ThinkPad T61p unstable with more than 4GB memory

[JNavas]

Although others have reported success with more than 4GB memory in the ThinkPad T61p,
my own experience was close, but no cigar:
Passed Windows Memory Diagnostic (Standard),
but system became slightly unstable, app crash every 1-2 hours,
failed Lenovo Toolbox memory test.
4GB: Micron MT16HTS51264HY
2GB: Kingston KTH-ZD8000C6
6GB TOTAL
Went back to 4GB (2GB + 2GB) and the system is rock stable again.
My guess(tm) is a slight/subtle timing problem with that much memory
on the Intel 965 Express chipset (which is speced for a max of 4GB).

Machine Type: 6460-DVU
BIOS: 7LETC7WW (2.27) 04/08/2010
Embedded Controller: 1.08

[TTY]

A friend of mine, who upgraded an Intel PM965 chipset notebook to 6 GB, said it depends on installation order. First, he installed all 6 GB at once. The notebook wouldn't even start. Then, he installed only one stick and booted. Shut it down and installed the second stick. Result: no problems. It could also be that the position of the 4 GB stick matters (slot #0/#1).

[JNavas]

I sincerely thank you for the suggestions, but:

• I had no boot issue -- 4GB+2GB passed POST and Windows Memory Diagnostics, and booted Windows.
• The only slot rule on the T61p is that Slot 0 (bottom) must be filled first -- which module is in which slot shouldn't matter. If it does, the system isn't stable. (FWIW, I had 4GB in slot 0 and 2GB in slot 1.)
• I think your friend's initial problem was probably due to not seating one of the modules properly. I reseated my modules to be sure that wasn't an issue here.

That it works for your friend on some other notebook of unspecified type and configuration is great, but it's not stable on the T61p here, and stability is critical to me.
Thanks again.

[richk]

I assume you have the latest BIOS. I have seen a few situations where mixed brands don't work quite right, even though they are certainly supposed to work. Problems always went away when I used matched sticks.

[JNavas]

I do have the latest BIOS.

[sjthinkpader]

The problem may be related to capacitive loading on the memory bus. 4GB+2GB may mean 16chips+16chips=32 chips. The memory address bus driver may not have enough drive for 32 chips or certain chips have slightly higher input capacitance. The preferred loading should be 8chips+8chips but 16chip+8chip should work fine.

[JNavas]

The problem may be related to capacitive loading on the memory bus. 4GB+2GB may mean 16chips+16chips=32 chips. The memory address bus driver may not have enough drive for 32 chips or certain chips have slightly higher input capacitance. The preferred loading should be 8chips+8chips but 16chip+8chip should work fine.

Thanks, but that theory doesn't hold up (in this case at least):

• The 4GB module has only 8 chips.
• Neither the Intel spec not the Lenovo spec says anything about the number of chips, just the permitted module capacities.
• There's no issue with 16chip + 16chip 2GB modules.

[RealBlackStuff]

There are often issues with Kingston RAM modules.
Try one from another brand, e.g. from
Crucial = Micron = stable.

[JNavas]

There are often issues with Kingston RAM modules.
Try one from another brand, e.g. from Crucial = Micron = stable.

Thanks, but I think it unlikely that standard grade Kingston (KTH, not KVR ValueRAM),
800 MHz as compared to the ThinkPad spec of 667 MHz,
OEM in HP/Compaq systems and otherwise rock solid in my system, is the issue.

[richk]

I think a timing problem that only affects you is less likely than something with your particular RAM modules and how they interact with each other

[JNavas]

I'd say a memory module issue is about the least likely cause.
You have no way of knowing:
how many others have tried and failed,
how well it is actually working for the few posting here,
how it's affected by such things as chipset rev, mainboard rev,
processor type and speed, machine configuration, etc,
and last but not least,
why Intel and Lenovo are on my side of the issue.
Every assumption posted in this thread thus far turned out to be wrong.
"Assumption is the mother of all screw ups."
[Wethern’s Law of Suspended Judgement]

[crashnburn]

Try the 4GB Stick alone? With MemTest+ / MemTest86 in addition to the Lenovo Toolbox test.

[RealBlackStuff]

Every assumption posted in this thread thus far turned out to be wrong.

If you know everything already, why bother asking here?
Besides, I did not assume, I merely suggested to try another module.
But if you want to play petulant, be my guest.
Plenty other forum members who appreciate my assistance...

[JNavas]

If you know everything already, why bother asking here?... Plenty other forum members who appreciate my assistance...

