Topic: KT4 Ultra MS-6590 - Win XP Home boots with 2GB DDR333 installed but not 3GB (Read 1699 times)

rjboomerang · « **on:** 18-July-12, 11:50:12 »

Spec...

KT4 Ultra MS-6590
AMD Athlon XP 3000+ (FSB set to 166MHz in BIOS)
BIOS MS-6590v1.4
Win XP Home 2002, SP3

Problem...

Previously, had 768MB RAM installed, and XP strangely identified CPU as a 2200+.

Recently upgraded the RAM to 3GB DDR333 (3 sticks of 1GB 184-pin unbuff DIMM 128MX64 DDR PC2700), and changed BIOS settings for CPU and DRAM.

In the DOS boot screen, Win XP progress indicator fills completely from left to right, then stops with a half-blinking cursor in the top left of the screen.

Removed 1GB stick and Win XP successfully loaded, also now recognising the Athlon XP 3000+, and running a heck of a lot faster.

Tried adding the /3GB switch to the boot.ini file, which didn't do anything, and after reading more forum advice, guessed this might be irrelevant. I don't think this is a 4GT and PAE issue, as I'm not trying to install above 4GB (if I've correctly understood).

Also, forum advice (various sites) conflicts on whether Win XP 32-bit is limited to 2GB or 4GB, and I can't find any workarounds for my current setup.

Any ideas?

Svet · « **Reply #1 on:** 18-July-12, 12:51:56 »

what exactly memory sticks you use? read >>Posting Guide<<

rjboomerang · « **Reply #2 on:** 19-July-12, 22:06:50 »

As specified above: 3 sticks of 1GB 184-pin unbuff DIMM 128MX64 DDR PC2700

I bought these recently from Crucial Memory after running their diagnostic tool.

Svet · « **Reply #3 on:** 19-July-12, 22:07:44 »

have you tested each one of the sticks alone ?

rjboomerang · « **Reply #4 on:** 19-July-12, 22:16:40 »

When Win XP Home failed to boot with 3GB, I tested with 2GB in the first two memory banks, and boot was successful. Then I swapped the 3rd stick (which was previously in the 3rd bank, when boot failure) with the stick in the 2nd bank, and Win XP successfully loaded with this combination of sticks. So, my conclusion was that there was nothing wrong with any of the sticks. I guess one possibility is that the 3rd bank slot isn't making contact with the pins? I just realised, I still need to test a cold boot with the BIOS 'fast boot' set to 'disabled' whilst the full 3 GB is installed, so I can see if the memory counter/test reaches the full 3GB (I'm assuming that's what it means).I'll try it and let you know the result.

Svet · « **Reply #5 on:** 19-July-12, 22:21:18 »

are all memory sticks are exactly the same?
At what memory timings they are running and voltage?

eagle2 · « **Reply #6 on:** 19-July-12, 22:58:50 »

The board probably supports 3GB at 133/266ddr bus, but not 166/333. You could try downclocking the the memory bus to 266ddr to check.
Not good running the memory below cpu fsb though...

If you want 3GB then you may have to downclock the cpu to 133 as well. And your xp3000 will become an xp2400 or thereabouts...

Ex Forum Mod · « **Reply #7 on:** 19-July-12, 23:11:40 »

Not sure if it makes any difference on that MB but are the RAM sticks high density (chips on 1 side only) or low density (chips on both sides)?

rjboomerang · « **Reply #8 on:** 20-July-12, 23:00:42 »

@Svet: I set the BIOS to auto, but will now reboot and make a note of the settings, assuming the BIOS will tell me.

@Easle2: Isn't the point of having DDR333 to be able to set the FSB to 166? The KT4 Ultra MS-6590 will theoretically support DDR400, but stability with overclocking is not certain (from what I've read) and only 2 memory banks are usable at FSB200. So, isn't expecting the board to perform at FSB 166 like asking a VW Golf 1.8 to perform at 150BHP when the engine can really handle over 200 BHP? (Please don't pull me up on the analogy, which is imperfect, like most are.) Or, maybe there's an issue between the 2GB and 3GB? -However, I'm successfully running at FSB 166 with 2GB installed, so... don't understand the rationale for downclocking. (However, I'll test it tomorrow just to eliminate the possibility.)

