on these boards...

Tried the fan, doesn't work, going to try again but it doesn't look like the big fan will help.

 

[imported_goku]

Bump, I've changed the thermal paste which is a PITA for these chipsets because it's this weird garbage that is hard get off. One problem I have with applying thermal paste is that I worry I put too much on or I accidentally create air pockets...

 

[Duvie]

Originally posted by: goku
I've got a fan blow at the system ATM, not sure if it's goanna help but not sure how that would help either as that wouldn't be practical. What would it mean if this does help? I don't think the chipset gets very hot on these boards...

Tried the fan, doesn't work, going to try again but it doesn't look like the big fan will help.

Yeah they do...especially when you are running them 92mhz or nearly 50% over stock....I did run my vagp at 1.6v and found it too help me when I was ocing the fsb,....i know you tried it to NO avail but keep track as the answer may be in conjunction with many of the the things you have tried.....


Also a HSF not being hot isn't a clear sign...Sometimes poorly installed heatsinks inhibit heat transfer and thus the heat i strapped at the core of the chipset and the HSF does not get hot..

Most successfaul ocers of IC7 mobo did this to their NB chipset...

Also what NB strap are you using? List all your bios settings....I remember there was some tricks to get past the 250 fsb wall that existed for some...

 

[Duvie]

Originally posted by: goku
Bump, I've changed the thermal paste which is a PITA for these chipsets because it's this weird garbage that is hard get off. One problem I have with applying thermal paste is that I worry I put too much on or I accidentally create air pockets...

If you are using AS just a very small dab and then take like a cridet card and smooth it out until it is very thin covering all of the glass raised core....

The crap that was there was tough t get off but one look at it makes sense as to why many replaced it and got better results

 

[imported_goku]

Originally posted by: Duvie

Originally posted by: goku
I've got a fan blow at the system ATM, not sure if it's goanna help but not sure how that would help either as that wouldn't be practical. What would it mean if this does help? I don't think the chipset gets very hot on these boards...

Tried the fan, doesn't work, going to try again but it doesn't look like the big fan will help.

Yeah they do...especially when you are running them 92mhz or nearly 50% over stock....I did run my vagp at 1.6v and found it too help me when I was ocing the fsb,....i know you tried it to NO avail but keep track as the answer may be in conjunction with many of the the things you have tried.....


Also a HSF not being hot isn't a clear sign...Sometimes poorly installed heatsinks inhibit heat transfer and thus the heat i strapped at the core of the chipset and the HSF does not get hot..

Most successfaul ocers of IC7 mobo did this to their NB chipset...

Also what NB strap are you using? List all your bios settings....I remember there was some tricks to get past the 250 fsb wall that existed for some...

I'm not sure how the NBstrap would effect anything. Also the Flattening out thing with the credit card confuses me as well as I thought all that was necessary was to paste it down, the pressure from the HSF would just level it out. On the guide for the PIII/Athlon XP they say flatten it out as you said but for the P4/A64 with IHS they say leave a half grain of rice sized bit of thermal paste and thats it.

Well I've changed the thermal paste but I have no idea how well I added in the thermal paste. The system works fine at 1:1 at 250mhz but if I 'heat it up' by raising the FSB to 272 and increasing the vcore, running some tests and then backing it back down to 250mhz, it will begin to error. Not sure how long it would take if I would let it 'warm up' otherwise..

 

[myocardia]

Interpretation of Duvie's last post:
For the NB chipset, "if you are using AS just use a very small dab and then take like a credit card and smooth it out until it is very thin covering all of the glass raised core...."

And for the cpu, just follow the instructions here: link to Arctic Silver's instruction page.