BIOS recommendations for GA-EP35C-DS3R

[sdog]

Memory: G.Skill F2-C8500CL5D-4GBPK
2 sticks (2x2G=4G) recently purchased
2 sticks (2x2G=4G) purchased for original machine build ~4 years ago.
DDR2-1066 CL-5-5-15 2.0-2.1v

Mobo: GA-EP35C-DS3R
Bios=v4

CPU: Intel Q6600 B3

Couldn't find my model on this forum's MOBO recommendation list.

Thought it would be simple to add another 4G of memory, particularly since all specifications on the motherboard and on the memory seem to allow this and all memory was the same brand and model, but no love.

Original symptoms were random blue screens and reboots as soon as I installed the memory. Now trying to troubleshoot and get a clean pass through memtest86+ before going back into the OS.

Any two sticks in the appropriate slots pass Memtest86+, but as soon as I put all four sticks in, Memtest86+ pukes out (usually somewhere around test 5 or test 8).

Original BIOS was using no overclocking and a lot of auto settings for the various bus frequencies, voltages, and DRAM timings. Since then I've tried all kinds of variations, including boosting the mem chip voltage a couple points, loosening the memory timings (but unfortunately having no idea "how loose" to make "which timings").

So far the only thing that has gotten me a clean pass with all 4 sticks through MemTest86+ has been to drop the memory bus multiplier from 4 to 3, which drops the the memory speed from 1066MHz to 800MHz. If this is the best I can do, then so be it, but it offends my sensibilities to have to run memory rated at 1066MHz at 800MHz.

I'm relatively tech savy, but I'm not a hardware geek or an overclocker, so I thought before I try every combination of memory timing settings known to man, maybe I'd check here and see if anyone has any familiarity with this board.

Any particular settings I should be using or trying in order to get this memory to play nice with this motherboard?

I'm presently testing Memtest86+ with a slight FSB overclock (300 rather than 266) with the memory multiplier set to 3 instead of 4, auto DRAM timings, and auto voltage settings. This brings the memory bus up to 900MHz. Memtest86+ shows the auto CAS settings as 5-6-6-17. So far it's made it about 80% of the way through the test without failure, so maybe I can push the memory back up close to 1066 with overclocking. And maybe it's good that I'm being forced grudgingly into the world of overclockers, BUT... it raises the question:

IF I am ultimately not able to simply plug in my 8G of 1066MHz memory on my 8G/1066MHz MOBO and have it run 8G at 1066MHz, what is the best approach to make the system as fast as I can?

Should I keep pushing the boundaries of FSB clock while dropping the mem bus multiplier while keeping timing cycles short? Or is it better to keep the memory bus speed up and loosen the timings until things work? Or something completely different?

Thanks in advance for any insight you can offer.

[Tradesman]

Biggest thing will be to bump the NB voltage a tad to help the MC (in the NB) handle 4 sticks, also what's your tRFC (memory advanced timing) at around 84. With your slight OC may want the vCore up just a bit also, possibly add .06 to current vCore. Ensure CR (Command Rate) at 2T

[GSKILL TECH]

DDR2-800 4-4-4-12 2.0V

This will perform roughly the same, it can also be used as a test to determine whether the modules can work together in a stable manner.

For DDR2-1066,

5-5-5-15 2.10V

NB Voltage 1.40V

tRFC 88

tWR 8

If problems persist, post some pictures and we can see what's going on.

Thank you
GSKILL TECH

[sdog]

Thanks T-Man and G.Skill Tech.

I need just a bit of help translating.

I DON'T have access to the following on this MOBO (or I'm just too stupid to see them):

tWR (write recovery)
tCR (command rate)
tRFC (row refresh cycle time)

I DO have access to these (BIOS v. F4):

Standard:
CAS Latency Time
DRAM RAS# to CAS# Delay
DRAM RAS# Precharge
Precharge Delay (tRAS)

Advanced:
ACT to ACT Delay (tRRD)
Rank Write to Read Delay
Write to Precharge Delay
Refresh to ACT Delay
Read to Precharge Delay
Static tRead Value
Static tRead Phase Adjust

Do any of the timing options I have access to relate to the timings you suggested?

For my NB setting ((G)MCH Overvoltage control), I don't know what the actual voltage is, but I can bump it from "whatever it is" + 0.1, +0.2, or +0.3V. Is the starting point for the north bridge voltage the same as the core voltage?

Sorry I don't know the terminology and technology just a bit better.

Thanks.