I am experienced and thorough, but I certainly don't "know everything already".
I was just reporting my experience, not asking for anything. Sorry for the misunderstanding.

Try the 4GB Stick alone? With MemTest+ / MemTest86 in addition to the Lenovo Toolbox test.

Thanks, but the 4GB module alone passes Windows Memory Diagnostic and is rock stable -- see my initial report.
However, it does not pass the Lenovo Toolbox memory test. Do you know whether other ThinkPad T61p machines have passed it with a 4GB module?

[crashnburn]

Just out of curiosity.. I never used this Windows Memory Diagnostic. I've always used MemTest+ / MemTest86/+ which were also suggested to me on the forums.

How does Win Mem Diag compare?
How do they all compare?

[AMATX]

Time = $$$

Buy a 4g chip, see if that fixes it. If you don't wanna pony up for a 4 gigger, get another 2g chip.

Memory's cheap; why waste time and hassles over-analyzing the situation.

You've probably already wasted more time on this issue than another chip is worth.

[JNavas]

Just out of curiosity.. I never used this Windows Memory Diagnostic. I've always used MemTest+ / MemTest86/+ which were also suggested to me on the forums. How does Win Mem Diag compare? How do they all compare?

Windows Memory Diagnostic is available in Windows 7 (and I think Vista) just after POST before booting Windows and in System Recovery. After booting Windows it can be accessed in Control Panel. It's also available as a free standalone download. With three levels of testing it seems rigorous and thorough. You might want to check it out.
I haven't used Memtest86 (or Memtest86+) in a long time -- my assumption(tm) is that Microsoft probably does a better job of testing memory for Windows than hackers or even Lenovo (PC-Doctor), and I already know the 4GB module fails the Lenovo Toolbox (PC-Doctor) memory test, which I assumed(tm) is due to the non-standard memory configuration.

Time = $$$
Buy a 4g chip, see if that fixes it. If you don't wanna pony up for a 4 gigger, get another 2g chip.
Memory's cheap; why waste time and hassles over-analyzing the situation.
You've probably already wasted more time on this issue than another chip is worth.

I already have two 2GB modules, 4GB modules aren't cheap, and I'd personally rather throw time at the issue than money.
The one open issue I see is that the 4GB module passes the Microsoft test but fails the Toolbox test.
I'd take another shot at this if someone can verify that a 4GB module can pass the Toolbox (6.0.5514.61) test on a T61p.
Anyone?
Thanks.

[ThinkRob]

I haven't used Memtest86 (or Memtest86+) in a long time -- my assumption(tm) is that Microsoft probably does a better job of testing memory for Windows than hackers or even Lenovo (PC-Doctor), and I already know the 4GB module fails the Lenovo Toolbox (PC-Doctor) memory test, which I assumed(tm) is due to the non-standard memory configuration.

Ahem.

Seriously though, memtest86+ is quite well proven. It's been around for a long time (much longer than the Windows memory test), and has a long track record of being a thorough, reliable test. As if its lengthy history, open codebase, and widespread usage weren't enough: you can run it from external storage (such as a USB stick) to help cut down on the number of variables.

[JNavas]

Seriously though, memtest86+ is quite well proven. It's been around for a long time (much longer than the Windows memory test), and has a long track record of being a thorough, reliable test.

Based on what? Something more than faith and anecdotal reports from hobbyists? What recognized experts have validated it? What major manufacturers are using and endorsing it? What specifically makes it better than the Microsoft test and the Lenovo/PC Doctor test?

As if its lengthy history, open codebase, and widespread usage weren't enough: you can run it from external storage (such as a USB stick) to help cut down on the number of variables.

Likewise Windows Memory Diagnostic. Have you not used it? Did you not read my post? What "variables" are being "cut down"?

But that doesn't address my question:
The one open issue I see is that the 4GB module passes the Microsoft test but fails the Toolbox test.
I'd take another shot at this if someone can verify that a 4GB module can pass the [Lenovo] Toolbox (6.0.5514.61) test on a T61p.
Anyone?
Thanks.

[rkawakami]

When I worked for Mosel Vitelic back in the DRAM era, the Applications group ran a correlation between the then-current version of memtest86+ (forget which one but somewhere in the early 1.xx days) and our Advantest production memory testers. The result was that this tiny and free diagnostic program was able to flag bit failures just as well as the multi-million dollar testers.

One of the standard memory scans that memtest86+ uses is the "Moving Inversion" test pattern. It is one of the more intensive patterns that can be run on a memory within a reasonable amount of time ("Gallop/Galpat/PingPong" is the ultimate; Google it).

