|
 |
| |
| |
테스트결과
김형우
조회 : 1733 ,2001/11/02 00:00 |
|
 |
|
좋은 정보 주셔서 감사합니다.
테스트한 결과
Bank 0 Bank 1
RAM (a) only pass pass
RAM (b) only pass pass
(a) + (b) pass
(b) + (a) not tested
모두 에러 없이 패스했습니다...T_T
램은 모두 삼성 PC133 256MB 단면이고,
램 설정은 fast, 2T-6T-3T 였습니다.
테스트 마치고 2000 들어가서 윈앰프 들으면서 웹서핑 하던중에
또 사운드 멎으면서 마우스 키보드 먹통으로 웜부팅 안되는...
화면 그대로 얼어버리는 증상 바로 나타나더군요...
-_-;;;
ːOCWORKSBENCH ? 인가 하는곳의 ECS 포럼에서
ː
ː미스터 애슬론이란 사람이 쓴 글입니다.
ː
ː제 보드는 정상이더군요 :)
ː
ː참고하시기 바랍니다.
ː
ː posted 31 October 2001 17:56
ː--------------------------------------------------------------------------------
ːFAQ 1.1 on K7S5A (M830) Data Corruption
ː10/31/01, Original V1.0 release.
ː10/31/01, New V1.1 Update with solution!
ː
ːCONTENTS:
ːPART 1: Overview
ːPART 2: First things to check
ːPART 3: The cause of the problem and “non-causes” of the problem
ːPART 4: The solution
ːPART 5: Summary of testing and research on the problem
ː
ː
ːPART 1: Overview
ː
ːProblem:
ːSome K7S5A (and PCChips M830) motherboards suffer from a serious data corruption problem. The problem is common with Athlon 1400 processors and occasional with Athlon 1333/1300 processors. The problem is rare (but possible) with other speeds.
ː
ːCommon symptoms:
ːMemtest86 Errors, 133/133 failures and problems, CONFIGMG Errors, Crashes, Blue Screens, Windows Protection Errors, OS installation failures, corrupted CD burns, Windows Registry corruption, general data corruption, bad BIOS flashing (can result in no-boot condition), etc.
ː
ː
ːPART 2: First things to check
ː
ːQ: How do I know if I’m affected by this problem?
ː
ːA: Download and run the Memtest86 program. (www.memtest86.com) If you can run through all tests without any memory errors, you do not have the problem and you should have a rock-solid setup. When the K7S5A is working, it works great.
ː
ːQ: Memtest86 reports errors, but I still think my system is “rock solid.” Do I really have a problem?
ː
ːA: It depends. The data corruption is generally “pattern sensitive” which means that when certain patterns of information are written to memory (especially cache memory) the data is corrupted. The patterns that are corrupted may or may not be written frequently by the OS or applications.
ː
ːSo yes, this is a real problem if you have data on your hard drive, or process data on your computer that must be 100% correct. This pretty much includes anyone using the computer for serious work.
ː
ːNo, maybe it’s not much of a problem if you just play some games and surf the ‘net, don’t have any data that is important to you, and don’t mind the possibility of reinstalling your OS eventually after you get file corruption.
ː
ːAlso, because the problem is related the driving impedance of the CPU, and the driving impedance changes with temperature, the number and type of errors you encounter may be strongly dependent on CPU temperature.
ː
ːQ: OK, I downloaded Memtest86, made the floppy disk, and I have the problem. Is my motherboard bad? What should I do?
ː
ːA: It’s possible that your motherboard is bad, but you must first check several other important components of your system before you can be sure -- you must check your BIOS Settings, Memory, Power Supply and CPU temperature. See below.
ː
ːQ: What BIOS and BIOS Settings should I use?
ː
ːA: Be sure your memory timing is set to Normal or Safe.
ːLeave all other memory settings at their defaults. If you’re not sure, use the clear CMOS jumper on the motherboard to reset to factory defaults. Then, set your CPU speed to 133/133 (or 100/100 for a 200FSB processor) and try again.
ː
ːA few users have reported improvements with the 011016 BIOS although this has not been confirmed by any controlled tests. Still, you may want to try this BIOS version.
