2sets of TWIN2X4096-6400C5DHX (8GB) on Asus Striker Extreme

Edit: Sorry was wrong thread title! It's 2x TWIN2X4096-6400C5DHX actually

(QUAD2X4096-6400C5DHX is the 4GB set with 4 modules of 1G each)

Hi there,

I'm using 8GB (2 pairs) of TWIN2X4096-6400C5DHX on an

Asus Striker Extreme. (64bit OS)

What are the suggested DRAM settings for this particular setup?

I did some research in the forums and am using these settings for now:

(The Striker Extreme gives a ****load of possible settings and I'm not really into it)

tCL: 5T

tRCD: 5T

tRP: 5T

tRAS: 15T

Command Per Clock (CMD): 2 clock

These are the timing-related settings I've found in the forum so far.

(See http://asktheramguy.com/v3/showthread.php?t=52351)

This is the voltage I've set manually:

DDR2 SDRAM Voltage: 1.9

BIOS is up-to-date (1305), legacy USB is disabled, the CPU is a QX6850, chipset nv 680i.

Edit: Skip this and continue with post #8 since I've used the old memtest which doesn't know more recent chipsets...

Thanks for your reply.

 

In addition to the above settings, have tested it with

 

NB Core Voltage: 1.25 V

MEM (DDR), Mhz: 667

 

Still hangs at memtest. I'm sorry I did not notice, but now it always

hangs at:

 

Test #0: [Address test, walking ones, no cache]

Testing 4096M - 6144M 8191M

Pattern: 00000000

 

The ranges from 0-2056 and 2056-4096 went so damn fast I didn't even see them. This is the point when the complete machine immediately hangs.

 

More Ideas?

 

In addition to your suggestion, I've tried it with these, no luck, same position = hang.

 

 

NB Core Voltage: 1.25 V

MEM (DDR), Mhz: 667

Memory Voltage: 1.9 up to 2.1

tRAS: 18T

Sorry for that, I was using memtest86, not memtest86+.

I'll give memtest86+ a try immediately.

By the way, I have just tested both of the dual channel enabled pairs on its own (which is 4 GB each) and it works without errors for around 3 hours (~3-4 passes) no problem...

I'll update the thread as soon as I've run memtest86+ on both pairs at once.

Thanks for your help.

Did the test with memtest86+ v1.7 and 8 GB installed, exactly the same problem. As soon as more than 4 GB (>=50%) is addressed -> System hangs!

 

Test 50% ###########

Test #0: [Address test, walking ones]

Testing 4096M - 6144M 8191M

Pattern: 00000000

 

Please help me out :-(

Edit:

Damn Folks. Since both memtest86 and memtest86+ stalled at the same position,

I began to rethink every setting and the way I've installed / used Memtest*.

I have compiled both from source using gcc 4.2.2, so I tried using the precompiled

binary of memtest86+ booting from an USB stick.

And guess what? Absolutely stable, 3rd pass just running by with 8191M usable memory;

without any errors. Don't know which compiler they've used, however there seems to

be a problem somewhere when using gcc 4.2.2 on x86_64.

For anyone encountering similar problems on the Asus Striker Extreme,

here are the settings (notice the low voltage) this RAM runs really stable on:

EDIT: Here comes the Phenomen: just for 5 days! read on..

(the latest BIOS v.1305 seems to set the voltage to 2.15 V when set to AUTO)

Asus Striker Extreme + TWIN2X4096-6400C5DHX + Intel QX6850

====================================================

Timing:

 

tCL: 5T

tRCD: 5T

tRP: 5T

tRAS: 18T

Command Per Clock (CMD): 2 clock

tRRD: Auto

tRC: Auto

tWR: Auto

tWTR: Auto

tREF: Auto

Async Latency: Auto

 

 

Voltage:

 

VCore Voltage: Auto

Memory Voltage: 1.85 V

1.2V HT Voltage: Auto

NB Core Voltage: Auto

SB Core Voltage: Auto

CPU VTT Voltage: Auto

DDR2 Controller Ref Voltage: Auto

DDR2 Channel A Ref Voltage: Auto

DDR2 Channel B Ref Voltage: Auto

 

 

Other (perhaps you want to know?):

 

USB Legacy Support: Disabled

SLI Ready Memory: Disabled

 

FSB - Memory Clock Mode: Auto

FSB - Memory Ratio: Auto

FSB (QDR), Mhz: 1333

MEM (DDR), Mhz: 800

 

ECC stuff: Disabled

 

====================================================

Thanks Ramguy for your help. Sorry for the problems which actually

were no problems ;-)

Cheers!

