TYAN Tiger MPX S2466-4M, problem with BIOS rev. 4.03?

[CougTek]

Hi,

I will configure latter this evening a dual-Athlon MP setup on the motherboard I named in the thread title. I think at least one of you has that motherboard and in order to avoid possible issues when I'll configure it (this IS serious stuff, no shit allowed), I'm seeking user experiences about the 4.03 BIOS that I'm about to install. I didn't build the system originally, but I have to do an in-depth check-up to insure that everything is ok. I'll even remove the retail heatsink and re-apply the compound (I bet there're thermal pads there instead). This thing will run Winblows 2K Server, but the OS installation shouldn't be a problem. It will, as you can guess, be a server (running MS SQL) and stability/reliability are THE words.

TYAN makes great stuff for SMP setups, so I'm not too worried. I mostly ask for the shake of it.

 

[Clocker]

Coug...I'm currently using teh 4.01beta BIOS. I'll update to the 4.03 right now and let you know how it goes. Thanks for the heads-up.

C

 

[Clocker]

Seems to be working great. No problems with my config. (see below).

The only annoyance I have (I had the same problem from the very beginning) is that my Hercules Fortissimo II sound card GAME PORT can't seem to find resources for itself. No big deal really, I never use the thing so I just disable it in Device Manager anyway...

 

[CougTek]

Thanks Kevin, I'm glad to know that the computer I was supposed to have in front of me now would work great. But the POS I currently have infront of me isn't what I expected to be.

[brief explanation]

You see, I spec'ed a dual-socket-A system for a customer needing a MySQL database about a month ago. The guy had a friend who happens to sell computers too. What the friendly seller proposed was, simply put, an overpriced piece of junk. He spec'ed a typical user box with some no name components into it and expected it to work fine for a database server. Our programmer, the guy who'll install and manage the database software on this server, requested that the config I proposed would be the one he'll work on or else he'd simply refuse the contract. Because our customer needed the services of our programmer, he told his friend to match my config. Even though I beated the other one's price by 500$, the customer chose to go with what his friend offered him because he felt that he would probably have better service from his friend than from myself. Quite understandable, no problem. However, my programmer requested that I install and configure everything on the server because it was obvious that the other wouldn't be able to do it right itself and we aren't too interested to fix the bugs of others.

[/brief explanation]

Fast forward today (it took the friend-seller one full month to deliver the box, unbelievable but true). I received the box to inspect it and configure Win2K Server on it. But what I saw instead of the config. I spec'ed last month is unreal. The box is generic and the metal is so thin that I think it must be 0.6mm thick instead of the 1.0mm sturdy Antec SX1080 I recommended. The power supply is a 300W generic unit too (for a dual-socket-A!!!) Inside the box, no fan in front, no fan in the rear. Even if there would be fans, the holes in the enclosure for the airflow are so scarse that it wouldn't help much. The motherboard isn't the Tiger MPX S2466-4M, but the previous S2466 with the broken USB port (both cost the same, why go with the bug-ridden one?!?). Graphic card : TNT2 no name. At least it could have put an ATI Rage something to get a decent 2D image. And so on and so on. Shit! I wouldn't even sell that to a home user (dual Athlon MP on a generic 300W PSU = BSOD for sure), much less as a server! And all that crap for 500$ more than what I proposed!?! I'm enraged! Outraged! I simply cannot believe it! Did the fucking idiot even powered it up to see if it would boot? And now I must configure and install a stable and reliable Win2K Server on that POS?!? I think I'll call the customer, tell that he's been screwed hard by his friend and to sue the robber who sold him that box.

I just can't believe it. 500$ dearer for something that cannot run stable. Amazing. And now I have to figure out to make a server out of this? Who the hell does he think I am, a wizard or something?

 

[Clocker]

Wow. What BS. If it doesn't cost too much, I'd refuse the business and say bye bye to the idiot. You will probably spend more time pulling your hair out trying to make the POS work that what it is worth....

What a statement that would be....of course it would cost you a bit of dough. But, I would not want to do business with that big of an idiot..

C

 

[CougTek]

Oh shit, it's getting worst! The front panel of the cheapo enclosure has one of its retention pin broken and it is loose. The bandit tried to fix it ... with glue! There's a yellow spot of glue on the side of the panel, I can't believe it! It looks like wood glue...to keep together a plastic piece to the metal structure.

And it's not over : the case is bent on the rear at the top side of the PSU. There was one of the power connector stuck between the top of the PSU and the enclosure just above it. It looks like the guy tried to scew it like that anyway and he bent the case by insisting too much with the screw driver. Unreal, I tell ya.

