Skynet!!

He's the reason they switched to lethal injection over the electric chair for executions. Not enough juice to get the job done. ;)
 
Not muchto show until I get the heatsinks on, they are at the parents place which I'm headed to right now. Once I get back I just have to do that Christmas stuff and I'll be ready to go (the lady may not let me play much tonight, but she's got work at noon tomorrow!)
Posted via [H] Mobile Device
 
MOAR PICS...... MOAR D....

quit slacking and enjoy your xmas
 
very quick update, getting ready for dinner now, but I measured them and they apear to be the right size

Yay!

and my ECWCS gear came in, no more freezing my ass off
Will post more pics tomorrow
 
Holy shit, these fans are crazy loud!!! Must get a controller. I've got 1 CPU in now, I'll put in the other 3 and go from there
Posted via [H] Mobile Device
 
Vaulter98c said:
Holy shit, these fans are crazy loud!!! Must get a controller. I've got 1 CPU in now, I'll put in the other 3 and go from there
Click to expand...
Vaulter, which HSFs did you finally get for this setup (sorry if I missed it earlier in the thread)? They may be crazy loud but they'll get the job done. In my experience, the loud server HSFs tend to also be the coolest, which is something that isn't always the case with enthusiast components. There really isn't very many choices out there and the quieter server stuff is usually not the coolest, unfortunately. If you can manage to OC (a difficult if) you'll need the additional cooling these fans will likely provide. Trust me, stability is paramount with -bigadv.
 
Ok, got all 4 CPUs mounted and 2 sticks per CPU. Takes forever to boot, but starts the post, all 4 CPUs are detected fine, and then it freezes during memtest, gets to 4 gigs even then stops, board number shows 60, which reads test extended memory. I've got to run real quick but I'll be back soon. When I get back, any more ideas besides swap some memory out?
Posted via [H] Mobile Device
 
Vaulter98c said:
Ok, got all 4 CPUs mounted and 2 sticks per CPU. Takes forever to boot, but starts the post, all 4 CPUs are detected fine, and then it freezes during memtest, gets to 4 gigs even then stops, board number shows 60, which reads test extended memory. I've got to run real quick but I'll be back soon. When I get back, any more ideas besides swap some memory out?
Click to expand...
The only thing I can suggest at this time without knowing the various BIOS settings is to test each module individually, and if that doesn't reveal anything, test each CPU individually with different memory configurations. My dual-socket Opteron is very sensitive to BIOS settings. This is NUMA architecture and it might be more complicated/time-consuming to troubleshoot your setup. Also, you might want to test the PSU on another system to see if it is outputting sufficient power, but I do not believe that is the culprit.
 
Ok, well, bios options are out because it won't let me far enoigh to get into the settings. Illtry playing arou d more but the manual offers no help
I'll keep you posted and I've got more pics, but they are phone pics
Posted via [H] Mobile Device
 
Vaulter98c said:
Ok, well, bios options are out because it won't let me far enoigh to get into the settings. Illtry playing arou d more but the manual offers no help
I'll keep you posted and I've got more pics, but they are phone pics
Click to expand...
Disappointed to see this project at a standstill. It's unfortunate the documentation isn't detailed enough to offer a possible solution. One thing I can relay from my Opteron experience is the placement of the memory modules can be very important. Depending on the type and size of the modules, sometimes only a specific order will work. Mind you, in such occasions the board would likely not POST at all but in your case it's freezing midway through the boot process, so I'm at a loss... :confused:
 
Ok, tried a bunch of things, but nothng worked. Finally I said screw it, populated every slot, now it works. Won't see beyond 8 gigs tho, so ivegot to find the missing 24 gigs

edit - all 32 gigs reporting for duty!
Onward to win home server. Damn these are loud LOL, I'll have to see if I can throtle them, but I'm all in the 20s LOL
Posted via [H] Mobile Device
 
Windows home server? Judas priest my man, a machine like this deserves a native Linux install! You'll get much better folding results.
 
Fantastic!! At least one thing finally worked to get the system up and running. Can't wait to see pics. Congrats! :cool:
 
O it will get it, but while I'm playing around I'd like to have something easy to fall back on
Anyways, I'm having some stability issues, I'd guess it's memory related, keep getting reboots trying to enter windows, even thru safe mode
Posted via [H] Mobile Device
 
Vaulter98c said:
O it will get it, but while I'm playing around I'd like to have something easy to fall back on
Click to expand...
My native Linux install has a dual boot with XP on another drive where I set my overclocks. If you install Windows first and then install Linux on another drive, the Linux installation will query if you want an OS boot manager. I suggest to select yes because it will be very easy to switch between OS installations this way.

