Overclocked system with 3 or 4 DIMM slots filled?

[Coki]

Are there any stability problems with 3 or 4 dimm slots filled with memory while running at 133 or 140FSB?

Thanks.

 

[utopia]

if your memory can take it, no..

 

[Coki]

The memory could take it I just remember reading something a while back saying that with 3 or 4 dimms filled system becomes a bit unstable.

 

[Plester]

depends on the mainboard - MANY do have problems o'ced w/ multiple dimms. check bxboards (i think) latest ram roundup - often when a dimm is in the last slot, a system won't overclock nearly as high as when a stick is in slot 1 or 2.

 

[IceStorm]

Depends on the chipset.

440BX systems don't run well at 133FSB with all four DIMM slots filled. If anyone's actually running a 440BX at 133 with all four DIMM slots filled, please let us know what DIMMs and what board. The problem isn't the RAM, it's the chipset.

The best I can do is 124FSB with my P3B-F, all slots filled. If I push it to 133, I have to drop to filling only the first and third DIMM slots.

The i815/E only handles two DIMMs when over 100FSB, as per Intel's documents on the chipset. I've gone up to 153FSB (the max on the SE6) with two DIMMs, but I had to retard the RAM speed by 33Mhz (120Mhz). It did 140FSB no problem with timings of 2-2-2-6/8. The board could never do timings of 2-2-2-5/7 unless I was at 130FSB or below. 133FSB, no good on 2-2-2-5/7 (this is CAS2 Crucial PC133 non-ECC RAM, mind you).

The P3V4X took four 128MB ECC PC133 DIMMs to 146FSB for me with no problems. The DIMMs were at CAS3.

I did try my SE6 with a 256MB ECC and non-ECC DIMM, both rated for CAS3. They could do 133, and 137, but nothing higher. I was a bit surprised to find two 128's doing better than one 256, but it could be the board - my SE6 is in the process of being RMAed, as it has flaked out (spontaneous reboots, RAM count in POST doesn't work, etc).

Just my $0.02

 

[Klosters]

All four of the Dimm slots are filled on my P3B-F. All four are 64Mb Samsung PC-133 CAS3 Dimms, and I'm running a ca2 PIII 550 FC-PGA at 133 Mhz FSB, for 733 Mhz Clock Speed. The Dimms aren't as identical as Mushkin led me to believe, at least according to a "shootout" I saw today. Three of the Dimms are "GA," one is "CT" or "TC" "-75." According to the shootout, the ~TC-75 won't run beyond ~134 Mhz FSB, even at CAS3. Perhaps that's true with the Soyo mainboard used in the review, but it runs fine at 133 Mhz CAS2 on the P3B-F.

I think. My new rig may not be 100% stable. I get occasional crashes while logging on to Juno. I've had 2 X Windows snap shut while online in Linux. I can't remember if the Juno thing happened with my old box, but an X Window spontaneously snapping shut is a quite rare event in Linux.

Heat could be the problem. I have a Delta 38 cfm 60mm fan on a little 40mm Vantec Si5 HS and I used Arctic Silver and a copper shim. The real estate available between the Iwill Slocket II and the first Dimm slot is tight. My guess is that a 50mm HS might fit. But would a GW "-12", "16" or "-32" HS be even a 1C improvement? Doubt it. 47 cfm 80mm's are on the way, to replace the no-name ~32 cfm case fans I have now. Rounding the IDE and floppy ribbons reduced mobo temp by ~2C, and CPU temp by 1C.

The rub is that I already bought a PAL35t, but decided against the CUBX. I'm reluctant to hang 7.2 ounces on the Slocket II, and I hate to lose 2 Dimm slots as well. Right now, MBM reads 30C mainboard and 42C CPU. The room temp is 26C. The CPU temp could be as low as ~39C with the Alpha at the moment. I'm forming the opinion that rapid heat removal might be a lot more important with the PIIIE's/CII's than it was with their good deal hotter predecessors. It's always something.

 

[IceStorm]

How many chips are on the 64MB DIMMs?

The max chip load per DIMM, unregistered, is 18. If you're using, say, 8 chips, that could explain why you can get away with all four DIMMs at 133 (less load). For me, I would "lose" RAM banks on boot when I had all four filled at 133, but they were all 18 chip PC133 ECC 128MB DIMMs.

You may want to try reducing one of the CAS timing values to 3, or increasing the cycle time to 10. That should make the box more stable, and not sacrifice too much in performance.

 

[JellyBaby]

The old rules of thumb (rule of thumbs?):

- Go with the smallest possible amount of RAM per DIMM (e.g. a 128 MB module is better than a 256 MB module).
- Go with the fewest number of chips per DIMM you can.
- Make sure the number of chips per DIMM is the same across all DIMMs.

This checklist has served me well for many years. I'm sure you can break one or more of these guidelines and live to tell about it. But I won't unless I have to!

 

[Klosters]

Thanks for the replies! They're all single sided 64 Mb Dimms, and they surely aren't ECC. Setting the Bios to 2-2-2-10 may have a destabilising effect, but without careful study it doesn't seem to matter if the Ram Bios settings are those, or more conservative 3-3-3-10.

Can anyone tell me if disabling L2 cache ECC may help, please? If so, how does this effect performance? What if I remove the first Dimm(in Slot 0) that's closest to the CPU? It would make room for an Alpha PEP66. Ugh! I already bought a PAL35t with a monstro Delta fan, but that kills two Dimm slots. 128 Mb? Whatever will the neighbor's say? jk.