Want to know how much he charged for that? IIRC and if it was indeed 500$ higher than what I offered : approximately 3800$CDN (with Win2K Server)! I just can't believe it! A school kid in his basement wouldn't have done worst.

 

[Cliptin]

I would also refuse to work on it unless you make perfectly clear that you are not responsible for the box's stability. Too much reputation at stake.

If the customer just received the machine, he is well within his right to return it and then buy yours.

 

[Clocker]

I feel dumb for asking but...what is 'IIRC'?

C

 

[Cliptin]

I feel dumb for asking but...what is 'IIRC'?

C

If I Remember correctly.

 

[CougTek]

I feel dumb for asking but...what is 'IIRC'?

C

More than 3000 posts on the old SR forum, nearing 700 here, doubtless several others elsewhere and you still don't know what IIRC means :eekers: Unreal. Do you have a yellow spot of glue somewhere? Just joking Clocker, I'm tired, sorry.

IIRC : If I recall correctly

others...

FWIW : For what it's worth
FYI : For your information
OTOH : On the other hand
IMO : In my opinion
IMHO : In my humble opinion (often sarcastical or ironical)

There are several others, but I'm tired, give me a break.

 

[The JoJo]

Cougtek, if I was you I would be yelling "abort mission, abort mission" !

Even if you get it up and running, what if the system is unstable and boots itself once a week. The customer will remember that YOU were the one to put it in order...

Carefull now! You don't want to loose the customer either (probably).

 

[CougTek]

I'll show the customer this document from AMD : http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25823.pdf

His PSU is very borderline for the +3.3V and +5V wres, while it is underpowered for the +12V wire (PSU rated at 15A while AMD requires 16.5A).

The PSU is a Young Year PSIV-400-2. The 400 in the model # is misleading because it is in fact a 300W unit. A sticker on it tells that it's certified for P4. That's far from being enough to insure me that it will drive a dual-Athlon MP setup. A search on Google regarding that name (Young Year PSIV-400-2) yields no result; very re-assuring. I haven't powered it up yet because I'm not sure if the PSU will survive the current pike at startup. There's no bottom interior fan on the PSU so air flow in this case will be horrible. AMD recommends that case temps do not exceeds 45C. Anyone wants to take bets as to how much time it will take before that limit is reached and broken for this box?

Else the motherboard/CPU, there are 4 other drives inside that box : three hard drives and one 52X CD-ROM. Would it be enough to kill the PSU according to you. Still unsure I'll take the risk.

Another look at AMD's recommend motherboard list for Athlon MP doesn't include the TYAN Tiger MPX S2466 that is inside the "server" I have here. Only the S2466-4M is recommended, which is quite understandable since the non-4M is crippled of its USB ports.

 

[CougTek]

Interestingly, the south bridge on that motherboard is the fixed "B2" revision. However, on the board's layout, it's written twice and quite clearly that it's the 2466, not 2466-4M.

Clocker, do you have the -4M and does the marking on the motherboard indicates 2466 or 2466-4M? I might have the good version afterall, although I would find it quite poor from TYAN not to have modified the marking on the new revision of the board.

 

[time]

Coug, stay calm. This definitely calls for major diplomacy. But the situation should be salvageable, nonetheless. You could really do with advice from Tony here. I'm certain he would be able to guide you through to a far more satisfactory state of affairs.

Remember, the customer insisted on paying an extra $500 for the privilege of dealing with his friend. He's probably a very loyal person, and therefore a desirable customer, if you can turn this to your advantage.

Cool down. The PC doesn't need to be that cool for proof of concept. And the thin case may be offensive for us, but it isn't likely to stop it working altogether. There's no reason you can't proceed.

The power supply is obviously dodgy, and I think you need to convey this to the customer in a clinical way that doesn't pit you against his friend. Over to you, Tony!

 

[Clocker]

Interestingly, the south bridge on that motherboard is the fixed "B2" revision. However, on the board's layout, it's written twice and quite clearly that it's the 2466, not S2466-4M.

Clocker, do you have the -4M and does the marking on the motherboard indicates 2466 or 2466-4M? I might have the good version afterall, although I would find it quite poor from TYAN not to have modified the marking on the new revision of the board.

I DO have the -4M.

IIRC, the only way to tell the physical diference between the non-4M and the -4M was to see if the chipset had the B2 stepping on it. I too was unsure if I had the 'good' S2466 until I found that out. So, you should be OK Coug.

