[H]orde build underway!

[TechFre@k]

Those $75 cpus.. are supposed to be somewhere along the lines of $600-800 each..

Lets see.. Newegg has the 8347 listed here: http://www.newegg.com/Product/Product.aspx?Item=N82E16819105200

How about you do some research before spouting off?

LMAO....and there it was

 

[sabregen]

Just a FYI, the PERC 5/i is going to be a bottleneck with more than 4 10K SAS drives on it. My 5/i topped out at roughly 250MB/s with 4 15K drives... when I switched to an Adaptec 5805 my speeds increased to about 350MB/s.

Thanks for the info. But I am only going to be running the SAS 5/iR initially (which, incidentally) is only 4 drives. Hell, it was cheap enough. The controller was $12, and the drive was $25. after shipping, $50 for both.

 

[sabregen]

also on a hardware related note, how do you justify spending $400 on a motherboard for a pair of $75 cpus?

I didn't spend $400, I bough an open box board. Mostly because I can't justify spending $400 on a board for 2x $75 CPUs . ultatryon@ecc is right though, anyone who bit on these CPUs got a steal, so the board expense is likely justfied for most.

 

[sabregen]

Okay, I'm going to be picking a new for the build tomorrow night before I even unpack anything, so there's about 20 hours left to drop in your suggestions.

I'm going to be spending the next few hours bringing up VMs (the double nested ones).

Also, I will take the first pictures tomorrow night, of the way the initial build was, and progress shots along the way.

I'll probably resort to keeping all of my responses to one post, instead of responding with a post to each comment/suggestion here in the thread.

 

[TechFre@k]

I didn't spend $400, I bough an open box board. Mostly because I can't justify spending $400 on a board for 2x $75 CPUs . ultatryon@ecc is right though, anyone who bit on these CPUs got a steal, so the board expense is likely justfied for most.

Wow, that was much nicer than I expected from you....must of been a good day.

 

[sabregen]

Wow, that was much nicer than I expected from you....must of been a good day.

Actually, it was terrible...jackass

 

[TechFre@k]

How about CubeZilla or Fustercluck? Actually [H]orde might be appropriate with all the drive space and what you intend to do with it.

 

[sabregen]

How about CubeZilla or Fustercluck? Actually [H]orde might be appropriate with all the drive space and what you intend to do with it.

added

 

[sabregen]

ESX servers figure out what I was up to. After adding the iSCSI storage to each of the ESX hosts Storage Configuration, I created a new VM called Sniffer on the LUN I created for it. When I went to power it on:

"You may not run a virtual machine inside of a virtual machine"

I've seen this discussed before, but I haven't the energy for it today. Looks like i have some digging to do. I'll go ahead and create the templates for the rest of the VMs.

 

[stainremover]

Those $75 cpus.. are supposed to be somewhere along the lines of $600-800 each..

Lets see.. Newegg has the 8347 listed here: http://www.newegg.com/Product/Product.aspx?Item=N82E16819105200

How about you do some research before spouting off?

how about.... some.... real research and not just newegg prices?

http://en.wikipedia.org/wiki/List_o...elona.22_.28BA.2C_65_nm.2C_high-efficiency.29

(1) they're BA stepping, so they have TLB bug. TLB bug means very few corporate environments will use them. at the very least no environment that needs an 8 chip platform dares to touch them. that means they do not hold any more value than 2347's. which newegg sells for $260, 1/3 of the "800 dollars" you claim them to be worth, without discounting for the fact that they are used AND that they have the TLB bug.

http://www.newegg.com/Product/Product.aspx?Item=N82E16819105187

(2) a quick search on ebay reveals that it's not hard to get this kind of price for similar or equivalent chips.

buy it now price of $160 - cheaper than what you paid.
http://cgi.ebay.com/AMD-Opteron-Qua...et-F-Pair_W0QQitemZ140273639351QQcmdZViewItem

slightly faster chip, only ended up selling for $100
http://cgi.ebay.com/AMD-Opteron-235...Quad-Core_W0QQitemZ110299498659QQcmdZViewItem


(3) many OS will enforce the TLB fix, which means you will be seeing terrible performance from the chips.

