Memtest intermittant failure
Reference Post 7427 : Cloning HDD
I'm still having problems Backing up and Verifying my system data to an external (firewire connected) drive.
The docs suggest checking my RAM using MemTest86+ (v4.0)
The test was run for several hours and it failed with the 2 existing modules (512 Mb) installed.
When the test is rerun using just a single module and even changing it through the 3 sockets, the test passes ..this is with either module being tested.
It's only when 2 modules are used together that it fails badly.
I'd like to fix this anomaly before moving to something else that might be causing ATI Validation will not work.
Any ideas of what is causing this Memtest failure?
MB = Asus A7n8x-x, Cpu = Athlon 2500+ (Barton); RAM = 2 x 512 DDR400 PC3200

- Anmelden, um Kommentare verfassen zu können

Further to you query --
Both module sticks are of different manufacturers: 1 - Samsung, 1 - Infineon
Asus MB supports single channel only.
Interestingly, CPU-Z shows the Infineon timings table with 3 settings (133, 166 ,200Mhz), Samsung shows only 2 (166, 200 Mhz).
I wonder if only being able to run the PC with 2 sticks and CPU set at 133 Mhz, and Bios Sys Performance set at Auto has anything to do with the above?
The CPU is rated at 2500 and can push it to 3200 via enabling frequency to 200 Mhz (166 is nominal, 333 FSB).
Currently I have it set at 133 Mhz which slows the CPU down so it can boot into Windows.
(at times it has to be set even lower (100 Mhz) to accomplish the same result)
As mentioned, Bios System Performance tab must be set to Auto rather than SPD for the computer to work with any stability.
PC will run sluggishly on just a single stick, but only at 133 Mhz and below.
Bios does allow voltage to be bumped up and .1v made no difference, either with 1 stick or 2.
Previous Memtest was run at _200 Mhz_ (overclocked?) and failed with 2 sticks -- singles ok.
Haven't tried Memtest run at 133 MHz as I felt that it was way below it's ratings and thus would probably pass ok.
Having only just recently run Memtest on these modules, I do now wonder if they have always been deficient (failed), and are just now showing signs of fatique (because they are not perfectly matched) -- one throwing the other off via their SPD specs ??
But before throwing good money after bad and purchasing new matched pair of RAM, I would like to know for sure whether it's the RAM or Motherboard, or CPU. I wish there was a good piece of software out there that could pinpoint this problem for me. Most seem to only be component-specific not testing the whole package.
- Anmelden, um Kommentare verfassen zu können

According to the Asus site the A7N8X-XE is a dual-channel board.
http://asus.com/product.aspx?P_ID=5vee2zoCMBLmLPzP
The fact that you have to slow the machine down to boot into Windows indicates something is very marginal on your system and it certainly could be RAM.
Diagnostic programs can only do so much to isolate a fault. The memory diagnostic tests by writing data and then reading it back and then doing a comparison. Unfortunately, anything that causes a bit to be altered will indicate an error. Usually the fault is a bad chip on a stick but other causes can be bad connector, bad cache memory, timing problems, even a bad CPU. In a case such as yours my first attempt at this stage would try to beg, borrow or buy some good matched RAM to try. Memtest is telling you it is having a problem but now you have to try and narrow down the possible causes.
- Anmelden, um Kommentare verfassen zu können

My board is an A7x8x-x, not xe. So in my case, single channel : 3 slots, no colour coding.
http://www.asus.com/product.aspx?P_ID=od83seU8RaneL4Cq
It is interesting that I can push single sticks to 200 Mhz (max) and they do not fail under Memtest, however, when they are paired (at 200 Mhz) they fail. So finding which stick is a fault, Memtest can't seem to parge it out.
If it's a bad connector, cache memory then I may need another MB.
CPU? -- same thing, a new processor.
One would think in this day and age there would be software available to find the culprit without using trial and error. Recycling ? bah -- this is a throw-away world isn't it ;-)
Thanks for your input !
- Anmelden, um Kommentare verfassen zu können

Memtest puts the address of the failing location in the report. See if it helps to make sense of the problem.
If it appears that the failures are on only 1 stick then swap them around and see if the failing addresses change. If the addresses are on both sticks then it appears that there is a conflict causing problems with the whole range of addresses.
My mention of the motherboard, cpu etc were just to demonstrate that almost any part of the PC could cause a memory diagnostic to show RAM failures. The odds are that a bad stick is the problem with most memory failures.
- Anmelden, um Kommentare verfassen zu können

