Dann-O, here's the marvelous new AMD MP platform that was supposed to cause so much damage to Intel's server market, ASPs, etc. Sounds like a round robin problem, you name it, among AMD, Tyan MBs and power supplies. Would you trust even your Mom and Pop candy store business to one of these, much less a Fortune one thousand company?
Jul 23, 2001, 07:17 PM
By: Dan Mepham
Last week, we started to question the Tyan Thunder K7’s apparent quirkiness and lack of stability. We also questioned both Tyan and AMD for their apparent lack of action toward remedying the situation. Before reading any further, look here, as well as here, and here.
There have been some developments since then, and today we’d like to update the situation. After scouring through forum threads, as well as speaking to Tyan and various other sources, we’ve come up with a few interesting discoveries.
First off, at this point, the problem with the board using the wrong FSB speed seems to boil down to a combination of user error and a sloppy manual. Apparently, the jumpers on the board, and the setting descriptions in the manual, are arranged in a somewhat confusing fashion, which, in turn, has led to several users configuring the jumpers incorrectly, and the board POSTing at the wrong FSB speed. Since then, Tyan has updated the User’s Manual on its website, and presented these settings more clearly. So, just a User’s Manual misprint. It would perhaps have been nice if this had been caught prior to printing of the manuals, but this certainly isn’t a big deal at all.
Further, it seems that, at this point, a great deal of the problems are occurring in systems using the Tyan-recommended NMB power supply. Tyan, of course, is blaming NMB for the problems (although we still can’t fathom why Tyan’s QA check didn’t find this sooner). NMB hasn’t said much. In a large percentage of the cases, swapping the NMB power supply for a different unit has greatly reduced or eliminated stability problems. So, were we wrong about the board, and this is all due to a Manual misprint and a bad power supply? We don’t think so. We’ve seen reports of users whose stability problems have disappeared after having the board, not the power supply, replaced. In other words, it seems that only certain boards don’t like certain power supplies.
As a measure of confirmation, we spoke with (on the condition of anonymity) an online retailer of Tyan products in the US, who told us that, of about 500 Thunder K7’s sold to date (a combination of S2462NG and S2462UNG), there have been a number of minor issues, and four serious problems. Furthermore, he has informed us that after exchanging several boards, customers are no longer experiencing problems. That’s rather odd, if this is all the fault of a power supply, wouldn’t you say?
With regards to certain SCSI, RAID, and other adapters not working, we’ve yet to discover any new information.
It seems to us that, quite logically, the dual Athlon platform is extremely picky about its power requirements, and that perhaps neither the Thunder K7 nor the NMB power supply are up to par in some cases.
Last week, we started to question the Tyan Thunder K7’s apparent quirkiness and lack of stability. We also questioned both Tyan and AMD for their apparent lack of action toward remedying the situation. Before reading any further, look here, as well as here, and here.
There have been some developments since then, and today we’d like to update the situation. After scouring through forum threads, as well as speaking to Tyan and various other sources, we’ve come up with a few interesting discoveries.
First off, at this point, the problem with the board using the wrong FSB speed seems to boil down to a combination of user error and a sloppy manual. Apparently, the jumpers on the board, and the setting descriptions in the manual, are arranged in a somewhat confusing fashion, which, in turn, has led to several users configuring the jumpers incorrectly, and the board POSTing at the wrong FSB speed. Since then, Tyan has updated the User’s Manual on its website, and presented these settings more clearly. So, just a User’s Manual misprint. It would perhaps have been nice if this had been caught prior to printing of the manuals, but this certainly isn’t a big deal at all.
Further, it seems that, at this point, a great deal of the problems are occurring in systems using the Tyan-recommended NMB power supply. Tyan, of course, is blaming NMB for the problems (although we still can’t fathom why Tyan’s QA check didn’t find this sooner). NMB hasn’t said much. In a large percentage of the cases, swapping the NMB power supply for a different unit has greatly reduced or eliminated stability problems. So, were we wrong about the board, and this is all due to a Manual misprint and a bad power supply? We don’t think so. We’ve seen reports of users whose stability problems have disappeared after having the board, not the power supply, replaced. In other words, it seems that only certain boards don’t like certain power supplies.
As a measure of confirmation, we spoke with (on the condition of anonymity) an online retailer of Tyan products in the US, who told us that, of about 500 Thunder K7’s sold to date (a combination of S2462NG and S2462UNG), there have been a number of minor issues, and four serious problems. Furthermore, he has informed us that after exchanging several boards, customers are no longer experiencing problems. That’s rather odd, if this is all the fault of a power supply, wouldn’t you say?
With regards to certain SCSI, RAID, and other adapters not working, we’ve yet to discover any new information.
It seems to us that, quite logically, the dual Athlon platform is extremely picky about its power requirements, and that perhaps neither the Thunder K7 nor the NMB power supply are up to par in some cases. It’s very likely that some NMB power supplies are having trouble meeting spike current demands or maintaining a clean, ripple-free signal. This would only be exacerbated by a few ‘intolerant’ boards out there. The fact that some boards seem intolerant, while others are fine, leads us to question the board design/production itself.
In any event, the issue may not be as serious as we had initially thought, but nevertheless, we think a 5% problem rate on a $500 server board is a little unacceptable. You get what you pay for -- quirks are to be expected of a cheap $60 board, but not of an expensive server board. Whether these issues are due to a sloppy User’s Manual, inadequate power supply, quirky boards, or likely a combination of all three, something still needs to be done. We note that, while Tyan continues to blame NMB for the problems, neither it, nor AMD, has yet removed the NMB power supply from its recommended list. Given the sparseness of the Delta unit, doing so would damage sales of the platform. Are we off the wall in thinking that, when you're talking about a $500 server board, quality and stability should come before sales?
For now, if you’re planning on buying one of these boards, we recommend spending a little extra time searching for the Delta power supply, since, to date, no one with a Delta unit seems to be encountering any major difficulty. Might save you some headaches.
As usual, keep the reports of experiences, both good and bad, coming so that we can gather as much information as possible. Thanks to all those who have contributed thus far.
Dan Mepham |