A good rule that I've used as a memory test engineer for the last 31 years is that if one program/pattern reports an error that another one doesn't (barring known bugs and/or coverage limits), then you have probably screened out a marginal failure mode. You can go to the memtest86+ website and follow the (broken) links to this page:

http://www.memtest86.com/tech.html (This is the original memtest86 program but applies to memtest86+)

This gives a basic description of what each of the test patterns is designed to accomplish.

re: Microsoft Windows Memory Diagnostic http://technet.microsoft.com/en-us/maga ... light.aspx

Are you using Extended mode to test your memory or just the Standard mode? I was not able to find any online references to what patterns Microsoft is using in either Standard or Extended mode.

Windows Memory Diagnostic supports x86-based platforms, including Intel Pentium and Celeron processors and AMD K6, Athlon, and Duron CPUs. However, the tool can address only up to 4GB of RAM and will not scan beyond that range.

However, it does not pass the Lenovo Toolbox memory test.

Exactly what error(s) does it list? Single bit fails?? Consistently the same address??

I'd say a memory module issue is about the least likely cause. <snip> "Assumption is the mother of all screw ups."

Don't be so quick to dismiss the source of the problem being the memory module (cf: Occam's razor). This can be tested by obtaining another 4GB module and see if it passes/fails the same way as your Kingston.

[JNavas]

I have no interest in an extended debate on memory testing,
especially since you have no experience with Windows Memory Diagnostic.
Had you actually used it, you would know what tests it runs.
I'm guessing you have no experience with PC Doctor either.

As for testing, there is no magic(c) -- test techniques are well known.

The bottom line for me is that Intel does not spec the 965 chipset
for more than 4GB, and I'm pretty sure there's a good reason.

As I've already said, all modules are solid when used within the 965 spec --
it only fails when used outside the spec, and then only occasionally,
a classic symptom of insufficient margin for reliable operation.

As I've also already said, memory is expensive (good memory at least),
and I have no interest is wasting money on a fool's errand with so little payback.

You have your perspective. I have my perspective. Let's leave it at that.

[rkawakami]

I have no interest in an extended debate on memory testing, especially since you have no experience with Windows Memory Diagnostic.

Quite true. I have preferred running memtest86+ for years and years (as anybody who has read any of my memory posts here can attest to). It has caught single-bit errors that POST and PC Doctor has not (see below). By the way, I'm still waiting for your answers to my questions about the memory failures you are getting with the Lenovo Toolbox program. I am curious as to the extent of the failure that the program is reporting.

I'm guessing you have no experience with PC Doctor either.

Quite false (you're sort of new here aren't you ). If you can spare a few seconds, use this site's Search button and plug in "doctor" and my username. Jump back to the 14th page to June 2006...

As I've already said, all modules are solid when used within the 965 spec -- it only fails when used outside the spec, and then only occasionally, a classic symptom of insufficient margin for reliable operation.

That's somewhat of an erroneous conclusion. By definition, a "spec" delimits those conditions which guarantees proper operation at all times. Once you go outside the spec, then you have no reason to assume that the device will reliably operate the same.

[JNavas]

One last time: I have no interest in an extended debate on memory testing.
And I have little tolerance for discourtesy.

[rkawakami]

One last time: I have no interest in an extended debate on memory testing.
And I have little tolerance for discourtesy.

dis·cour·te·ous/disˈkərtēəs/
Adjective: Showing rudeness and a lack of consideration for other people.

Sorry, but I don't see anywhere in my posts that I was rude or inconsiderate to you. Quite the opposite in my opinion. You have (twice) ignored my request for some additional information which could be used to diagnose your problem. I can only conjecture that the reason why you bothered to start this thread was to solve this issue. If I'm mistaken, then accept my apology for wasting your time.

[Harryc]

As the OP stated, "I was just reporting my experience, not asking for anything. ", so lets leave it at that. We're not accomplishing anything here. My advice Ray is to lock this thread...up to you.

[JaneL]

Thanks for reporting your experience with your T61, John, but please don't bite the others who are making comments. There are those here who are equally as skilled, experienced and knowledgeable with as many or more years in the business as you have.

I'm going to follow Harry's advice and lock the thread as it's served it's purpose.

[BillMorrow]

One last time: I have no interest in an extended debate on memory testing.
And I have little tolerance for discourtesy.

my irony detector is going off and pointing toward the northern end of the SF peninsula..!