@Henry: The sticks have chips on both sides and all three are identical.

@all: I just did a test booting up with the BIOS quick boot parameter set to disabled. When 2GB is installed, the count = 2097152 and Win XP successfully loads (in BIOS, FSB IS set to 166). When 3GB is installed, the count = 3407872 (looks like 3GB to me) and Win XP loading progress bar in DOS fills from left to right, then stops, and a half-blinking white block appears in the top left of the screen (still in DOS mode) and nothing else happens.

Unless I've missed something in the BIOS, could the evidence be pointing towards something in Win XP Home as the source of the problem?

By the way, many thanks for all your suggestions!

flobelix · « **Reply #9 on:** 20-July-12, 23:18:04 »

What seems so far still unanswered is the question if those sticks are identical. And i don't just mean same size.
If not this could also be a compatibility problem with the sticks. The old Via chipsets are quite picky when it comes to ram.

rjboomerang · « **Reply #10 on:** 20-July-12, 23:21:22 »

@Eagle2: Just thinking about your comment - When I had only 768MB installed (DDR333), I recall that the FSB was only set to 133 (last time I fiddled with it was years ago), so perhaps this explains why Win XP had identified the processor as only 2200+? Now I have 2GB installed and set the FSB to 166, Win XP sees the processor as 3000+ (which is in fact a Barton).

rjboomerang · « **Reply #11 on:** 20-July-12, 23:24:38 »

@flobelix: I haven't examined all three sticks yet (as I'm typing on the very machine we're talking about), but they were all new this week from the same supplier (Crucial Memory), so the likelihood is (from the supplier's longstanding reputation) that the sticks are all the same spec.

flobelix · « **Reply #12 on:** 20-July-12, 23:32:48 »

Has nothing to do with manufacturer or supplier. As I guess the three sticks were not a kit (as I'm not aware of a ddr1 three stick kit) these are most likely not identical. Same specs (if so) means not those are identical. Could have entirely different chips. Get CPUZ and read out the SPD of the sticks.

rjboomerang · « **Reply #13 on:** 20-July-12, 23:51:33 »

@flobelix: OK thanks, I'll check out CPUZ (I'm guessing it's a free utility I can download?) and report back.

@Svet: I've just noted the BIOS voltage settings and PC Health status, which are as follows:

Freq/Voltage Control Spread Spectrum +/- 0.25% [I'd guess this is the default setting, which helps to smooth any spikes]
FSB 166 MHz
CPU Ratio auto
CPU Vcore auto
DDR Voltage auto
- Termination auto
AGP Voltage auto

PC Health Status: [I'm guessing this will give a snapshot of the conditions, which might indicate a stable/unstable environment?... but obviously not possibly to determine fully without a time series... though I can say when in BIOS setup mode, the temps and voltages were fairly stable, so probably no cause for concern with the PSU and motherboard voltage regulators (but obviously, I'm not an expert on this)?]
CPU Temp 55 deg C
Sys Temp 46 deg C
Vcore 1.616V
+5.0V 4.893V
+12.0V 12.228V
-12.0V -11.825V
-5.0V -4.976V

Battery 3.28V
+5V SB 4.824V

[Probably a lot of superfluous detail, but here you have it, just in case something is relevant!]

flobelix · « **Reply #14 on:** 20-July-12, 23:54:35 »

Quote from: rjboomerang on 20-July-12, 23:51:33

@flobelix: OK thanks, I'll check out CPUZ (I'm guessing it's a free utility I can download?) and report back.

Right! CPUZ is free.

rjboomerang · « **Reply #15 on:** 20-July-12, 23:56:31 »

@Svet: Memory timing is 2.5 CAS latency, which was auto-detected and consistent with the memory specs.

flobelix · « **Reply #16 on:** 21-July-12, 00:04:07 »

For using three sticks you could try setting command rate to 2T if not done already.

rjboomerang · « **Reply #17 on:** 21-July-12, 00:10:16 »

@flobelix:

OK, here are the results fresh in from CPU-Z!