C

 

[Buck]

Coug,

Your customer will try to uphold the reputation of his friend, so attacking the quality of the system will certainly work against you. What I would do is describe the situation with the PS and offer to replace it yourself since you understand how busy your customer's friend is. Explain how this could be done along with the Windows installation and configuration in a matter of a few days. Either explain the price of the PS upfront (which would not be my first option), or include it in your price to install and configure Windows (obviously, I don't know what you've already estimated). As hard as it may be for you, searching hard for the odd commendation about the system won't hurt either.

 

[CougTek]

Thanks guys. I will try to make him change the PSU at least, but the whole enclosure should be. It's not just the fact that it's very thin, it's because there's no airflow within it. We recommended to go for a dually so that the customer should be good for 5 years without upgrading, With such a poor cooling, I doubt everything inside the case will last that long.

Anyway, I have slept since yesterday's horror show and I ad the time to cool down. I'll see what I can do.

Thanks Clocker, I must have the -4M afterall.

Oh and one last thing. Although we asked for Registered RAM so that it would be possible to upgrade later on with additional DIMM, I discovered that our friendly seller put ECC but unbuffered RAM on the board. Wonderfull isn't it? Now if more than the 1GB of RAM is needed later, we'll have to change all sticks. A true pro that guy, really. I bet he doesn't even know it about Reg. vs unbuffered.

 

[Mercutio]

Traditionally video cards don't matter much in servers, but just to contradict time, I'd like to point out that a thicker case is going to be a slightly better EMI shield than a thinner one. Since servers tend to be put in places with lots of EMI (with all the HVAC and power requirements typically asociated with "server" environments), there *is* a compelling reason to choose a heavy-duty case.

Personally, at this point I would diplomatically say that the hardware Coug's been given is not acceptable for the task at hand. That machine will be nothing but trouble, and CougTek needs to, at a minimum, get an assurance that his competance won't be questioned until the objectionable elements of that machine are either replaced or returned.

 

[time]

I'd like to point out that a thicker case is going to be a slightly better EMI shield than a thinner one.

Eeeek! Don't misunderstand me, I loathe and detest cheap cases. I was just trying to keep CougTek on a pragmatic path.

All things considered, perhaps it would be best to try to shoot for a whole new case and PS?

Or maybe Coug should leave it as is with the qualification that it doesn't meet either his spec or AMD's design guide, and he therefore takes no responsibility?

 

[Pradeep]

Oh and one last thing. Although we asked for Registered RAM so that it would be possible to upgrade later on with additional DIMM, I discovered that our friendly seller put ECC but unbuffered RAM on the board. Wonderfull isn't it? Now if more than the 1GB of RAM is needed later, we'll have to change all sticks. A true pro that guy, really. I bet he doesn't even know it about Reg. vs unbuffered.

Shocking. IMO the TigerMP shouldn't be run with anything but Reg. memory. Now the MPX may permit the use of two unbuffered modules, but for maximum stability Reg ECC is definitely a requirement. I think it is time to tell your customer that you cannot take responsibility for the stability of such a POS machine. Now if it were my friend who had built me that, and I found he'd charged me $500 more for cheaper, crappier parts, he probably wouldn't be my friend anymore. Friends don't build friends cheap shitty computers.

 

[Cliptin]

Or maybe Coug should leave it as is with the qualification that it doesn't meet either his spec or AMD's design guide, and he therefore takes no responsibility?

This is where the diplomacy breaks down in my mind. You don't want to bash his friend because that the guy defensive for his friend. But the next question from the customer is likely to be "Well, what is wrong with it."

You can go over your original design decisions with the customer and indicate why you did not make the decision the friend made. Regardless, you are going to be put into an odd position.

Here's what you do. Reaffirm to the customer your commitment to service after the sale and emphasize that you can not be confident in the hardware as it is based on it not meeting the design guidelines of AMD and yourself. Couch it in such a way that the ball is back in the customers court as to the responsibility for stability.

He will either decide that the box is good enough and hence you are obsolved of responsibility or decide to have you fix the problem whether that means upgrades or complete replacement.

 

[CougTek]

Couch it in such a way that the ball is back in the customers court as to the responsibility for stability.

That's what we are currently trying to do. I re-checked the list I offered him instead of the box he got and I dropped a tear. I will ask him to change the whole enclosure, not ust the PSU. Anyway, the moron bent the case when he screwed in the PSU so the case is already damaged.

About the RAM, yep, unreal. Without Reg. sticks, our customer won't be able to upgrade without replacing the sticks he already has, instead of just adding sticks if it would have been reg. memory.