(4) on top of the terrible performance, they are clocked very low. they're the slowest chip in an already poorly performing family (the family of BA stepping barcelonas). this is the equivalent of spending $200 on a motherboard because you got a killer deal on an Athlon X2 4400+

 

[sabregen]

ultatryon@ecc & stainremover - take it to PMs, please. this is a worklog thread. thank you.

Update: Athena Power 5.25" 4-in-1 enclosure purchased for $30 shipped. Do I buy the other SAS drives now...or wait? Better wait on those.

 

[ultatryon@ecc]

you are right, sab

 

[[H]ea\/y B]

Dude...BOHICA? as in Bend Over Here It Comes Again? appropriate, for the amount of $ this is requiring.

What the hell is WOPR?

WOPR was the super computer in the 80's movie WarGames

 

[[H]ea\/y B]

Or maybe project Nuk Nuk

 

[sabregen]

Well, it's time to select the name of the beast. Given the backstory of this machine, and how I got the original build parts (they were given to me), and the fact that I really do gather all the pieces I can get my hands on (and always find a use for them), and the idea of folding on this machine...and incorporating the beloved [H] into the name:

I HAVE SELECTED [H]orde AS THE NAME OF THE BUILD.

Now for pics of the new parts: In the boxes

 

[Arcygenical]

I HAVE SELECTED [H]orde AS THE NAME OF THE BUILD.

Good man

 

[sabregen]

Here we go! I'm going to say now that tonight ended bitter sweet. It all came together, but when I went to boot the machine, I got no POST, no video. The attempt at a hotflash (on the previous server board, whose BIOS chip was only different from the new boards' by 1 digit) resulted in an error message that basically said the BIOS image I was attempting to flash didn't match the one that was in the board. Looks like I will have to do some research in the morning. Props to TechFre@k who took the pics, handed me zip ties, gave me insight....and drank my beer. Couldn't have done it without you (minus the beer part).

Without further aduei, I present tonights activities!

Parts out of the box

Naked board, CPU socket covers removed. Notice that there's no standard Socket F CPU HSF retention brackets on this board. It literally arrived to me in a static bag, no IO plate, cables, etc. The CPU sockets did not have the requisite Socket F 3.5" pitch mounts....what to do?

Luckily I had some Socket 939 brackets left over from previous build were the HSF retention brackets had been removed to install aftermarket coolers that used backplates, so the entire rentention assemblies had been removed on those builds to install the aftermarket cooling. Like I said, I tend to [H]orde, so I still had these laying around. However, when laid out on the Socket F CPU area, it's obvious some mods have to be done to get the new HSFs mounted.

Took some time to think about the CSF HSF / retention bracket issue, and decided to clean the CPUs with Arctic Silver 2 step cleaner

CPUs gooped up, and in sockets...locked and loaded. RAM is installed.

I determined that the existing Socket 939 brackets could be modded to accomodate Socket F mounting. Having built a lot of Socket F systems at work, and knowing how the rentention mechanism works, I figured out what had to be removed from the stock plastic bracket. I tried scissors initially, no go, plastic was too thick. I tried a pair of Henckle kitchen shears that go through chicken bones, but that was a no go. TIME FOR THE DREMEL!

Dremel work on brackets

Notice that there's a normal (black) set of brackets cut, and also a yellow pair. The yellow ones came from the infamous eVGA mATX NF4 SLi Socket 939 board, which is in my wife's machine. I'll never use them for anything, and THE [H]orde MUST GO ON! Time to file the rough edges created by a 30,000RPM cutting wheel on plastic. Files at the ready!

It's not very elegant, but I figured that the least that I could do is situate the yellow HSF retention brackets on the inside mounts. The Scythe Andy Samurai is pretty big, bigger than I had expected (although I'm no newbie to large 120mm air coolers). The reasoning was that the inside mounts would likely be the least visible ones, if you were someone looking at the guts of the build.

Scythe Andy Samurai's installed. Notice the wrench? Due to the modification of a Socket 939 HSF bracket, the wrench was necessary to clamp the Andy Samurai's locking mounting mechanism closed. My fingers just couldn't manage it.