[Slot#1]
Module Size: 1024
Max bandwidth: PC2700 (166 MHz)
Manufacturer: Micron Technology
Part Num: 16VDDT12864AY335F2
Week/Year: 19/18
JEDEC#1: 133 MHz, CAS 2.0, RAS to CAS 3, RAS Precharge 3, tRAS 6, Voltage 2.50
JEDEC#2: 166 MHz, CAS 2.5, RAS to CAS 3, RAS Precharge 3, tRAS 7, Voltage 2.50

[Slot#2]
Module Size: 1024
Max bandwidth: PC2700 (166 MHz)
Manufacturer: Micron Technology
Part Num: 16VDDT12864AY335F2
Week/Year: 19/18
JEDEC#1: 133 MHz, CAS 2.0, RAS to CAS 3, RAS Precharge 3, tRAS 6, Voltage 2.50
JEDEC#2: 166 MHz, CAS 2.5, RAS to CAS 3, RAS Precharge 3, tRAS 7, Voltage 2.50

I haven't tested the 3rd stick yet, as need to replace one of the above sticks with it in order to boot up the desktop... watch this space!

rjboomerang · « **Reply #18 on:** 21-July-12, 00:12:46 »

@flobelix: I'll just replace one of the sticks with the 3rd one and re-run CPU-Z so we have the full results, then I'll reboot again and change the command rate to 2T with 3 sticks installed.

rjboomerang · « **Reply #19 on:** 21-July-12, 00:48:13 »

@flobelix: Ok, so replacing the 2nd with the 3rd stick produces exactly the same results in CPU-Z. The only difference is the serial code. Also, I changed the BIOS settings as follows, and no success in loading Win XP Home with 3GB installed...

Latency: auto
Row precharge: 2T
RAS pulsewidth: Auto
RAS to CAS delay: 2T
Bank Interleave: 4-way
DDR DQS Input Delay: Auto
SDRAM Burst Length: 4QW
SDRAM 1T Command: Enabled
Fast Command: Ultra
Fast R-2-R Turnaround: Enabled

Do you think we might be converging on 'Win XP Home' as the issue or are there other h/w or BIOS config factors to test?

xmad · « **Reply #20 on:** 21-July-12, 01:40:30 »

Did you try raising the voltage to the ram slightly/ memory controller? Sounds like the extra bandwidth is causing a problem.

Svet · « **Reply #21 on:** 21-July-12, 12:13:38 »

Quote from: rjboomerang on 21-July-12, 00:48:13

@flobelix: Ok, so replacing the 2nd with the 3rd stick produces exactly the same results in CPU-Z. The only difference is the serial code. Also, I changed the BIOS settings as follows, and no success in loading Win XP Home with 3GB installed...

Latency: auto
Row precharge: 2T
RAS pulsewidth: Auto
RAS to CAS delay: 2T
Bank Interleave: 4-way
DDR DQS Input Delay: Auto
SDRAM Burst Length: 4QW
SDRAM 1T Command: Enabled
Fast Command: Ultra
Fast R-2-R Turnaround: Enabled

Do you think we might be converging on 'Win XP Home' as the issue or are there other h/w or BIOS config factors to test?

at what freq. they are running with those parameters?

Disable this:
SDRAM 1T Command: Enabled

And relax this value if possible:
Fast Command: Ultra

rjboomerang · « **Reply #22 on:** 22-July-12, 10:38:43 »

@Svet: See my CPU-Z results posted earlier for individual RAM frequencies with 2 chips installed (JEDEC#1 and JEDEC#2 on the two sticks are both running at 133MHz CAS2.0 and 166MHz CAS2.5). As you've suggested, I'll set 'SDRAM 1T Command' to disabled. The fast command can be set to normal, fast or ultra, so I'll relax the value to normal.

@xmad: After trying Svet's suggestions, if it doesn't work with the 3 sticks, I'll try increasing the voltages to RAM slightly.

