Offler

So the summer returned here for some days and the OCZs failed the test... Will see if i can return them, else I have to sell them.

At all the nanyas are rock solid stable since last testing, however not much overclockable. At all it raises my interest in 1.5v dimms for X48. Right now i try to re-test the CPU because of possible lower voltages and then i try to OC the board a bit.

Offler

Soo... If the Ch1Ch2 CommonClock Setting is set to "nominal" while the ChX Clock Crossing Setting is on Auto i get BSODs regarding "Memory Management" sometimes...

Offler

so the OCZ rams have been sucessfully tuned for the board. Took half of year to suceed.

That bios is killer one

However - I want to upgrade in some time to 16gb ram. Currently possible with 4x 4gb, 2x 8gb will be better.

Any reccomendations?

Offler

Ok lets talk a bit about disk adapter solution on this board.

Its ICH9R + some JMicron stuff. 4x SATA II by intel, 2x SATAII by JMicron, 2x ATA by JMicron.

PCI-E Spread Spectrum test = The JMicron Chip fails when the spread spectrum is on. Internal Intel SATA adapter works fine.

Whats wrong with this solution? Well the Intel ICH9R is connected via single PCI-E 1x Version 1.0 lane to the system. This allows 250 mb transfer from the system to the adapter.

Each SATA II connector allows up to 300mb/s (250 in real) and you have four of them. Once you connect good disk drives to all connectors you will notice that this connection slows down data transfer if the system needs more than one drive at the same time.

Therefore I am currently testing Adaptec 1430SA Raid controller, but not as a Raid just as a SATA adapter. It has PCI-E 4x connection which allows up to 1000mb/s. This configuration allows me to connect 4 disk drives or other SATA II devices and each of it can work without waiting for bus.

The adapter uses Marvell 88sx7042 ... and it seems it has also failed at PCI-E spread spectrum on...

Offler

Yesterday I got some time to OC at last. I hit 430Mhz without any major problem. memory was set to 400/1600, I just set CAS from 7 to 8. OCing was easy on this board for the very first time and the limit I usually hit was bit higher. Howerver at 435 the journey ended this time.

Not sure what happened there - maybe the memory hit its limit (1740 Mhz), so I will try to start again with 400/800 setting. At all since fixing the ram i have no occasion to re-test the CPU stability on bit lower voltages, so its worth a try.

Also the problem with new Sata disk adapter was confirmed - not as good as I expected.

At all this OC is still far from my expectations.

Offler

Even next attempts to overclock finished at 430Mhz FSB. Thats bad.

More interesting seems other testing I did with memory.

FSB400

LGA775 can acess to memory with 64bit data width, at 400MHz Quad data rate - 1600Mhz effective. Theoretical bandwidth is then 12,8Gb/s

Memory at 1600Mhz Cl7 Dual Channel - Bandwidth 25,6 Gb/s
Read 9400Mb/s
Write 8400mb/s

Memory at 800Mhz Cl7 Dual Channer - Bandwidth 12,8 Gb/s
Read 6400mb/s
Write 8400Mb/s

(the read values were varying - seems dependant on the DRAM data strenght values)

Memory at 1600 Mhz CL7, Single channel - Bandwidth 12,8Gb/s
Read 9400Mb/s
Write 8400mb/s

That leads me back to the questions:
1. If memory bandwidth is larger than CPU bottleneck can other devices such as PCI-E cards acess memory independently on CPU? Therefore the rest of bandwithh can be used for hi-speed SSD PCI-E cards without slowing down the CPU.

2. Single channel mode use the bandwidth more effectively so the question is how can this help to overclock.

eva2000

Not sure about #1, depends on architecture and how cpu communicates with PCI-E. Would differ from LGA775 vs LGA1366 etc

#2, for LGA775 held back by NB and mem controller for oc'ing, so yes less memory used/populated = less stress on NB/mem controller = potentially better oc'ing of cpu

Offler

Well the OC attempt at this point ended at 415Mhz. Stability was very poor and nothing helped to stabillize it.

There was also another thing which is complete mystery from me. I keep DImm Fine delays on board automatics set to "more aggresive". After each power unplug the board generates new set of them.

With one specific setting I was able to run the memory on Command rate 1. After cold reboot the setting was gone, and system was unable to boot

Have to check yours values.