Moving on the the drives and the case now. These 6x WD 320AAKS drives were purchased from axan. Thanks man, they showed up in great condition! These 6x SATA drives will be attached to the Nforce 3600 Professional SATA headers (there's 6x) on the motherboard. They will be configured as a RAID-5 array.

The drives are in. I don't know, with everything that I'm intending on throwing into that case, if the Enlight 470-MP PSU will hold up (especially once the 8800GTX goes in tomorrow) to the task. I may have to serisouly consider getting a beefier PSU for this build. The reason that this one is being used is because it's the PSU with the longest leads in my posession.

Mortherboard and drives are in. Those coolers are really big. I'd eventually like to go water, or find something a bit smaller, but for now, these will do. At least I got them mounted!

Power cables run as neatly as possible. Like I said, I wish I had a beefier PSU with longer leads, I could have run the power to the board somewhere not so obvious, but I had no choice, at the time, with this PSU. It will be interesting tomorrow when the 8800GTX shows up. It will have to go under the power supply cables that go to the board

All power cables run, zip tied, etc. Also did the SATA cables as best as I could. There's 2x SATA cables that it is possible to see the zip tie on them. This bugs me, and I will likely fix it tomorrow.

Well, I attempted a test fire of the system, and I got no post. I downloaded the newest BIOS to a flash drive, and dug out my trusty old DOS Boot CD. I manged to pry the current BIOS chip of the H8DC8 (old board) out of the socket with an eyeglass screwdriver (after trying a few other things that weren't cutting the mustard). What you're looking at is a plastic coated paperclip that I normally use for shorting the power header on motherboards (when they are being tested) that has been bent to measure, and inserted in a U shape in the bottom of the BIOS chip socket. The BIOS chip is then pressed into the socket until the pins engage. The paperclip has a diameter sufficient to keep the BIOS chip from sinking into the socket too far (in which case I'd be unable to extract the sacrificial boards BIOS chip to perform the hot flash). The idea is that after the DOS boot CD comes up, I can just lift on either side of the paperclip, and remove the original chip, and insert the BIOS chip from the KFN32-D SLI to perform the flash. I fired up the H8DC8 with it's original BIOS, let the DOS boot CD do it's thing, and had the thumbdrive in the USB port with the flashing utility and the BIOS image on the USB thumbdrive. I then swapp the H8DC8's BIOS chip for the KFN32-D SLI's BIOS chip but when I went to flash the BIOS, I got an error. I'm six hours, and just as many beers into it. I have (for the moment) hit a wall, so I'm calling it a night.

Tomorrow the 8800GTX shows up, courtesy of [H]eaVy B (thanks bud), and I have to figure out why the hotflash isn't working. If anyone has any ideas, please let me know. More info tomorrow, stay tuned!

 

[sabregen]

Exact error message usinf the AFUDOS utility from ASUS:

"ERROR: ROM file ROMID is not compatible with existing BIOS ROMID."

Looking for a way to disable the BIOS check.

Update: The BIOS chips have identical identifying information on their surface. I had incorrectly thought that they were only 1 digit apart. They are not...they're exactly the same. If I can figure out a way to disable AFUDOS's ROMID check, this flash should be cake.

 

[sabregen]

Okay. The AMI flash utility that ASUS has listed on their page for this board is not AFUDOS, it's another version of AFU. I downloaded all of the BIOS's listed for my board, and all of the flash utilities. I finally managed to get my hands on the actual AFUDOS.exe. The BIOS ROM name was KFN0401.ROM. Syntax was this:

C:\>afudos.exe /iKFN0401.ROM /n /pbnc
The re-write took, but the system will still not boot. I have power LEDs on the board, and front of the case, but no video, and no POST. Still digging for some more answers on the web, but I am getting ready to pull the board from the case and start checking surface level components and the pins inside the sockets, and check for shorts caused by the case. Here's the output of the above command:

 

[sabregen]

Well folks, I tried everything I could think of after getting the BIOS flashed to the newest release, but the board would still not give video output, no POST, no keyboard lights, and no beep codes. All the LEDs on the motherboard were green and all the fans turned on. I took the board out of the case, and dropped the config down to 1 stick of RAM, PCI video card, 1 CPU, 1 HSF, PS2 keyboard, and the monitor. I got nothing. I then removed the video card (which would normally cause beep codes), and got nothing. I then removed the RAM entirely from the board...nothing. I then removed the CPU, nothing. I swapped the CPU into another socket, and put 1 stick of RAM in it....nothing. I got no video, no keyboard lights, no POST, and no beeps. I called ASUS and they suggested an RMA (but would have to wait until Monday).