Thanks guys.

flobelix · « **Reply #23 on:** 22-July-12, 16:13:31 »

We're excited tell us how it goes.

rjboomerang · « **Reply #24 on:** 22-July-12, 22:29:21 »

Ok, so here are the results:

[Test1]

* Installed 3rd stick of 1GB DDR333
* SDRAM 1T Command = disabled
* Fast Command = Normal

Results: Win XP Home failed to load

[Test 2]

* Installed 3rd stick of 1GB DDR333
* Increased DDR Voltage through available options of 2.6, 2.7 and 2.8V and reboot with each change
* Also tried the above with DDR Termination Voltage set to 1.27 or 1.29V
* Also tried combinations of SDRAM 1T Command = disabled / enabled
* Also tried combinations of Fast Command = Normal / Fast / Ultra

Results: Win XP Home failed to load

rjboomerang · « **Reply #25 on:** 22-July-12, 22:32:35 »

One thing I haven't checked yet is the voltages displayed in the BIOS PC Health Status when 3 sticks are installed - this might indicate if the 3rd DDR appears to be affecting the Vcore voltages...

flobelix · « **Reply #26 on:** 22-July-12, 22:45:04 »

You also havent posted the spd details of the 3rd stick. I suppose one of the sticks to be faulty or the three are simply not matching and not working altogether.

rjboomerang · « **Reply #27 on:** 22-July-12, 22:49:47 »

Quote from: flobelix on 22-July-12, 22:45:04

You also havent posted the spd details of the 3rd stick. I suppose one of the sticks to be faulty or the three are simply not matching and not working altogether.

@flobelix: I already stated above "replacing the 2nd with the 3rd stick produces exactly the same results in CPU-Z. The only difference is the serial code." See the post previous to this with details of the sticks.

rjboomerang · « **Reply #28 on:** 22-July-12, 22:57:31 »

With the 3 sticks installed, the BIOS PC Health Status shows the following voltages:

Vcore: 1.632V
+5.0V: 4.920V
+12.0V: 12.167/12.228V
-12.0V: 11.743/11.825
-5.0V: -4.976

With only 2 sticks installed, the status is as follows:
Vcore: 1.616V
+5.0V: 4.893V
+12.0V: 12.228V
-12.0V: -11.825V
-5.0V: -4.976V

So, the only real difference with 3 sticks is a slight rise in the Vcore and the +5.0V.

No idea what this signifies, if anything...

flobelix · « **Reply #29 on:** 22-July-12, 23:03:12 »

Quote

The only difference is the serial code

Have overlooked that. Conclusion sticks are not identical. With a different part number these could have a different type of chips. As those are not matched and the old mem controlers in via chipsets are quite picky they are simply not working together. Maybe you are lucky and they will work with DDR266 or 200.

rjboomerang · « **Reply #30 on:** 22-July-12, 23:27:53 »

Another test I've done is to enable boot logging with Win XP. This appends command lines to the file 'ntbtlog' in the windows directory, which mainly shows drivers being loaded as Win XP is launched. The log is normally date and time-stamped for every boot-logging session.

When all 3 sticks are installed, I cold boot, then press F8 and select 'enable boot logging' from the options menu. This is what appears on the screen as the system, attempts to load Win XP.
--------------------
Loaded driver \WINDOWS\system32\ntoskrnl.exe
Loaded driver \WINDOWS\system32\hal.dll
Loaded driver \WINDOWS\system32\KDCOM.DLL
Loaded driver \WINDOWS\system32\BOOTVID.dll
Loaded driver ACPI.sys
Loaded driver \WINDOWS\System32\DRIVERS\WMILIB.SYS
Loaded driver pci.sys
Loaded driver isapnp.sys
Loaded driver viaide.sys
Loaded driver \WINDOWS\System32\DRIVERS\PCIIDEX.SYS
Loaded driver MountMgr.sys
Loaded driver ftdisk.sys
Loaded driver PartMgr.sys
Loaded driver VolSnap.sys
Loaded driver atapi.sys
Loaded driver disk.sys
Loaded driver \WINDOWS\System32\DRIVERS\CLASSPNP.SYS
Loaded driver fltmgr.sys
Loaded driver sr.sys
Loaded driver PxHelp20.sys
Loaded driver KSecDD.sys
Loaded driver Ntfs.sys
Loaded driver NDIS.sys
Loaded driver Mup.sys
Loaded driver BTHidMgr.sys
--------------------

Then the process freezes and a solid white block appears in the top-left of the screen, with the top-right corner of the block blinking.