Thanks.
I ran Memtest yet again overnight (at safe 133 Mhz -- max it will run and still boot). Both modules did not fail even after 18 passes! They only fail above the 133 Mhz threshold -- and that's only when in a pair. Yep, my feelings is that there is some compatibility issues when pushed too much. So it seems the system can't handle 2 sets of (different?) timing parameters when over 133 Mhz. I'm thinking that one of the modules is pooping out and affecting the timing of both, especially when Bios is set to SPD. Does that make sense ?
None of my friends have spare DDR's around so I think I'll just order a pair of 1 gig and run them in my single channel Mboard. I guess the single channel Mboard will utilize a dual channel memory modules as single sticks (as they will come pre-matched) ?
- Anmelden, um Kommentare verfassen zu können

SPD is Serial Presence Detect and is used to set the RAM timing parameters from data included in the RAM stick. If the 2 modules are providing the BIOS with differing values it can cause problems. It depends on the sticks, the motherboard etc. If a stick, regardless of how it is labelled, happens to be able to handle different timing then it won't be a problem and some motherboards are just more tolerant of things than others.
Buying the dual-channel modules will ensure that you get matched modules and they will work. Just buying single sticks of the same model from the same manufacturer would be OK as well. Is there much of a premium these days on dual-channel sets?
You could also try going into BIOS and manually setting the memory timings to slightly slower identical timings. It could be that the problem isn't really timing but some other parameter of the RAMJ. Chasing all of this may be flogging a dead horse and the quickest and best solution is to buy some RAM. I have found the Crucial site to be handy in finding RAM - just put in your MB and they tell you what is recommended.
Good luck.
- Anmelden, um Kommentare verfassen zu können

Their individual SPD values seem to be the same according to CPU-Z and Sandra but, yes I suspect their parameters are different than as advertised, perhaps somehow changed since they've been in the computer for a few years now.
I can adjust the timings but this Cpu/RAM combination is synchronus and apparently won't work otherwise without getting well into some the risk (raising voltages, etc). Perhaps I can play with this after obtaining replacement RAM (?)
All this time I thought my Mb accepted DDR400 pc3200 (per the owner's manual). But after reviewing some old forum postings (and now experience) I find that this Mb will not take kindly to being pushed past its limits and does not like DDR400 modules. I guess it's part and parcel of buying a stripped down version of their best seller, huh -- live and learn.
To reinforce the specs on module types, I went to both the Crucial and Kingston sites and Kingston recommends only DDR333 PC2700 with a max of 2 gigs -- 3 gigs (max for this Mb) for even lower frequency values.
Crucial also recommends only DDR333 PC2700 but oddly enough it says my Asus Mb is dual channel -- which it isn't (modules currently in the computer are both DDR400 PC3200).
But I am also coming to the conclusion that I've spent way too much time on this "problem". I'm again becoming like a dog with a bone.
I may pick up some modules and check them out in this machine. If the fault proves to be somewhere else, then onward to either another Mb or complete system. Hmm, I wonder if just updating to another Mb should be the way to go anyway ..as this one is getting long in the tooth ?? And can I even reuse DDR400 modules in recent Mb's ??
Again, I appreciate your input !
- Anmelden, um Kommentare verfassen zu können

Virtually every motherboard made in the last few years is double-rate DDR2 (now DDR3 is taking hold). In other words, your DDR is obsolete and won't be useable in a new board.
A computer store I deal with never recommends the user who isn't looking for absolutely the fastest performance to buy the fastest RAM for a board - more stability with 1 step lower than the fastest is their point.
What about buying the RAM at a local store and either have them install it and run memtest for a while to demonstrate all is fine or making a presale arrangement that you can return the sticks if you still have a problem?
Note that since DDR is obsolete there are likely local stores that may have good used DDR available at a low cost.
- Anmelden, um Kommentare verfassen zu können

Whahoo! I was finally able to do a full system Backup WITH Validation to the external (firewire) disk last night. Obviously lowering the CPU frequency to 133, setting the latency timing up one notch to 3.0 and +.1v to 1.7v seemed to have worked
..at least for now to buy some time while I search out my options.
Yes, I figured that the DDR stuff was most likely finished in this marketplace -- worth a try -- thanks.
Are there any good versatile Mb's that you would suggest upgrading to (using AMD and new-type memory)?
As for replacement RAM: locals want $100/2 gigs + $70 for install & test -- no returns on OTC cash and carry!
Seems there's still good demand for DDR (especially 512 +) as folks are trying to keep what they have alive.
No used stuff around except at an electonics recycle depot : used Dell PARITY (9 chip-which my Mb won't take); DDR400 PC3200 -- $25, no warranty. He had a big box of loose beater modules ..I don't think so.
But there's a National online shop (NCIX) I normally deal with and for new pricing they are quite reasonable, but limited when it comes to older product.
- Anmelden, um Kommentare verfassen zu können

