VOGONS


8GB of DDR1 in a socket 939 system

Topic actions

Reply 101 of 119, by Billtm

User metadata
Rank Newbie
Rank
Newbie

hmmm, GTX660 only successfully worked with: HT - 1x\2x , AMD K8 C'n'Q - Enabled, Overclock Profile - Auto or HT - 1x\2x\3x , AMD K8 C'n'Q - Disabled, Overclock Profile - Manual, PCI E Clock - 100MHz, PCI Clock - 33.33MHz, with HT 4x\5x\Auto - black sceen. GTX580 works only in the safe mode and not boot OS in normal mode =\

Reply 102 of 119, by shamino

User metadata
Rank Oldbie
Rank
Oldbie
shamino wrote:
Billtm wrote:

I have found in what a problem, you were right - CPU Stepping D0 =( , my CPU stepping E6 and it's working.

Interesting. Well I think that's good news - this explains what's going on and at least I know that if I get a different CPU then it can be solved. I'd rather it be a CPU problem than a board problem.

I just got an X2 4200+ E6 stepping CPU in the mail and have installed it in the Asus A8N-SLI. This replaces an old D0 stepping 3500+ I had in the board before. The new CPU is an LCB9E, which is one of those codes that also appeared on Opteron CPUs, which implies to me it ought to have the best chance of handling the same RAM as the Opteron.

The "H/W DRAM Over 4G Remapping" option has now appeared. My BIOS options on that page now look like the picture from Billtm posted on 4/21. So the later CPU stepping did fix that problem.
ECC support was enabled in the BIOS, but I still spent about 15 minutes swapping modules trying to convince the BIOS on this board (Asus A8N-SLI) to POST with my 2GB Registered ECC modules. As was the case before, it was willing to POST with a 512MB and (eventually) a 1GB, but the 2GB require some kind of witchcraft to get the board to stop beeping about them and actually use them. I only persisted so long because I knew I had done it before, and once it starts working, it stays working unless the CMOS gets wiped. I guess there's some hidden parameters in the CMOS that have to be tricked into the correct position or something.

Anyway, I eventually got 2 of them to boot so it has 2x2GB installed now in dual channel at DDR400. With the SW and HW Remapping options enabled (the latter of which was previously unavailable with the old D0 CPU), the full 4GB is now available for use, and shows up both at POST and in WinXP64.
I have 2 more identical 2GB modules but I haven't been able to get that other pair to POST. Whatever is wrong, I'm not real worried about getting it to 8GB anyway.

The problems I'm left with both relate to my desire to control the power consumption of this system.

1) Even with the E6 stepping CPU, I still can't get P-state switching (which I think is what "Cool n Quiet" is) to work when registered memory is installed. Any attempt to change P-states with registered RAM still causes an immediate hard lock. With unbuffered memory it works fine, as it did before on the D0 CPU. For whatever it's worth, changing P-states works perfectly fine on my Socket 940 Opteron server with registered memory, so it's not an inherent limitation of the K8, it's something funny about this board or maybe the Athlon64 CPUs. Since WinXP doesn't utilize CnQ by default, and the number of people attempting to use registered RAM on Athlon64s was/is already limited, I'm not sure how well this scenario has been tested by the internet at large. For me at least it doesn't work.

2) The other problem is that with this E6 stepping I'm no longer able to set the voltage any lower than 1.100v in a software undervolting utility. I haven't tried doing it with the BIOS. This isn't relevant when running full speed but when using CPU throttling, it's nice to be able to set the idle voltage lower than 1.1v.

I primarily built this machine to be a power efficient alternative to my main desktop, so these 2 problems are a disappointment. On the point that's relevant to this thread, I will probably revert back to unbuffered RAM because the difference in idle power with and without working P-states is significant.
With unbuffered memory I can use P-states again, but as mentioned in point #2 I still can't set the idle voltage any lower than 1.10v now. The old D0 had been running at 0.85v. As a result, the X2 still uses a bit more power than the old 3500+ CPU did at idle. I might put up with that for the performance increase, but it's irritating to have to make that compromise since I bet the difference is almost entirely due to the voltage.

Even if power usage was not a concern, I'd be hesitant to commit to Registered 2GB (or even 1GB) modules in long term operation on the A8N-SLI because the BIOS has some strange problems getting into the mood to POST with them, and I'd be worried about the configuration failing and having to fight with it again in the future.