At this point, I turn off the PC using the main power switch, then remove the 3rd memory stick. After booting up again, with only 2 sticks installed, opening the text file 'ntbtlog' reveals that this boot event wasn't logged.

So, I reboot Win XP with only the 2 sticks installed, and select 'enable boot logging'. This time, the above list of loaded drivers is shown in the text file, plus a whole lot more. The potentially interesting observation is the next in the list after "Loaded driver BTHidMgr.sys" which is "Loaded driver \SystemRoot\System32\DRIVERS\amdk7.sys".

'amdk7.sys' is a driver file which seems to come with one of the Win XP Service Packs, and which "is necessary to enable Windows XP support for AMD PowerNow! technology on computers that use a mobile Athlon 4 or mobile Duron processor." http://support.microsoft.com/kb/306458

Various other posts refer to it in relation to AMD Athlon XP 3000+ Mobile, whereas mine is a 3000+ in a desktop. Also, if 'amdk7.sys' isn't available, then Win XP will load the generic driver 'processr.sys'. In fact, my 'ntbtlog' includes the entry "Did not load driver \SystemRoot\System32\DRIVERS\processr.sys" which would make sense.

However, my curiosity is that with the 3 sticks installed, Win XP fails to continue loading when reaching 'amdk7.sys' in the driver list.

Another test I did was to change the power management settings in the BIOS.

"IPCA Function: This item is to activate the ACPI (Advanced Configuration and Power
Management Interface) function. If your operating system is ACPI-aware, such
as Windows 98SE/2000/ME, select Yes. Available options: Yes, No."

This was normally set to Yes. When I set it to No and rebooted with 3 sticks installed, instead of a half-blinking white block (as described above), the screen went black and then the system rebooted itself...

The plot thickens!...

Maybe there's a clear explanation for this - however, I'm wary of doing something to the 'amdk7.sys' driver file (whether removing or disabling it somehow) in case it prevents Win XP from launching with just my 2 sticks installed. I'd guess there are precautions, but need further info on whether this is relevant, and what the implications might be.

rjboomerang · « **Reply #31 on:** 22-July-12, 23:36:43 »

Quote from: flobelix on 22-July-12, 23:03:12

Have overlooked that. Conclusion sticks are not identical. With a different part number these could have a different type of chips. As those are not matched and the old mem controlers in via chipsets are quite picky they are simply not working together. Maybe you are lucky and they will work with DDR266 or 200.

But the part numbers are identical (see above). The serial numbers are different, which I didn't post, as assumed this is a unique number to identify every new stick which comes off the production line?... Unless this assumption is wrong.

For the two sticks currently installed...
[Slot #1]
Part Num: 16VDDT12864AY335F2
Serial Num: D47699EE
[Slot #2]
Part Num: 16VDDT12864AY335F2
Serial Num: D4766A21

Also, I already tried dropping the SDRAM speed to 266 and no joy.

flobelix · « **Reply #32 on:** 22-July-12, 23:59:47 »

If serial number is different makes no difference. Definitely it is not identical even a different batch can have other chips than the batch before. None the less the sticks don't work together. With three sticks installed this is the max the controler can handle. If then there are other problems like sticks beeing not identical this can be too much t handle. If no further relaxing the timings (ddr200) und setting slowest possible latency timings you will just end up with 2 Gig.

rjboomerang · « **Reply #33 on:** 23-July-12, 14:44:29 »

@flobelix: So, to recap on my obervations and your diagnosis of the root cause of the problem:

* The old mem controlers in via chipsets are quite picky and so when the MS-6590 motherboard has the full 3GB RAM installed, the memory sticks need to have identical chips in order to function. One way to reduce sensitivity might be to relax the timings to DDR200 and the slowest possible CAS latency settings.