I tried using the precompiled binary of memtest86+ booting from an USB stick.

It's always best to use a precompiled version. Glad to hear you are working fine and yes, that's a very good voltage for 800Mhz. :):

This was interesting reading, how do you get your Striker Extreme to boot from a USB stick ? All I see is removable, which is my floppy drive.

If I can run memtest off a USB stick would be a massive help, as the system hangs running it from a CD ROM (Not always but most of the time)

Did your system just hang, no BSOD, just literally freeze and only option was to power down ? And did it then hang again on POST ? DET DRAM in the LCD ? This could be the resolution im looking for. How stable is your system now ?

Although Biz is using 8 gig, im only on 4, am I ok to set those same values for the timings etc ?

I'll try it at 1.85 see if that helps, im currently using 1.9v

To boot from an USB drive with the A-SE you have to set the following:

In the boot priority list set Harddisk to the highest priority,

and the USB Stick / Drive is not under Removables but under Harddisk Drives along with real harddisks, you have to give it the higher priority.

I have another question, don't think I need another thread for it.

Just to get rid of every Auto-Detection this crappy Bios does,

what are the right settings for these advanced memory timings using the above-mentioned memory? I want to manually set everything memory related.

(If power fails, Asus' CRASHFREE Bios kicks in and sets everything to auto again,

so it won't boot anymore. This Bios just sucks...)

Menu > Extreme Tweaker > Overclocking > Memory Timing Setting

tCL (CAS Latency): 5

tRCD: 5

tRP: 5

tRAS: 15

Command Per Clock (CMD): 2

** Advanced Memory Settings **

tRRD: Auto

tRC: Auto

tWR: Auto

tWTR: Auto

tREF: Auto

tRD: Auto

tRFC: Auto

Async Latency: Auto

here is what I've found in the qualifying / testing thread, but its not all I need to know :-(

Testing parameters for TWIN2X4096-6400C5DHX:

Testing parameters for QUAD2X4096-6400C5DHX:

TWIN2X4096-6400C5DHX Data Sheet

QUAD2X4096-6400C5DHX Data Sheet

 

Motherboard make and model: Asus M2N32-SLI Premium/Deluxe

Each pair is tested and packaged together using the following settings:

AI Over Clock Tuner: Manual

CPU External Frequency: Manually set to 200MHz/(266Mhz for Intel)

DDR2 SDRAM Clock: Manually set to 800MHz

PCI Express Freq: Auto

PCI Freq: 33.33

Performance Mode: User Define/Manual

DRAM Timing Selectable: Manual

SDRAM CAS Latency: 5T

SDRAM RAS to CAS Delay (tRCD): 5T

SDRAM Row Precharge (tRP): 5T

SDRAM Active to Precharge Delay (tRAS): 15T

SDRAM Write Recovery Time: 5

Write to Precharge Delay: 15 (If it's Listed)

Command Rate: 2T

DDR2 SDRAM Voltage: 1.9

Advanced voltage control

CPU Vcore offset:_Disabled

DDR2 Termination: 0.5x

CPU to NB HT:____1.200V

NB to SB HT:_____AUTO

NB Core/PCI-E:___AUTO

SB Core/PCI-E:___AUTO

SB Standby core:_+1.500V

Output driver control

CKE:______AUTO

CS/ODT:__1.00X

ADD/CMD:_1.00X

MEMCLK:__1.00X

Chipset Core Voltage: Auto

All other settings are motherboard defaults.

Maximum DDR2 recommended VDIMM: 2.1 Volts

Please give me the remaining Auto settings :-)

These settings are mostly for high end overclocking to attempt to find stability on a system that is not reaching full stability.

I would not advise you changing these settings.