Reply 103 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie

Will either of these work on my Asus A8N-SLI Premium with my overclocked Athlon 64 X2 4400+?
http://i.ebayimg.com/images/g/3ycAAOSwZd1Vdv8U/s-l1600.jpg
http://i.ebayimg.com/images/g/H4oAAOSwubRXK1YH/s-l1600.jpg

Reply 104 of 119, by shamino

User metadata
Rank Oldbie
Rank
Oldbie
jm764 wrote:

Will either of these work on my Asus A8N-SLI Premium with my overclocked Athlon 64 X2 4400+?
http://i.ebayimg.com/images/g/3ycAAOSwZd1Vdv8U/s-l1600.jpg
http://i.ebayimg.com/images/g/H4oAAOSwubRXK1YH/s-l1600.jpg

I don't know about the second picture with the heatsinks, the chips aren't visible so specs would be needed.

The first picture (the Micron MT36VDDF25672Y) are the ones I've used on my A8N-SLI most recently. Before that I used Qimonda HYS72D256320HBR-5-C which are equivalent and behaved the same. Either of those have the right chip organization to be compatible, and they work perfectly on my Socket940 board of the same era, but I've found the A8N-SLI s939 setup to be quirky with them (and it doesn't support them officially so maybe the BIOS is to blame).

Those Micron MT36.. will work but the board (at least my non-Premium) required a strange "convincing" process to get it to start POSTing with them. I had to start with smaller registered modules. It was totally willing to post with 512MB RegECC, but less willing to post with 1GB and especially resistant to the 2GB modules. Through some combination of swapping modules in and out of the machine eventually it started to work with the 2GB. Once you get it working it will continue to work, but if the CMOS ever gets cleared you'll have to screw around with it again.
It was a strange experience and I don't claim that it makes sense, and maybe your board won't be so weird about it.

My A8N-SLI (non Premium) is working with 2x2GB of those Microns but I didn't get it to run with 4 of them. However, I didn't put much effort into that so maybe it can work. For all I know, maybe one of my 4 Micron modules is at fault. Pretty sure I tested them all on the 940 board but maybe the A8N-SLI doesn't like one of them - that could explain why I had to fiddle around with them so much.

Reply 105 of 119, by havli

User metadata
Rank Oldbie
Rank
Oldbie

Just a heads up - apparently ECC reg DDR can work on socket 754 too.
Specifically Asus K8V and Athlon 64 3200+ (ClawHammer core)... the catch is there are significant limitations. Only first (blue) DDR slot works and neither RAM frequency or timing can be adjusted (both depend on SPD). So in this case DDR 266 is the maximum speed.

s754_ecc_reg_ddrruj1n.png

HW museum.cz - my collection of PC hardware

Reply 106 of 119, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

Asus generally enables ECC where the chipset (or CPU in this case) has that capability built in, so I'm not surprised.

All hail the Great Capacitor Brand Finder

Reply 107 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie

Well the Micron MT36VDDF25672Y-40BF3 ram I ordered came in today and with it brought a 2GB DDR1 ram revelation with it for my Asus A8n-SLI Premium.

As it turns out not all "MT36" are made equal.

To be specific I got 4 modules with 2 different models and 2 similar ones.

The modules that worked were numbered "200639" and "200649" and booted with the system with no problem separately and together.

The ones that failed were both labeled "200635" and didn't boot at all with the "memory warning" beep going off once power was pushed.

Sadly I don't have a Socket 940 to test these two sticks with to make sure they aren't broken, but I'd say it's safe to assume that "200635" and lower Micron "MT36" Ram wont work with any Asus Socket 939 board at least.

I'll see what I can do with the other two sticks on an ASUS A8N-VM CSM Rev.1.01 (if it'll even take it) I have in another rig of mine.

Hope this helps, and let me know if any of you get these sticks working, it could just be my board or my sticks acting up.

Reply 108 of 119, by shamino

User metadata
Rank Oldbie
Rank
Oldbie
jm764 wrote:
Well the Micron MT36VDDF25672Y-40BF3 ram I ordered came in today and with it brought a 2GB DDR1 ram revelation with it for my As […]
Show full quote

Well the Micron MT36VDDF25672Y-40BF3 ram I ordered came in today and with it brought a 2GB DDR1 ram revelation with it for my Asus A8n-SLI Premium.

As it turns out not all "MT36" are made equal.

To be specific I got 4 modules with 2 different models and 2 similar ones.

The modules that worked were numbered "200639" and "200649" and booted with the system with no problem separately and together.

The ones that failed were both labeled "200635" and didn't boot at all with the "memory warning" beep going off once power was pushed.

Sadly I don't have a Socket 940 to test these two sticks with to make sure they aren't broken, but I'd say it's safe to assume that "200635" and lower Micron "MT36" Ram wont work with any Asus Socket 939 board at least.

I'll see what I can do with the other two sticks on an ASUS A8N-VM CSM Rev.1.01 (if it'll even take it) I have in another rig of mine.

Hope this helps, and let me know if any of you get these sticks working, it could just be my board or my sticks acting up.