 

[CougTek]

The customer decided to take the risk and to continue with that box. The seller will replace components if they fail. Fine with me.

So I powered up the box as is this afternoon. Just installing the first components of Win2K Server and both CPU were at 79C after the first restart. As I was re-assured that I wouldn't be responsible if a failure would occur, I completed the installation and updating process (install IE6, SP3, blablabla). In order to be sure that the system will be stable and reliable under all situation, I'll have to make a short term burning test (two instance of FAH with 100% CPU usage for each CPU) that I'll continue untill a)system melts or b)they come to search their box. Expect my FAH stats to increase slightly over the week-end (It might not last that long though...)

Currently, putting my hand in front of the power supply exhaust fan is a heat endurance challenge. Internal temp must be between 55C and 60C. Would I want to be entirely sure of the reliability of the dually, I would close the door and window of that room and put a 17" monitor box on the enclosure and leave it like that for a while, ...but I'm not that evil, am I :twistd: Afterall, it's only about 4000$CDN, so it shouldn't be such a big deal?

No, I'm not that evil.

 

[Pradeep]

Coug, what app are you using to check the CPU temps?

My duallie also runs a fair bit of hot air out of the PSU.

 

[Clocker]

My dually gets hot around the PSU too. It blows a lot of hot air and the PSU enclosure is very warm to the touch. Internal case temp is 40C. My CPUs are 56C and 40C (the one by the AGP is hotter). According teo MBM, their voltages are slightly different too (by about 0.03v)...heck I guess this pic is easier (the X15-36LP is listed too):

I have a 120MM intake in front, and an Antec temp adjusted 92mm exhaust. Oh yeah, I also have a Panaflo 80mm L1A in front of my drive cage that is basically silent (it is only about 20cfm). The PSU is a Traditional Technology Group 420W:

http://www.ttgitech.com/PS-420ATX.htm (28W/42W/16W) on the (3.3v/12V/5V) rails. THis PSU is really quiet. The fans are slow and temperature controlled. Actually, all the fans in my rig are temp. controlled except for the 120MM intake which is manually adjustable. All the fans (including the two in the PSU) seem to bias to the slow side which is nice .

Here is a screen shot of MBM after I had the side open for a few minutes to check the wattage on my PSU and then closed it back up. Notice the temp and fan speed difference.

C

 

[CougTek]

Coug, what app are you using to check the CPU temps?

The BIOS, when I reboot the computer. I know there are better ways, but I don't know much about these programs. Could you suggest one? Motherboard Monitor is a name I ear often.

Clocker, does it take something like 10 or 15 seconds for your motherboard to resume when you reboot. I know duallies take longer than single CPU systems, but I find it quite long everytime I reboot.

Oh and Pradeep too, I forgot you own a dually too.

 

[Clocker]

Yes. It is somewhat slow. For some reason it seems there is about a 10 second delay for the video signal to be initialized to energize the monitor (even on a soft boot, the monitor goes off). In fact, it is so slow that I have never seen the monitor show the Radeon spash screen at the very beginning of boot. I disable the fast boot option in the BIOS (so that the memory counts) just so that I can see how everthing is detected during boot (i.e. the CPUs etc.). You can always hit any key to cancel the memory count after that if you don't feel like waiting forever.

C

 

[CougTek]

Oh and the generic TNT2's (M64 too, if it wasn't bad enough already) image quality is aweful. There's no other word. At 1280x1024, everything is so blurry that I sometimes have to guess a few letters in small font caracters. Even though graphics aren't a priority on a server box, there's low and LOOWWW. The ATI Rage IIC 4Mo PCI I use in my own once-was server box is infinetly crisper than the piece of generic junk found in the 4000$ server (I like to remind myself how much the crook charged for this) that sit aside of me.

Never use cheap generic graphic cards, even when graphics don't count much.

 

[CougTek]

Update : I installed TYAN's own monitoring software. I started FAH on both CPUs in within 30s, their temps past from 61C (when at idle) to 64C and climbing. All temps across the board (VR1, VR2, AGP, etc) are at 61C and above when at idle and now (2 minutes after I started FAH) at 62C and above.

The guy is fool to assure this stuff. It will fail sooner than later. He would lost less money just by taking back the cheapo case + PSU (this junk costs 45$CDN combined) and either replace it with a decent enclosure or let me do it for him. Oh well, I guess that's how we learn...

 

[Pradeep]

Coug, what app are you using to check the CPU temps?

The BIOS, when I reboot the computer. I know there are better ways, but I don't know much about these programs. Could you suggest one? Motherboard Monitor is a name I ear often.