ːIMPORTANT: You must have a stable system configuration before you update the BIOS or you could end up with a corrupted BIOS that may not boot. Try setting your CPU speed to 100/100 and use Memtest86 to see if your system is OK before you update. If you still have errors, go into the BIOS and turn off the L1 (and L2) CPU cache. This will make your computer slow, but should allow you to update without error in most situations.
ː
ːQ: How do I know if my Memory is OK?
ː
ːA: Confirm that your memory works properly by changing your processor to either a slow Athlon/Duron (1000MHz or less), or to an Athlon XP processor. In most cases, a slow processor or an Athlon XP processor will allow the motherboard to work well enough to show that the memory is OK. Be sure to use the same 133/133 CPU speed settings if possible. If you continue to have the same problems in Memtest86, check the Power Supply and CPU cooler, and try your memory in another computer, if possible.
ː
ːIf you don’t have access to another processor, just try setting you CPU clock to 100/100. This will allow you to test your system at a lower speed that just might pass the test. Many “bad” motherboards still cause errors at 100/100. If you still have errors at 100/100, you might have low-quality memory, a bad motherboard, or some other problem. You’ll have to keep testing to find out.
ː
ːBe sure to use high quality PC2100 or PC133 Memory. If possible, get memory from the approved list in the manual or on the web site.
ː
ːThe data corruption problem has been confirmed with both PC2100 and PC133 memory, but is generally less severe with PC133 memory.
ː
ːQ: What Power Supply do I need? Should I really buy a 350W or 400W super supply for $$$?
ː
ːA: You need a good power supply that can provide 5V with enough current to run your motherboard and all other components:
ːThe best option is to buy an “AMD Approved for Athlon 1400MHz” Power Supply. Note that just “AMD Approved” is not enough, because it might only be approved for some slower Athlon (like Athlon 1000) and that’s not going to work if you have a 1400/1333/1300 processor. AMD’s web site (www.amd.com) has a list of every power supply they have tested and approved. There are plenty of units that support the Athlon 1400, and they can be found on the Internet for under $40 (I have fully tested the AMD approved 1400 power supply, model KME PX-300W, that comes with the Stratitec CPUBuilders ATX mid-tower case at www.samsclub.com for about $40, and I recommend it for anyone in doubt). If you don’t have one of these, get one.
ːIn general , you will not need to get a super 350-400W power supply. Unless it is rather expensive, one of these is often not AMD approved at all, or is not as good as a basic 300W AMD Approved unit. The possible exception where you may need to spend the $$$ on an AMD Approved 350W+ supply is when your system is loaded with other power-hungry items like a super fast Video Card (or two for dual monitor support), multiple 7200-10000RPM hard drives, multiple CD-RW drives, SCSI cards, etc.
ː
ːQ: What CPU temperature is OK and what CPU Cooler do I need?
ː
ːA: The Athlon processors at 1000MHz to 1400MHz are rated to operate correctly at temperatures up to 95C (about 200F!). Typical CPU coolers and case setups will keep the temperature of the CPU in the 45-65C range and this is very acceptable. If the BIOS reports a CPU temperature above 65C, you should consider a better CPU cooling system, and/or a better case cooling system, and make sure you have a good thermal grease/thermal pad on your heatsink where it touches the CPU.
ː
ːQ: What CPU voltage is OK?
ː
ːA: The Athlon Thunderbird and XP are specified to work between 1.65V and 1.85V. The BIOS should show the Core Voltage within this range. 1.74-1.76 is the most common and is ideal. If your voltage is less than 1.65V or more than 1.85V, you may have a problem of some kind related to the CPU power supply.
ː
ːSome users have noticed that the reading for the Core Voltage jumps between different levels in the BIOS. This can happen even on good motherboards and is not a concern unless the voltage jumps are .1V or more (i.e. if the voltage jumps from 1.66 to 1.76 to 1.66 to 1.76, etc.)
ː
ː
ːPART 3: The cause of the problem and “non-causes” of the problem
ː
ːQ: I have checked and double-checked everything in PART 2 of this FAQ and everything looks good. What’s causing the Memtest86 errors?