Anyways, I'm having some stability issues, I'd guess it's memory related, keep getting reboots trying to enter windows, even thru safe mode
Click to expand...
Go into the BIOS and see what memory options are in there.
 
Ok, still gettng the reboot issue, I've told bios to do more extensive men tests, nothing

Any ideas?
Posted via [H] Mobile Device
 
Ok, tried loading Linux, as I boot into the installer I get a kernal panic - not syncing - fatal exception
Gonna try WHS again
Posted via [H] Mobile Device
 
Disable ECC in the BIOS as well as anything else ECC-related. If you can get a stable enough OS installation, run a memory tool like memtest or other utility to verify that all modules are in optimal working condition.
 
Still rebooting when I try to get Into windows (right after I boot from disk and it says windows is loading files, once that bar loads it reboots)
Reboot when I try to boot into an old whs Install it reboots once the win Dow should pop up
Linux install gives me a fatal error, no reboot tho
And now I'm noticing that not all the ram shows up everytime, soMetimes it does, sometimes it doesn't, 8 gig when not, 32 when working
It's not temps, I've watched them for a while, fanless HS's on chips get about 50, CPU's stay around 20-30
Shouldn't be power, not seeing fluchations and it's a brand new unit, well under half it's rating

I'm clueless and I think I'm gonna call it a night, I've got to work again in a few hours, got to get dinner going
Posted via [H] Mobile Device
 
Vaulter98c said:
I'm clueless and I think I'm gonna call it a night, I've got to work again in a few hours, got to get dinner going
Click to expand...
OK, no problem but we got to find what the problem is eventually, so I'm going to ask you to do something you might find tedious. I'm thinking like you it's a strong probability that the issue lies somewhere in the memory subsystem but what precisely I'm not certain without having the system in front of me. Give us if you can the following information:

Board model and revision
BIOS version
Memory module manufacturer and model
CPU steppings

Nothing may come of this but it's worth a shot. Someone may think or know of something. These are not systems most enthusiasts encounter every day or every year for that matter. If worse comes to worst, post your issue in the Multiprocessor forum or even better at 2CPU.com forum. The latter forum has the biggest congregation of multi-socket enthusiasts in any computer tech forum and the archives are replete with information.
 
The memory may not be compatible with that board is what I can figure.

My ASUS board did that with the Mushkin RAM I used to have. Deteced anywhere from 2-4 out of the 6 gigs I had. Once I got new RAM it worked like a charm. Something to consider anyway.
 
Memory should be fine, I got all the way thru and into windows with 1 quad and 2 sticks

I'm not with it anymorebut the manual says Tyan S4985 (Thunder n4250QE) rev 1.0
I don't remember much more, I think the stepping is B1, not sure tho, but I know all 4 are the same
Memory, don't know without being there, maybe nitro knows
Bios, don't remember
Posted via [H] Mobile Device
 
Well reason I say that is because when I had 2 sticks in it worked just fine. Same thing with 3 sticks. When I put any more than that it was behaving erratic.

But then again, completely different board, DDR3, and Nehalem architecture.
 
About ECC, acording to the manual, once I redid the bios ECC was disabled
Posted via [H] Mobile Device
 
Damn, just to add insult to injury I was almost late for work when I got a speeding ticket 100 yards from the parking lot. First one...

Right when I get a handle on my medical bills I get another 400$ from radiology and now this

Fuck
Posted via [H] Mobile Device
 
Sorry to read about your misfortune, Vaulter. :(

I have another question. Which PSU are you running? Is it the Corsair in the OP? Reason I ask is due to the board's dual EPS specification. Is the PSU powering the board dual EPS compliant?
 
Valuter

What's the problem? I think I missed something further up.

Just to get up and running I would start one cpu at a time with one stick per cpu installed.

Also what memory are you using? There were two sets, 4x1 set and 16x2 set. I would use the 2gb modules and don't mix.

Also disabled everything not necesary in the bios
 
I got 2 8 pins, asked around and people said it would be fine. Every power plug on the board is populated

It was fine with 1 quad and two sticks, only setup I havnt tried was 2 quads and 4 sticks
Posted via [H] Mobile Device
 
yeah sorry about your trouble vaulter, you might need to memtest all the sticks of ram separately, it's a pain but it looks like this is what's giving you the reboots. Good luck mate
 
Vaulter98c said:
I got 2 8 pins, asked around and people said it would be fine. Every power plug on the board is populated
Click to expand...
OK, there is no consensus about whether power plug adapters are fine or not. There are people who favor them and those who do not. I generally fall into the latter camp. For enthusiast-level boards and many workstation boards power adapters are OK. I have only used molex>4-pin P4 adapters and even they don't always work with every board I tried. I would be very wary of using an 8-pin EPS adapter. However, if there is no other choice, get the adapters that run off the PCI-E cables on the PSU, not the molex cable. Reason I advise this is due to better current available. Look here: http://www.overclock.net/power-supplies/551470-dual-eps-850-watt-power-supplies.html

Either way, I strongly advise against any kind of power adapters for a quad-socket board. When I say strongly, I mean don't use them if you're populating all processor sockets and installing many memory modules, etc. What may be fine on enthusiast or dual-socket boards might not fly on a quad-socket platform. Even if it works for a while you may see problems down the road. You should purchase a dual EPS compliant PSU and I can recommend a few PSUs to look at. This may not be your problem but it could become an issue down the road. Thought I'd give you a heads up.