[rjboomerang: I'll test this later, however, I suspect the performance gain from 2 to 3GB RAM running at DDR200 (FSB100) will be of less net benefit than the Athlon XP 3000+ running at FSB166 with only 2GB of memory running at DDR333 and the fastest memory settings possible in the BIOS.]

* When testing the SPD of the mem sticks using CPU-Z, the part numbers were identical for all 3 sticks, but serial numbers were different. Neither the part numbers nor the serial numbers provide an indication of whether the chips are identical or not. Therefore, CPU-Z is useful for diagnosing memory type and current performance/settings, but not for identifying the physical chips.

[rjboomerang: I haven't yet examined each stick to see if there are identifiable markings on each set of chips to distinguish them, but this might be pointless anyway?]

So... some unanswered questions:

1)
I'm curious as to why MSI would sell a motherboard with a max capacity of 3GB, which is possible in theory, but not in practice? - Was it only tested by MSI using perfectly identical memory chips before market release?
I can understand that the DDR400 feature of the motherboard was somewhat experimental when it was first released (at DDR400, the max memory may have been limited to 2GB), but surely using 3GB of DDR333 is then below max performance and shouldn't be considered to be experimental?

2)
If the Via memory controller will in fact manage 3GB DDR333 if the mem chips are identical, then from where/how does one purchase these to ensure they are identical?

I can imagine some of you might be reading this and thinking "why doesn't rjboomerang just upgrade the motherboard and avoid this issue?" - I guess it's because I'm trying to squeeze the last bit of life out of it (including the Athlon XP 3000+, which is still a reasonable processor for a home desktop) before having to build from scratch.

flobelix · « **Reply #34 on:** 23-July-12, 15:04:11 »

Quote

I'm curious as to why MSI would sell a motherboard with a max capacity of 3GB, which is possible in theory, but not in practice?

Has nothing to do with MSI. It's always a gamble mixing not matched sticks especially with the old controlers in Via chipsets. It also works with good matched or at least identical sticks. If those are not only trying will tell and might also result in that not working. Also keep in mind the technology you are using is antiquated. There is only a little choice of ram still available. The good quality sticks that could be bought in the time when your board was up to date aren't in production or for sale anymore.

That also leads to question two as you will probably not be able to buy proper DDR1 these days. Just face the fact that by using old not anymore supported technology you may not be able to use it like you could have with supply of proper parts. As all these days sold ddr1 sticks will be remainders it will be more or less a matter of luck.

With modern ram or in the old days of ddr1 you could buy kits of ram what means preselected ram to work together. If you choose not to go for that you have to have luck as you will not know from which batch seperately bought sticks are and if the manufacturer maybe changed the chips for the new batch.

rjboomerang · « **Reply #35 on:** 28-July-12, 10:55:53 »

Thanks flobelix, I learned a lot! It seems unlikely I can return the third stick to the vendor as 'unneeded', but fortunately it didn't cost a fortune, so I'll put it down to 'experience'.

After all this, I'm satisfied my desktop is dancing a lot faster now with the 2GB installed, the processor is recognised as a 3000+, and I can set the BIOS parameters high without any issues, and FSB to 166.

flobelix · « **Reply #36 on:** 28-July-12, 11:51:30 »

Glad I could help and good to hear you have at least a little improvement with the 2GB.

eagle3 · « **Reply #37 on:** 29-July-12, 22:12:12 »

It's the Via KT400 chipset. Supports 3GB @ 133/266ddr but only 2GB @ 166/333ddr.

You could get a 133fsb xp2800(runs at 16x133) and use all 3 sticks. With the xp3000 you'd be running at 13x133. You could run the xp3000 at 166 bus and the 3GB of memory at 133/266 ddr but I don't think it would be any better performance-wise.

I'm still running a KT600 board with only 2x512MB..

News:

Author Topic: KT4 Ultra MS-6590 - Win XP Home boots with 2GB DDR333 installed but not 3GB (Read 1699 times)