That's great. If you haven't, try running it with the latency at 3 or possibly 3.5 if available, 1.7V but at 200 Mhz. The TI archive validation procedure is a good test both for reliabilty and speed. If it seems to work at 200 then let memtest run overnight.
Also, make sure you can validate the image using the TI rescue CD. It is Linux, uses different drivers, and it is essential that it be able to handle the archive when doing a restore of the active partition.
No real comment on motherboards other than I only buy Asus but might consider a Gigabyte. I tend to find when upgrading that you need to do board, processor, RAM, likely graphics and possibly power supply. I tend to buy a basic box with the board etc and then add my own disks, optical drive, whatever although these days with highly integrated boards that's about it.
Like everything else buying a motherboard can be a good or not so good experience and some models are better than others. You are likely to have less trouble if you don't rush out and buy the latest and greatest but get a more mature one. These may have gone through a revision cycle and are much more likely to have had some BIOS updates to fix issues. You'll also have the benefit of reading comments on-line but remember it is typically only the people who have grief that post comments with the happy ones remaining silent.
- Anmelden, um Kommentare verfassen zu können

Just about every permutation has been tried, including running ATI @ 200 Mhz. The computer worked ..but just barely and for only a little while.. but of course failed Backup AND Validation. Only my current setting of 133 Mhz will work satisfactorily both in bootup and ATI processes.
Good point about Validation using the CD. I'm going to try that as well as it is ultimately what I will need to do during the Restoration process.
Hah.. yes, not many folks seem to surf fixit forums unless they have a problem with something. All others are just helpful interested folks I guess? -- thank god for that !
As mentioned, NCIX is the site I usually go to for my parts. I also found a very useful thread about putting together a computer using recent suggested components. But the tread's mostly populated by younger folk who are into gaming (I mostly just surf, and do photoshop and video manipulating).
Anyway, the site is at:
http://www.redflagdeals.com/forums/ncix-custom-built-computers-1000-700…
As it's good for research in all types of purchasing, lot of amazing deals have also come out of that site.
In the past I had completed a couple of full assemblies and they weren't difficult. The much added bonus is in knowing exactly where everything is and being able to self-augment the system. The cases were also big server boxes so there was lotsa room to move around and with lotsa airflow ..just like this current one.
So it's nice to know that some places (like NCIX) will also assemble your online parts purchase with a put-it-altogether, test, and warranty option ($50).
But as you suggest, the main trick is picking out the parts and having them work the first go-around.
After a few years of being out of market it's quite easy to be out of the loop too.
- Anmelden, um Kommentare verfassen zu können

Seekforever wrote:Also, make sure you can validate the image using the TI rescue CD. It is Linux, uses different drivers, and it is essential that it be able to handle the archive when doing a restore of the active partition.
Followup :
Hah.. you must have realized this was coming...
Using the Rescue CD (full version), I tried this out and could not access the backup file or any other information on the USB external drive (I:).
However, oddly enough the complete folder shows up on an internal logical drive (E). -- why there? I can Validate at that location ..but it still isn't via the USB drive (I).
ATI version is Home 2009 (bld 9770)
- Anmelden, um Kommentare verfassen zu können

Do some investigating and see if what is showing up as E isn't really your USB drive. Does the volume label check out for example.
There is often confustion with drive letters because Linux doesn't use the same mechanism to assign them as Windows does.
- Anmelden, um Kommentare verfassen zu können

Gosh! I forgot it was Linux. Sure enough the drive designations are interchanged when booting from the CD. D is E, and E is I. Well, son of a gun -- I guess my attention was on finding Windows I:
Everything now seems to show up -- thanks for the tip, bud !
Back on the topic of Hardware :
I can't seem to let go of the idea that the Memory is the culprit here. As a test was done with single modules in any one of the slots and run past the safe frequency of 133 MHz (ie; 166~200), Windows booting fails, just as it does when in pairs. The machine used to run fine at the higher frequencies ..with both modules.
Jeez, maybe it is either the CPU or Mb afterall?
Even at this 'safe' 133 Mhz, the performance seems to be ever so slightly deteriorating (slow surfing with other IE windows opened, plugins hard to run, screen refresh is "roiling" at times when under heavy load).
We have Memtest for testing memory, now if we could just find some good software tests for the CPU & the motherboard ..
- Anmelden, um Kommentare verfassen zu können