Clocker, does it take something like 10 or 15 seconds for your motherboard to resume when you reboot. I know duallies take longer than single CPU systems, but I find it quite long everytime I reboot.

Oh and Pradeep too, I forgot you own a dually too.

There is a bug in the BIOS (at least with TigerMP) where the temp readings in the BIOS are very high. If you hit the left or right arrow key, the temps will update and tumble from 80's down to a more reasonable 30-40, which is the actual temp. MBM therefore doesn't work with the TigerMP, I use an app called K7TT.exe, it gives the proper temps.

Oh and with the long delay on boot up, that is caused by the ECC memory check, the delay gets longer the more memory you have. If you turn of ECC checking it should speed up your boot (defeats the purpose of ECC tho).

With 100% load I have temps not exceeding 60 C (dual 1.47's).

 

[CougTek]

The temperature stabilized at 66-67C for both CPUs. Problem is all the other places inside the enclosure are between 64C and 67C too. But at 67C, the processors (unfortunately) won't melt. I won't kill it during the week-end, but the box will experience stability problems soon and I don't expect it to live more than 2 years (at most) at such a high interior temperature.

Oh and TYAN's own monitoring software doesn't share the bad temp reading bug of the BIOS, so 67C really is the accurate temperature.

The voltage on the +5V is currently a bit low at 4.6V while the +12V is at 11.2V. Most of the others seems ok though. The monitoring tool (v2.07) also reports great variations between the max and min voltage of many wires. This isn't a good sign for the future. Well, a good sign for the server as it currently is.

I find it sad to ruin such a good piece of hardware (motherboard + CPUs I mean) inside a so poor enclosure.

 

[Clocker]

Pradeep-

Where can I get that K7TT.exe...google does not find it.

C

 

[Pradeep]

It was originally released on the 2cpu forums, this is just a copy I have:

http://vdpc.org/K7TT.exe (20KB )

I don't know if it works on the TigerMPX, it was designed for the TigerMP. Let me know how it goes

 

[Clocker]

K7TT seems to produce more reasonalbe numbers.
CPU two is really close. CPU one seems off....I'll put an offset into MBM...

 

[Clocker]

Interestingly enough...MBM has never shown this big of a difference in the CPUs before. Usually, the difference was about the same as K7TT is giving. MBM was always about 10-12 degrees off between the CPUs, not 17 like it is now showing....

Weird

 

[Clocker]

DISREGARD MY POST ABOVE

Somehow MBM was reading the temp of the SCSI HDD for the temp of my CPU for CPU1. That is why the reading was so low.

MBM and K7TT give me the same numbers for CPU temp on BOTH CPUS.

C

 

[Pradeep]

Interesting. Mine always has CPU 0 running about 6 degrees hotter than CPU 1.

 

[Clocker]

Mines about 11 right now.

C

 

[time]

The temperature stabilized at 66-67C for both CPUs. ... I won't kill it during the week-end, but the box will experience stability problems soon and I don't expect it to live more than 2 years (at most) at such a high interior temperature.

Perhaps, but for heaven's sake don't tell anyone that. Claiming that it will die will only make people doubt your expertise. And remember the temperatures are not that unreasonable for a duallie, the only clearly suspect item is the power supply.

I truly, truly feel your pain in this, but from one obsessive personality to another: "Let it go!" There are lots of computers out there with cruddy components, yet most of them work anyway. I don't like it either, but it's a fact. We would both benefit from learning to compromise more often.

 

[CougTek]

My expertise won't be affected at all if the crappy box melts Time, for the simple reason that I recommended something else than this configuration. Since it's different from what I told them to get (different in a bad way), it's no wonder that the thing overheats and if it ultimately fails during my tests or later, I will deny any responsability because ït's not the configuration I told them to buy.

After one full day of FAH, both CPUs are at 70C and no point inside the enclosure is cooler than 68C. Tomorrow, it's supposed to reach 31C during the afternoon here, but that's not as bad as the 35C to 40C temperatures in which the server will have to operate next summer since it will sit in a rather hot environment.

So yes, in a way, I'd like it to fail now, but only because I'm sure it won't last long on the ground like it would have been supposed to. But I'm not actively trying to kill it, my tests remain fairly regular. Sure I wouldn't test another box as much, but only because I know that a well-designed cooling setup would be effective enough not to worry about it. In the current case, I'm worried so I test a bit more. In fact, I doubt I will kill it this week-end. But it would be better if I would because we would save more trouble in the future and that it's obvious it will die soon anyway.

Thanks for the advice, but I won't follow it this time.