[Tradesman]

tRFC = Refresh to ACT Delay tWR = Write to Precharge Delay tCR is normally just standalone as COMMAND RATE or COMMAND TIMING something to that effect and will be either 1 or 2 with a T or N designator like 1T, 2T, 1N or 2N will want the 2 N or T

For NB Voltage go +.1 and it should then show what it is actually at and can adjust from there

[sdog]

O.K. Thanks Guys. I've gotten one full pass through memtest86+ for each of 2 different configurations based on your parameter setting suggestions.

once using the slower speed (800) with faster timings (4-4-4-12)

and once using a faster speed (not quite 1066, but 1000 based on my odd combination of other bus multipliers) with slower timings (5-5-5-15)

I already had my various voltages at or a point or two hotter than your suggestions, so I'll slowly lower all the non-NB voltages a point or two and do some longer stress tests to see if everything stays stable as those come down.

NEVER MIND!

Just walked over to the machine again and it's got errors on the second pass for the 1000MHz, 5-5-5-15 configuration.

I'm going back to the 800 4-4-4-12 config and let it run more than 1 pass and see what happens.

[sdog]

G.Skill Tech, are you saying that, in terms of overall memory speed, the slower mem speed with the shorter cycles (800 4-4-4-12) will be roughly as fast as the faster speed with more cycles (1066 5-5-5-15)? Or did you mean something different by "perform roughly the same"?

If the overall speed is similar, and I can get several passes with the 800MHz configuration, maybe I can just stop and be happy.

[Tradesman]

Yes, and it is roughly, the higher freq provides more bandwidth but the lower CL (CAS Latency) means less clock cycles before doing a task....so....roughly 800 at CL4 is just a little less performance wise than 1066/CL5 which is slightly less than 1333/CL6 etc

[sdog]

O.K. At this point I have a working memory configuration w/ basic timings at 4-4-4-12. Overnight (7 clean passes) on memtest86+.

## FREQUENCIES :
FSB freq=333MHz
CPU Mult=9 (3.0GHz)
Memory Multiplier 2.4 (800MHz)

## TIMINGS:
Basic timings = 4-4-4-12

Advanced timings all on auto except:

Write to Precharge Delay = 8
Refresh to ACT Delay 88

## VOLTAGES:
DDR2/DDR3 Overvoltage Control = +0.3V (have no idea in relationship to what, but on other BIOS screen, memory voltage shows at 2.144V)

FSB Overvoltage Control = +0.2V (Have no idea what the actual voltage is because I have no idea what this overvoltage is in relation to and the voltage itself doesn't show up anywhere is in the BIOS. The only two actual voltages shown are VCore and DDR2 Voltage)

(G)MCH Overvoltage Control = +0.3V (Have no idea what the actual voltage is because I have no idea what this overvoltage is in relation to and the voltage itself doesn't show up anywhere is in the BIOS. The only two actual voltages shown are VCore and DDR2 Voltage)

CPU Voltage Control = 1.5V
__________________________________________________ ___________________

My preference would have been to run the memory at 1066 5-5-5-15 since in theory both memory and board should be able to do that. But... Since it seems I can't, and since I've started to learn a little bit about overclocking, I'd like to see if I can make it as fast as I can from the 800CL4 direction.

My plan from here would be to nudge memory speed up, timings down, or voltages down, one parameter at a time, little values at a time, with an overnight memtest or Prime95 test between each change, until it goes unstable, then back the changes off a notch.

Since everything seems to be running fairly cool, even with long Prime95 tests, I think I lean toward dropping voltages last and starting with speed tweaks first.

Anything I've said so far sound like a bad idea?

If it were you trying to push this memory a little closer to spec on this MOBO, would you start by pushing up the memory speed (by playing with FSB speed and multipliers) or by dropping some of the timing cycles I've listed? If so which ones? Maybe "Write to Precharge Delay" and "Refresh to ACT Delay" since both were bumped up a little higher earlier?

Finally, one more possibility. Would it be possible to run at 1066 CL6? If so, how would the overall speed of 1066 CL6 compare to the performance of 800 CL4? If faster, what would you recommend as a starting place for the advanced timings for 1066 CL6? Or are the differences between the two too close to bother with?

Sorry if it seems like I'm asking overclocker questions in a memory forum, but I'm really just trying to understand which BIOS settings have what effect on the speed and stability of the memory since I wasn't able to just plug it in and go.

Thanks.

[GSKILL TECH]

Try

FSB freq=375MHz for DDR2-900

Did you increase MCH Core Voltage?

Download a memory benchmark test, then you can see exactly which configuration performs better, and in what way.

Thank you
GSKILL TECH