With symptoms as you describe it could be almost anything. However, usually when RAM fails it either causes a serious problem or no problem (may be in a location that really doesn't matter at all). A PC does not know if RAM data is bad it just uses what it gets. If a disk gets bad data because the CRC is wrong it will do retries to get it right. The retries will slow down the system considerably or it will give up and tell you that the file is unreadable.
A bad disk or MB will certainly cause boot failures but so will flakey RAM, cables, etc.
When the PC starts up, check to see if it has detected all the RAM. If the RAM sizing check fails on the second module it will get eliminated. Also see what Windows is reporting as the RAM size.
There are system tests out there but the more popular one's name escapes me right now. If I think of it, I'll reply. Google is your friend and you should be able to find some. However, if something is reported as bad don't accept it as the gospel truth right away unless it is confirmed by some other method. In my previous life as a system manager the first thing we did with new hardware was to run the supplied diagnostic on a hard-copy terminal and file the report away so we knew what the results should be. Diagnostic programs are like everything esle - they have their faults and quirks too.
- Anmelden, um Kommentare verfassen zu können

Good advice -- thanks.
Borrowed copies of Passmark and Prime95. After hours of running neither one picked up any hardware problems. But again, this is with CPU set at the safe 133 Mhz --- still can't test when CPU is at higher frequency as the sys won't boot into Windows in order to lauch and retest with those diags.
Aarg!
Another thing that is starting to cross my mind is perhaps some sort of software pollution or hyjacking is taking place. There seems a ****load of software starting up with Windows. When I try to multitask anything, the computer still slows to a crawl. I've run malware checkers and run regcleaners that detected stuff, but still not helping.
Is there software that will help in leaning out the riff raff?
- Anmelden, um Kommentare verfassen zu können

Finally found a new memory module and tried it in the Mb.
Same result : CPU still won't run at it's posted speed, no matter what I do inside of Bios. Checked over the power supply and Bios reports all voltages and temps ok.
As a last ditch attempt at trying to repair Windows (system) I tried to Restore and Validate ATI from Rescue CD.
Took over 6 hours to 'restore'! ..but only completed to 90%.
Had to reboot system and used Windows System Restore instead to get it back to previous (useable) state. Most everything on the C: drive is now in duplication.
Don't quite know how to solve that situation.
I think what I would like to do now is to take note of any software programs that I need & see if I have them on disks somewhere, then reformat the C: drive with a fresh Windows XP install. I can then see how everything plays out vis a vis CPU, ATI, et al.
Question: what would be the best approach to saving and getting the major settings, IE faves, OE emails, etc. back in working order after the reinstall? Do I just save Docs & Settings? -- I don't want to also include any garbage and faults that may affect an new complete Windows install.
Thanks.
- Anmelden, um Kommentare verfassen zu können

You can export your favorites and OE mail. Somewhere in XP I think there is a wizard that lets you do what you want or most of it. It is intended for changing the OS etc.
Can you open your existing image by Windows Explorer or by the TI mount command. If so, you should see your files and be able to copy them out of the tib file. However, since you've had trouble, I'd consider this a secondary backup.
Your data files are the important ones since they aren't on an install CD anywhere. My first step would by to attach a USB drive and copy them using WIndows Explorer to the USB drive. I would also manually copy the exported files from OE etc to the USB drive as well.
You could also create and validate another TI image for good measure.
- Anmelden, um Kommentare verfassen zu können

Ok, so I might also do a fresh install of the OS into a new active partition and keep the old C: drive handy in case it doesn't pan out. This way I might be able to finally prove a bad HDD.
But..
If I do decide to use the existing 'TI Backup image' (which is now stored on USB drive) for the new active partition, can I just use Windows to "copy" it over to the partition or is there another TI Restore process I should be using?
- Anmelden, um Kommentare verfassen zu können

You can't just copy via Windows and have the drive bootable.
- Anmelden, um Kommentare verfassen zu können

Ok, how about if I made another active partition (on the second drive), then used ATI to Restore a "banked" (stored on USB drive) system image to that partition -- I could then also use a tool such as BootMagic to make my computer dual booting ?
At POST the computer would give me a choice of whether to use the existing C: drive OS or the new partition OS.
This could negate the chore of swapping the ribbon cable Master / Slave configuration.
I could also allow me to "toggle" between OS examples in order to test whether my old C: drive is in fact interferring with my cpu speed and bootup problems.
Do you think this idea will work?
- Anmelden, um Kommentare verfassen zu können

If you can setup the dual boot then it will work.
- Anmelden, um Kommentare verfassen zu können