Ive spoken to Biz in PM about this, and thankfully got my USB stick to boot with memtest, although for some odd reason when I came to use it, I still used the CD :s Force of habit I guess.

Ive used the settings that were shown to me in a different thread, and im sure Biz will use these, would be great if this solved the issue, would be a great fix for the Striker Extreme.

Damn. It gets no end... I was really happy to have it running stable for nearly a week.

Please note this:

With the following settings, I have run memtest86+ v1.70 on

- each stick on its own: no errors

- each pair on its own: no errors

- both pairs at once: no erros

testing them after warm and coldstarts, the last test (full load) was done after a cold start for a whole night - stable - no errors at all.

BIOS VERSION 1305

 

FSB - Memory Clock Mode: Unlinked

x FSB - Memory Raio: Auto

FSB (QDR), Mhz: 1333

Actual FSB (QDR), Mhz: 1333

MEM (DDR), Mhz: 800

Actual MEM (DDR), Mhz: 800

 

Vcore Voltage: Auto

Memory Voltage:1,85 V

1.2v HT Voltage: Auto

NB Core Voltage: Auto

SB Core Voltage: Auto

CPU VTT Voltage: Auto

DDRII Controller Ref Voltage:auto

DDRII Channel A Ref Voltage:auto

DDRII Channel B Ref Voltage:auto

 

tCL (CAS Latency): 5

tRCD: 5

tRP: 5

tRAS: 18

Command Per Clock (CMD): 2 clock (2T)

 

Advance Memory Settings

tRRD: AUTO

tRC: AUTO

tWR: AUTO

tWTR: AUTO

tREF: AUTO

tRD: AUTO

tRFC: AUTO

Async Latency: AUTO

 

CPU Spread Spectrum: Disabled

PCIE Spread Spectrum: Disabled

MCP PCIE Spread Spectrum: Disabled

SATA Spread Spectrum: Disabled

LDT Spread Spectrum: Disabled

 

CPU Internal Thermal Control: Disabled

Limit CPUID MaxVal: Disabled

Enhanced C1 (C1E): Disabled

Execute Disable Bit: Disabled

Virtualization Technology: Disabled

Enhanced Intel Speedstep Tech: Disabled

LDT Frequency: 5x

 

PCIEX16_1 Frequency (Mhz): 100

PCIEX16_2 Frequency (Mhz): 100

PCIEX16_3 Frequency (Mhz): 100

SPP<->MCP Ref Clock, Mhz: Auto

 

USB Legacy Support: Disabled

So guess what?

I've mapped nearly the whole 8G within my OS and had no problems at all.

Just today, without even entering the bios for the last days and without adding or

removing ANY hardware - My kernel OOPses after around 6 hours of continuous work.

So to be sure its hardware related (and not an OS/Software issue),

I ran memtest86+ v1.70 again.... Errors ALL over the place just at test #1.

I'm still trying to get an answer for that.

To be sure its not the memory, I unmounted 3 sticks and then ran each one on

its own again. No errors at all.

Then the second test.. each matched pair on its own - No errors at all.

I mounted both pairs - 4 sticks - ran memtest - errors all over the place.

I did not change any BIOS setting since then; I'm still at the

above-mentioned and previously stable settings!

Seems like its this ****ed up memory controller!

If anyone wants to help me - here is what I've tried so far without success:

TRIED THIS WITH BIOS VERSION 1301, 1303, 1305

 

FSB - Memory Clock Mode: Unlinked

x FSB - Memory Raio: Auto

FSB (QDR), Mhz: 1333

Actual FSB (QDR), Mhz: 1333

MEM (DDR), Mhz: tried 667 and 800

Actual MEM (DDR), Mhz: ...