The numbers you mentioned are date codes. I looked at a bunch of Micron memory that I have and they all have 6 digit date codes like that down to some SDRAM memory that was made in 2001. I found a few older SDRAM modules with Micron's older logo (an M without any circle behind it) and on those the date code was different. If those had a date code on them at all then it was a 4 digit code that started with "00" (2000) instead of the 6 digit that they started using later.

It's interesting that the older date codes didn't work but the later ones did. Unless it's just a coincidence, then I wonder if Micron changed something in the SPD programming after a certain date that made the difference.

Unfortunately my A8N-SLI isn't accessible right now so I can't check the codes on mine.

Reply 109 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie
shamino wrote:

The numbers you mentioned are date codes. I looked at a bunch of Micron memory that I have and they all have 6 digit date codes like that down to some SDRAM memory that was made in 2001. I found a few older SDRAM modules with Micron's older logo (an M without any circle behind it) and on those the date code was different. If those had a date code on them at all then it was a 4 digit code that started with "00" (2000) instead of the 6 digit that they started using later.

It's interesting that the older date codes didn't work but the later ones did. Unless it's just a coincidence, then I wonder if Micron changed something in the SPD programming after a certain date that made the difference.

Unfortunately my A8N-SLI isn't accessible right now so I can't check the codes on mine.

I see.

20160908_075227.jpg
Filename
20160908_075227.jpg
File size
3.33 MiB
Views
1038 views
File license
Fair use/fair dealing exception

Well I did a bit more investigation and found out that the chips in the center of the "200635" vs the "639" and "649" are different. as shown in the image. The "635" has a "SSF859" chip in the center and "639" and "649" both have a "ICS A16859CFLK" chip in the center.

I'm looking into what differences the two chips have now, but if you find any "200639" or later in the meantime that work or any other chip that will work post it's information below.

Also here's the Micron site for the ram (though I'm not a "ram buff" so I can't make much of the information listed) for more information on it: https://www.micron.com/parts/modules/ddr-sdra … 889E087AE182%7D

Reply 110 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie

https://youtu.be/g8IVI0sZ6F8
"Good News Everyone"! 😁

The company that hosted the ebay listing I bought the sticks from was kind enough to freely send another set of two "200639" and "200349" modules to me and I am glad to say all 4 sticks boot and run fine on my ASUS A8N-SLI Premium just fine.

It seems that it's probably the "SSF859" chip on the "200635" that doesn't like the board, and a few later models also have the chip instead of the "ICS A16859CFLK" on the 639s and 649s.

I'll post some screenshots with CPU-Z and some benchmarks tomorrow once MemTest 86+ (ver 5.01) is done testing them so I can make sure there's no problems with my sticks.

Here's the old EBAY listing for reference: http://www.ebay.com/itm/Micron-PC3200R-ECC-8G … =p2047675.l2557

Reply 111 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie
8GB_DDR.jpg
Filename
8GB_DDR.jpg
File size
567.33 KiB
Views
973 views
File comment
CPU-Z Results
File license
Fair use/fair dealing exception

It took a bit of fighting but I got the ram running stable now after a couple of bad memory settings in the bios.

Sadly it seems that the only way it will work (at least for me) is on auto so no overclocking (not even setting it to 400mhz works) sadly.

Reply 112 of 119, by jm764

User metadata
Rank Newbie
Rank
Newbie
20160928_205418.jpg
Filename
20160928_205418.jpg
File size
4.66 MiB
Views
863 views
File comment
Bios Settings. Ram Voltage set to 2.6 (lowest my board will go)
File license
Fair use/fair dealing exception
20160928_195613.jpg
Filename
20160928_195613.jpg
File size
2.12 MiB
Views
863 views
File license
Fair use/fair dealing exception

Well after a bit of tweaking I got the ram to 400 Mhz stable.

Sadly my system locks up in "Normal" mode as there seems to be a problem with my Sound Blaster X-Fi Pci sound card as it's totally fine when the system runs in "Diagnostic Startup" mode in Windows 7 Ultimate 64-Bit.

I'll try some more test to get the on-board sound working then the SB later this evening.

Reply 113 of 119, by shamino

User metadata
Rank Oldbie
Rank
Oldbie
jm764 wrote:
Well the Micron MT36VDDF25672Y-40BF3 ram I ordered came in today and with it brought a 2GB DDR1 ram revelation with it for my As […]
Show full quote

Well the Micron MT36VDDF25672Y-40BF3 ram I ordered came in today and with it brought a 2GB DDR1 ram revelation with it for my Asus A8n-SLI Premium.

As it turns out not all "MT36" are made equal.

To be specific I got 4 modules with 2 different models and 2 similar ones.

The modules that worked were numbered "200639" and "200649" and booted with the system with no problem separately and together.

The ones that failed were both labeled "200635" and didn't boot at all with the "memory warning" beep going off once power was pushed.