It was fine with 1 quad and two sticks, only setup I havnt tried was 2 quads and 4 sticks
Click to expand...
TBH Vaulter, I would start folding on this setup even without all the components installed just to determine if the basic configuration even works for prolonged periods. You could start with two processors and run regular SMP if you harbor doubts about long-term stability with -bigadv. Once you have a better handle on the situation, install the remainder of the processors and memory modules.
 
APOLLO said:
OK, there is no consensus about whether power plug adapters are fine or not.
Click to expand...
Then allow me to provide a consensus: they're fine. Based on the current capacity of the pins and wires, there is no way that the power draw will ever exceed specifications or cause failure.
 
Zero82z said:
Then allow me to provide a consensus: they're fine. Based on the current capacity of the pins and wires, there is no way that the power draw will ever exceed specifications or cause failure.
Click to expand...
Here, I will disagree with you Zero. Having run multi-socket boards since the mid-late 90s, I would be very reluctant to veer away from the manufacturer's recommendations based on the specifications of the componentry involved here. I'm actually quite surprised to see you being this certain about success with adapters and this platform, especially after having unpredictable results in my experience with lesser hardware. Even if it works, it will be hard to troubleshoot an issue of instability if all one has at one's disposal is an enthusiast PSU and adapters to test with. Best to adhere to manufacturer recommendations and eliminate as many potential uncertainties as possible with setups as complex as this.
 
APOLLO said:
Here, I will disagree with you Zero. Having run multi-socket boards since the mid-late 90s, I would be very reluctant to veer away from the manufacturer's recommendations based on the specifications of the componentry involved here. I'm actually quite surprised to see you being this certain about success with adapters and this platform, especially after having unpredictable results in my experience with lesser hardware. Even if it works, it will be hard to troubleshoot an issue of instability if all one has at one's disposal is an enthusiast PSU and adapters to test with. Best to adhere to manufacturer recommendations and eliminate as many potential uncertainties as possible with setups as complex as this.
Click to expand...
If I remember correctly, the current limit for each molex pin is 9A. If that is indeed the case, then there is no chance of any failure so long as all the CPUs remain within their TDP (and it is very unlikely that they will exceed it by enough to cause an actual failure).
 
Zero82z said:
If I remember correctly, the current limit for each molex pin is 9A. If that is indeed the case, then there is no chance of any failure so long as all the CPUs remain within their TDP (and it is very unlikely that they will exceed it by enough to cause an actual failure).
Click to expand...
Yes, no chance of failure. I wouldn't be concerned with failure but with adequate current being supplied to the board to properly feed all the installed componentry. It might not be noticed at all but for the occasional lockups or crashed WUs that may occur once in a while. Something as innocuous as a random failed WU will normally be perceived as matter of course with folding, but it may also be out-of-spec hardware as the underlying cause. How could one really know the culprit? The user would not be able to confirm it because it's not something that can actually be tested without having an equivalent but compliant PSU at hand. The link I posted above gives me additional pause for concern on top of my inherent doubts.

Bottom line: I might run other hardware with PSU adapters but no way, shape or form would I run adapters on a quad socket board. I'd rather not run it at all if I had no other choice.
 
hmm, good points all. U fortunatley due to lack of funds I'm gonna side with zero on this
I'll update in a few, gotta dispatch for a sec
Posted via [H] Mobile Device
Were in snow control, yeah...

Anyways, like I was saying, I'm siding with zero on this. If money weren't an issue I'd use a server box, but I did this project for 2 reasons.
To play around and learn some things (never attempted a build this large)
And because it will be cheaper to run than the box it replaced (from which the sale money made this an even trade, no extra moneys)

I'll try a 2p set up tomorrow if I have the time, I just found out I'm getting my hours cut next week so I'm cramming in all I can get now
 
Last edited:
APOLLO said:
Yes, no chance of failure. I wouldn't be concerned with failure but with adequate current being supplied to the board to properly feed all the installed componentry.
Click to expand...
Using adapters won't affect the amount of current supplied to the board. That depends entirely on the PSU and the board itself.
 
You must log in or register to reply here.
Back
Top