 

Vcore Voltage: Auto

Memory Voltage: tried 1.85V, 1.9V, 2V, 2.1V and Auto

1.2v HT Voltage: tried 1.25V, 1,35V and Auto

NB Core Voltage: tried 1,4V, 1,5V and Auto

SB Core Voltage: Auto

CPU VTT Voltage: tried 1,4V and Auto

DDRII Controller Ref Voltage: tried +10mv, +20mv, +30mv and Auto

DDRII Channel A Ref Voltage: tried +10mv, +20mv, +30mv and Auto

DDRII Channel B Ref Voltage: tried +10mv, +20mv, +30mv and Auto

 

tCL (CAS Latency): 5

tRCD: 5

tRP: 5

tRAS: tried 15 (as tested in xms qual - thread) but for most tests 18

Command Per Clock (CMD): 2 clock (2T)

 

Advance Memory Settings

tRRD: AUTO

tRC: AUTO

tWR: AUTO

tWTR: AUTO

tREF: AUTO

tRD: AUTO

tRFC: AUTO

Async Latency: AUTO

 

CPU Spread Spectrum: Disabled

PCIE Spread Spectrum: Disabled

MCP PCIE Spread Spectrum: Disabled

SATA Spread Spectrum: Disabled

LDT Spread Spectrum: Disabled

 

CPU Internal Thermal Control: Disabled

Limit CPUID MaxVal: Disabled

Enhanced C1 (C1E): Disabled

Execute Disable Bit: Disabled

Virtualization Technology: Disabled

Enhanced Intel Speedstep Tech: Disabled

LDT Frequency: 5x

 

PCIEX16_1 Frequency (Mhz): 100

PCIEX16_2 Frequency (Mhz): 100

PCIEX16_3 Frequency (Mhz): 100

SPP<->MCP Ref Clock, Mhz: tried 200, 250, most time Auto

 

USB Legacy Support: Disabled

I just want to share my experience with these:

Changing ANY Voltage except Vdimm from Auto to a manually set Voltage,

the system freezes randomly after a given time (even within bios!).

So you really have to pick the right Voltages (if set manually): too high - freeze - too low - freeze.

The exception to this rule is Vdimm.

You can "play" with it (within the RAMs specs) and all that changes is the time

errors appear within memtest and the memory range they appear in

(is this a direct link to which bank is addressed?)

I think the trick for successfully addressing 4 banks is:

Using the perfect combination of these:

Memory Voltage: tried 1.85V, 1.9V, 2V, 2.1V and Auto

NB Core Voltage: tried 1,4V, 1,5V and Auto

DDRII Controller Ref Voltage: tried +10mv, +20mv, +30mv and Auto

DDRII Channel A Ref Voltage: tried +10mv, +20mv, +30mv and Auto

DDRII Channel B Ref Voltage: tried +10mv, +20mv, +30mv and Auto

I thought about the problem beeing lost signals or a changing signal strenght

between the NB and each of the 4 banks.

The design in terms of distance of the banks and the controller seems to be:

NB -- A1 -- B1 -- A2 -- B2

To get a stable and constant signal strength from both

A1 and B2 compared to each other is quite tricky?!

I'm not really into that, but I've tried to realise that using the "Channel A/B Ref Voltage", eg.

Channel A Ref Voltage: +10mv (less distance)

Channel B Ref Voltage: +20mv (more distance)

Actually, this is noticable within memtest86+ and also... it's duplicable!

I can shift the errors appearing in whatever test# from a lower

memory range (channel A?!) to higher memory ranges (channel B?!)

Just for the test I've set the 2 voltages to extreme offsets, eg.

Channel A Ref Voltage: -30mv

Channel B Ref Voltage: +30mv

That gave me nearly all errors in the lower memory range.

Channel A Ref Voltage: +30mv

Channel B Ref Voltage: -30mv

That gave me nearly all errors in the higher memory range!

I still had no success finding the right combination.

And all that just to get the full dimm load running stable.

Actually I'm using 1 pair in everyday-use now and testing around with

full load until it works. (really sucks to have this much money lying around with no use..)

The last bit I've never tried to change actually are the advanced timing settings,

which you suggest to let them stay on Auto.

After all the research, this problem seems to be present with all

4 dimm solutions on various 680i boards.

Who has designed that? After all these years of experience.. its unbelievable nvidia

releases such a buggy chipset which is even unable to operate medium speed ram

(I don't think ddr2 667/800mhz is bleeding edge anymore)..

I'm checking the Asus ftp-site every hour to see if they've released a new Bios.

Did they even contact you, Corsair, in the long run of developing and bugfixing their BIOS?