Sadly I don't have a Socket 940 to test these two sticks with to make sure they aren't broken, but I'd say it's safe to assume that "200635" and lower Micron "MT36" Ram wont work with any Asus Socket 939 board at least.

I'll see what I can do with the other two sticks on an ASUS A8N-VM CSM Rev.1.01 (if it'll even take it) I have in another rig of mine.

Hope this helps, and let me know if any of you get these sticks working, it could just be my board or my sticks acting up.

jm764 wrote:
Well I did a bit more investigation and found out that the chips in the center of the "200635" vs the "639" and "649" are differ […]
Show full quote
20160908_075227.jpg

Well I did a bit more investigation and found out that the chips in the center of the "200635" vs the "639" and "649" are different. as shown in the image. The "635" has a "SSF859" chip in the center and "639" and "649" both have a "ICS A16859CFLK" chip in the center.

I'm looking into what differences the two chips have now, but if you find any "200639" or later in the meantime that work or any other chip that will work post it's information below.

Also here's the Micron site for the ram (though I'm not a "ram buff" so I can't make much of the information listed) for more information on it: https://www.micron.com/parts/modules/ddr-sdra … 889E087AE182%7D

I checked the markings on my pair of MT36VDDF25672Y modules that are working in an Asus A8N-SLI (not Premium), and I took note of them in a text file but I forgot to ever post this. Sorry it took so long.
My working pair both have the date code 200635 PR
the buffering chip is marked like this:
ssf859
ti 67z
nf67

both modules of the working pair are marked the same. After rereading your posts I see that these match the 200635 modules that didn't work on your board. So I guess the nature of the compatibility quirk remains a mystery.
I have 2 more of these modules that are not in the machine, because I couldn't get them all to work. Unfortunately I can't find them - they've been misplaced so I can't report the markings on those. I seem to have a problem with misplacing RAM, this isn't the first time I've done that.
I'm confused enough about the behavior that I'm not totally sure if my troubles could be pinned down to 1 or 2 incompatible pieces of RAM, or if it was some other issue. It seems that maybe some of them were incompatible for some unknown reason, but I wish I could find them to investigate more.

Reply 114 of 119, by rukwar

User metadata
Rank Newbie
Rank
Newbie
Billtm wrote:

PS. I was wrong, i read this topic
for work trim the SSD-disk does not need RAID 0 mode and the modified driver, just use standard IDE mode and standard driver MS. SSD-disk must be connected to the port SATA3\4 (SATA rev. 2.0 (3 Gbit/s, 300 MB/s)) and to use MS IDE-driver (ATA Device), when the nVidia driver is installed, the SATA3\4 port uses SATA-driver (SCSI Device) and Trim doesn't work. The SATA1\2 port (SATA rev. 1.0 (1.5 Gbit/s, 150 MB/s))
You do not need to modify the BIOS, use original 1604 from Asus site.

this thread is old but i'm hopping that anybody can help me on this. the manual says that the SATA1/2 have the setting master and are used for boot disks. and the sata 3/4 are slaves and are used to data disks. The manual doesn't specifie that the connector have different speeds. it says that the sata has "up to 3Gb/s data transfer speeds".
How can i ensure that windows doesn't install the nvida driver to control the hard drives. do i need to change the driver that windows use? how can i do that? windows 7-10 dont' use the default driver?

Thanks in advance
Best regards

Reply 115 of 119, by matze79

User metadata
Rank l33t
Rank
l33t

There even 4Gb Sticks.

https://www.ebay.de/itm/4GB-4GBx1-Samsung-PC3 … 2gAAOSwIWtfvOG6

https://dosreloaded.de - The German Retro DOS PC Community
https://www.retroianer.de - under constructing since ever

Co2 - for a endless Summer

Reply 116 of 119, by matze79

User metadata
Rank l33t
Rank
l33t
kithylin wrote on 2016-04-16, 02:03:

...

What a shame.. why not simply let your account anonymize and delete.. instead destroying whole Conversation.

https://dosreloaded.de - The German Retro DOS PC Community
https://www.retroianer.de - under constructing since ever

Co2 - for a endless Summer

Reply 117 of 119, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

There even 4Gb Sticks.

Double stacked! In all seriousness, even 8 Gb is a bit over the top.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 119 of 119, by PC-Engineer

User metadata
Rank Member
Rank
Member

I have an A8NSLI Premium working with a kit of 4x2GB PC3200R (Infineon) and 8GB usable in Win7 (64bit) - with one Graphics Card installed.
BUT: with a second Graphics Card (SLI) it recognize "8GB installed (4GB usable)", same in Win7 (64bit).

Does the A8N32 SLI has the same effect to the available RAM with two Graphics Cards installed?

1994/1995 - Socket3 - ASUS SV2GX4 / POD 100MHz / 64MB / SCSI - Windows 95