ː
ːA: Most likely, a “bad” motherboard.
ː
ːQ: What exactly is wrong with the “bad” motherboards?
ː
ːA: Two things:
ː1) 99% of the problem: Incorrect “Driver Impedance.”
ːExtensive testing has found that “bad” motherboards do not correctly program the “AMD bus driver impedance” on the CPU. These “bad” boards program the CPU to approximately 60 Ohms, while the AMD “Revision Guide” from April 2001 calls for 40 Ohms. This mis-match allows the CPU’s to sometimes work at lower speeds, but not at full speed.
ː
ːAn additional variable is that this motherboard value is supposed to match the “impedance” of the actual circuit board (PCB) material of the motherboard itself. Since the circuit boards vary from batch to batch, there are some motherboards that work pretty well, even with the problem, and others that really don’t work.
ː
ːThe Athlon XP and Athlon MP have special capabilities that allow them to work at either 40 Ohms or 60 Ohms. If your Athlon XP or MP processor is not working in your board, check the items in PART 2 of this FAQ again, and check the noise problem below.
ː
ː2) 1% of the problem: Noise.
ːThe Power Switching FET’s and Diodes in the CPU power supply circuit generate large amounts of noise. The motherboard has a “leak path” that passes this noise directly onto the 12V power line. Typical boards have 3-4 volts of “bursting high-frequency” noise on the 12V line that may be the cause of occasional data corruption. This is not a big problem usually, but can result in a few errors, usually in test #7.
ː
ːQ: Why do some Memtest86 tests pass and others fail?
ː
ːA: Memtest86 will typically find errors in Test #3, #4, #5 and #6. These tests seem to be more sensitive to the impedance mismatch problem.
ː
ːQ: What are the errors that Memtest86 is reporting and why does it find them?
ː
ːA: Memtest86 usually reports pattern errors and block move errors in large quantities. These problems are usually caused by the degraded signals that result from the impedance mis-match.
ː
ːQ: I pressed ‘c’ in Memtest86 and turned the Cache memory to OFF. Then all the tests passed, why?
ː
ːA: It appears that the CPU’s cache memory is not able to operate correctly on a motherboard with the problem. Turning the Cache memory off makes Memtest86 only run the non-cache tests. These tests communicate directly with the main system memory, and that always appears to work correctly.
ː
ːQ: Does this mean I can just turn off the CPU cache in the BIOS and the problem will be fixed?
ː
ːA: Possibly, but don’t bother. The system will be slow like a Pentium 100 (or worse) with the cache turned off.
ː
ːQ: Does anything else in the system have an effect on the problem? What about my video card, hard drives, OS, etc?
ː
ːA: As far as testing can determine, nothing else has any significant impact on the problem.
ː
ːQ: Could it be the clock speed? My system shows 1394MHz instead of 1400MHz and that’s not quite right.
ː
ːA: The motherboard actually runs at 99.5% of the standard 133.33MHz. This is caused by the limited clock options built into the ICS clock chip used on the board. This chip has a 133.33 setting that is used by the motherboard, but there is a -.5% speed variance at that setting. Other settings, like 133.9MHz have a +/-.25% speed variance, but would result in the CPU running a bit too fast.
ː
ːQ: I got the CONFIGMG errors while using DDR memory, then I switched to PC133, and the problem went away. Doesn’t this mean that the problem is with DDR memory, or at least my stick of DDR memory?
ː
ːA: No. Running the motherboard with PC133 memory usually reduces the number of errors, and this helps make the CONFIGMG error less frequent. Extensive testing has found that almost all bad motherboards that have errors with DDR memory will also have errors with PC133 memory. The reduction in memory errors does not appear to be directly related to the different memory, but the motherboard behaves differently when using it.
ː
ːQ: Could it simply be that the 1400MHz processor takes too much power and that’s the cause of the problem?
ː
ːA: No, the motherboard has no problem supplying the power for 1400MHz. This has been verified by careful measurements with a high speed oscilloscope. In addition, additional power loads of 14W (equivalent of about 200MHz in additional CPU speed) have been added to the core voltage while running an Athlon 1400, and the motherboards have no problems with that either.