Whatever... just signed the petition to get Nvidia wake up and work out a solution with all the MB makers. :-P

Back to topic, please share your ideas... :-)

After all the research, this problem seems to be present with all 4 dimm solutions on various 680i boards.

This issue is indeed problematic and VERY hard to isolate because you can take to identical boards and have one work with all four banks populated and another will just not make stability.

I personally go with Intel Chipsets for Intel CPUs. I have had no problems installing 8GB on the P35 chipset (P5K/P5K-e/Pk5-Deluxe) and this is my advice for those when they ask me the board for 8GB. There are issues, just not anywhere near as many as the 680i. The 680i is often far more wonky with high end DRAM and four banks filled. However, this is just my personal opinion. I would say that there might be some board overvolting of the DRAM and that you are getting a resonance of the DRAM strobe. The way to find out if this is the case is to remove the DRAM and set it aside for three or four days. Insert the DRAM into the system again and memtest. If the dram tests out fine, then your system is overvolting. Keep in mind, if you continue to use the system in this way, you will very likely damage the DRAM and have to RMA. Repetition with the RMA'd DRAM will repeat this damage.

Hey.. thanks for your reply :-)

I personally go with Intel Chipsets for Intel CPUs. I have had no problems installing 8GB on the P35 chipset (P5K/P5K-e/Pk5-Deluxe) and this is my advice for those when they ask me the board for 8GB.

I thought about switching.. really.

Perhaps its the last resort with the downside of loosing SLI GPU support.

However, this option has to wait until the next Bios release and the next payday :-)

Actually I don't think this is even related to the 8G?!

I have read so many threads dealing with 4GB memory problems

(the common setup now) in the form of 4*1 GB (which is 4 banks - full load again).

I would say that there might be some board overvolting of the DRAM and that you are getting a resonance of the DRAM strobe. The way to find out if this is the case is to remove the DRAM and set it aside for three or four days. Insert the DRAM into the system again and memtest. If the dram tests out fine, then your system is overvolting.

Yes, indeed I've noticed less errors after a cold start (even without removing

the DRAM) than after a memtest which I've run after some hours of testing.

I will definitely follow this idea, and test the DRAM with my previously stable settings after 2 days without surge.

Lets have a look what will happen. If it runs stable through the whole test again,

the overvolting is a real thread.

What could we / I do in this case?

Keep in mind, if you continue to use the system in this way, you will very likely damage the DRAM and have to RMA. Repetition with the RMA'd DRAM will repeat this damage.

Damn. I want to prevent this, really.

Especially because I'm from Germany and don't know the RMA way of Corsair in Europe..

Actually I don't think this is even related to the 8G?!

I have read so many threads dealing with 4GB memory problems

(the common setup now) in the form of 4*1 GB (which is 4 banks - full load again).

This issue is indeed problematic and VERY hard to isolate because you can take to identical boards and have one work with all four banks populated and another will just not make stability.

I agree. The problem is just aggravated with all four banks filled with 8GB vs all four banks filled with 4GB. All four banks filled is the determining factor

What could we / I do in this case?

I have already told you what my advice is to those who wish to populate all four banks and doubly so with 8GB.

Damn. I want to prevent this, really.

Especially because I'm from Germany and don't know the RMA way of Corsair in Europe..

I understand. The longer you continue this "Testing" and using the DRAM in this way will increase the possibility of final physical damage. I have seen this occur myself in more than a few systems and then RMA is the only recourse with repetition yielding the same result. I tell those people that to keep doing what they are doing will achieve their prior results.

I know of no other recourse. I personally would sell the board and buy a P35 based board. Look at the Striker issues on this and other forums with regards to 4 banks filled and doubly so for all four banks filled with 8GB. Look at the P5K-Deluxe issues with the same event and you will have your answer. And the Striker is the best of the 680i boards. Check the EVGA website out.

By the way.. this could be the cause for the 4 modules running stable

for several days with just 1,85V. :eek:

The 680i Overclocking Guide by Nvidia is not really in-depht regarding Voltages.

But I think this is more a question of MB Design, and Asus did their own.

(They didn't follow the reference design -huh? But its present on the EVGA, too...)

Now I'm getting even more confused.

Beer is the only solution..