ː
ːQ: What about the theory about different power FET’s on good and bad versions of the board?
ː
ːA: Not correct. Both the IR and NEC power FET’s used on different versions of the boards are able to handle the power requirements with no problem. Testing has found that the noise signature of these two types are a little different, but special motherboard changes have been made to greatly reduce their noise (at the expense of more wasted power) and the noise does not appear to be related to the problem.
ː
ːQ: Can I just raise the Core voltage to 1.85V to solve the problem?
ː
ːA: Generally, no. A higher core voltage usually makes the problem worse not better. Feel free to try if you’d like, the processor is designed to work at 1.65 to 1.85 so it shouldn’t hurt anything.
ː
ːQ: So, is there some kind of clock quality problem?
ː
ːA: No. The “spectral signature” (FFT) of the clocks on good motherboards match those on bad motherboards.
ː
ːQ: Even if it’s AMD approved, couldn’t it still be some problem with the Power Supply?
ː
ːA: Not likely. A bad power supply can definitely cause problems, but no power supply can fix the data corruption problem. Testing on more than 8 power supplies, and testing with lab quality power sources has confirmed that there is no power supply solution to the problem.
ː
ːQ: Why isn’t the problem just bad memory? It’s Memtest86 that finds the errors and it says they are memory errors.
ː
ːA: Memtest86 is actually testing the CPU, Cache and Memory all at the same time, and it can’t tell exactly where the problem is. If the memory really is bad, Memtest will find errors at specific locations in the memory chip, and should find them on any motherboard. When Memtest finds the same “errors” on every memory chip, but only on one motherboard and at one CPU speed, it points to the motherboard and CPU as the problem.
ː
ːQ: I’ve seen references to a Memory voltage adjustment that will move the memory from 2.5V to 2.6V. Will this help?
ː
ːA: Not likely. A higher memory voltage is really only good when you need the memory to run faster for overclocking. Extensive testing with a range of memory voltages has found that it has no impact on the problem.
ː
ːPART 4: The Solution
ː
ːQ: So what can I do?
ː
ːA: At this point there are only three real options if you have the problem:
ː
ː1) Get another processor. Most bad motherboards will work correctly with a slower Athlon (1200MHz or less) or an Athlon XP processor.
ː2) Get another motherboard. Try exchanging your board. Ask for a motherboard with the “driver impedance” problem fixed.
ː3) Wait for a fix. If you have really steady hands, good eyes, and can solder small parts, you might be able to fix your motherboard yourself, see below.
ː
ː
ːQ: Can I just set the BIOS to CPU speed of 100/100 to fix the problem?
ː
ːA: Sometimes. Testing on a couple dozen motherboards has found that many will work correctly at 100/100, and many will not. If you have good results in Memtest at 100/100, your board should be safe to use.
ː
ːQ: Is there anything I can solder/unsolder from the board that might make the problem better?
ː
ːA: Yes, but not easily. There are two resistors on the motherboard that control the drive impedance. Changing these to a value close to the proper 40 Ohm value fixes the problem. Unfortunately, they are so small and hard to solder that it’s just about impossible to do unless you have an ultra-fine-tip soldering iron, and a very steady hand. Plus, if you mess up and get a little solder where it’s not supposed to be, the motherboard can be destroyed when you turn the power on. To make matters worse, the resistors are not clearly labeled or numbered on the board. They are side by side, inside the CPU socket area on the corner that is the farthest from the AMR slot.
ː
ːAlso, there is a small capacitor that causes a noise leak path from the CPU power supply onto the 12V line. Inserting a ferrite bead (or small inductor) in series with this capacitor will choke the noise down to around 1V on the 12V line and can clean up any remaining errors that might persist after the resistor fix above. Again, you’ll need a very steady hand, a small soldering iron, and you’ll need to find a tiny surface mount ferrite bead or inductor to install. This capacitor is located just off the corner of the CPU socket that is closest to the AMR slot. It is a large (thick) surface mount capacitor, not clearly marked, but right next to the two through-hole diodes that are right below the through-hole 16pin power supply control chip. One side of the Capacitor is connected with a heavy trace right into the heart of the CPU power supply circuit. If you unsolder the Capacitor, and then solder it in an upright position, you can then solder the inductor in an upright position where the other side of the capacitor was and bridge the tops of both together.
ː
ː
ːPART 5: Summary of testing and research on the problem
ː
ːJust about every possible component has been switched on every motherboard in every combination that makes sense. The problem always follows the motherboard, and does not appear to be a problem with any of the added components.
ː
ːOn all tested “bad” motherboards, the problem is completely eliminated (or at least reduced by 99.9%) by changing the resistors to a value close to the AMD recommended 40 Ohms.
ː
ːOne “good” Motherboard, a Rev 0 board, always works perfectly and is 100% solid, passes every Memtest86 test with every item listed below. This motherboard has 60 ohm resistors, but works anyway. Since this is the only motherboard that works with 60 Ohm resistors, maybe this is just a “lucky” board.
ː
ːWith 1.4GHz processors, the 26 Rev 4 and Rev 6 “bad” motherboards fail Memtest86 at 133/133. About half also fail at 100/100.
ː
ːWith Athlon XP and MP processors, all 27 motherboards (“good” and “bad”) are 100% solid with every item listed below.
ː
ːDifferent BIOS versions have no impact on the problem.
ː
ːMemtest86 always finds errors in Tests #3, #4, #5, and/or #6 on “bad” boards.
ː
ːThe Athlon MP 1200 runs rock-solid and passes all Memtest86 tests, even at 150/150, on 5 randomly tested “bad” motherboards. All DDR memory was tested at this speed (300MHz) as well, and works perfectly on the “bad” motherboards when using the MP processor.
ː
ːA stand-alone lab power supply was used to supply all power lines for the motherboard to ensure that none were lacking from the standard ATX supplies. No voltage variations on any power lines had any impact on the problem.
ː
ːA stand-alone lab power supply was used to adjust the 3.3V and memory 2.5V/1.25V on the motherboard itself, with no effect.
ː
ːAll the main clock and power signals into and out of the CPU were carefully compared between the “good” board and “bad” boards. No conclusive differences could be found.
ː
ː
ːConfirmed failure symptoms:
ːMemtest86 errors (of course)
ːCONFIGMG errors
ːCD-RW write corruption
ːWindows Blue Screens
ːWindows Registry corruption
ːOS install failure (Linux)
ːWindows Protection Errors
ːWindows lockups
ːBoot failure
ː
ː
ːTest Setup:
ː
ː27 K7S5A Motherboards: One Rev 0, two Rev 4, 24 Rev 6 boards.
ː
ː10 CPU’s: Two Athlon 1400 266FSB, One Athlon 1400 200FSB, One Athlon XP 1800+, One Athlon MP 1200 266FSB, One Athlon 1200 200FSB, One Athlon 1333 266FSB, One Athlon 1300 200FSB, One Duron 650, One Duron 1000
ː
ː15 Memory sticks: Two Samsung PC2100 DDR 128MB, Four Kingmax PC2100 DDR 256MB, Two Winbond PC2100 DDR 512MB, One Crucial PC2100 256MB, One Kingmax PC133 256MB, One PC166 256MB, One PC166 128MB, One Micron PC133 256MB, Three generic PC133
ː
ː3 BIOS versions: 010920, 010911 (OC), 011016
ː
ː9 Power Supplies: One Antec 350W AMD Approved 1400MHz, One 300W AMD Approved 1400MHz “Deer” brand, One 250W generic, Six 300W AMD Approved 1400MHz “KME PX-300W”
ː
ːEquipment:
ːProtek 506 4 digit Digital DMM
ːTektronix TDS3034 300MHz digital oscilloscope
ːVarious lab grade power supplies and other equipment
ː
ːSoftware:
ːMemtest86 2.8
ːWindows 98SE
ːWindows Me
ːWindows XP
ːRedhat 7.1
ːRedhat 7.2
ː
ː
ː
|
|
|
|
 |
|
|
|
| |
|
|
 |
|