I decided to go one step further and check the surface of the board for broken solder points, capacitors, obvious stuff. Here's what I found (I must not have noticed it last night, either in my haste to build, or because the light isn't so great in this room. I mis-read the BIOS chip information last night, so I'm leaning towards the latter):

CPU Socket #1 - The screwdriver wielding Socket Fairy must have visited upon the previous owner of this board.

CPU Socket #2 - Sigh...not good. I've fixed bent pins on Socket 771 and Socket 775 boards before, and straightened more pins on more processors that I care to count. But, when I fixed those, I had a really well lit area, and a huge magnifying glass to do it, and I still went a little cross-eyed. There's no way I'm trying my luck on Socket 1207. Plus, if I did, and made it worse, I'm sure the warranty would be voided.

So, the 8800GTX should be here any minute, the entire board area components of the build have been disassembled and put back in their hiding places until a replacement board comes. I think I'm gonna give Newegg custoemr service a shot, as they will likely be much faster than ASUS could ever hope to be. For now, the hardware portion of the build is on hold. The Athena Power 4-in-1 enclosure, and the Dell SAS 5/iR should be on their way to me, and the 8800GTX should show up today. I'll restart my search into why I'm getting the "can't start a virtual machine in a virtual machine" error in VMWare Workstation.

 

[vsboxerboy]

Damn, I just read this from post 1 and got to the sad, sad ending.

Sorry to hear that and I hope that she will be up and purring as soon as possible.

 

[GenesisFactor]

wait...what happened??? NO!!!

 

[Arcygenical]

CMOS checksum destroyed.

Well, that doesn't sound scary at all...

 

[sabregen]

CMOS checksum destroyed.

Well, that doesn't sound scary at all...

I know, sounds scary right? Well, according to ASUS, that's what it's supposed to say when you have a successful completion of the process. It doesn't translate very well, that's for damn sure! I have the Newegg RMA printed out. I guess by the time the board comes back, the rest of the parts will be here. Here's to hoping that Newegg doesn't send me another botched board. If that happens, I will just RMA to ASUS. I would hope (if Newegg was smart) that they would sell freshly RMA'd boards as open box, and not just keep rotating the same 10 boards or so around. I have no idea how they work that, but I think the most likely would be:

1) Customer buys NIB board from Newegg. Board isn't right, or customer is an idiot. Newegg gets the board back.

2) Based on customer reason for the return to Newegg, Newegg may opt toRMA the board to manufacturer. If the customer lies and doesn't admit to something they did that was retarded, then I imagine the board makes at least 1 more trip to someone's house before being RMA'd/

3) If the customer said the board was defective, or admitted some sort of snafu that is hardware related, hopefully Newegg RMAs the board to the manufacturer

4) RMA'd board come back from manufacturer, and then Newegg relists available stock of the whatever the model # is, and somone else buys it.

The part of the process that always causes the breakdown of efficiency...the human element. Some people don't even know enough to know if what they did broke it...sigh...and I end up buying the damned thing with hopes on high for a sweet new rig by weekends end.

 

[TechFre@k]

The part of the process that always causes the breakdown of efficiency...the human element. Some people don't even know enough to know if what they did broke it...sigh...and I end up buying the damned thing with hopes on high for a sweet new rig by weekends end.

Unfortunately there is no patch for human stupidity. I am totally gutted about that, was hoping to see what it would do today, and of course drink more of your beer....actually I was going to bring some over as well.

I think I would probably just RMA to Asus, I know that it would take longer, but getting it right the first time would be great if you know what I mean.

 

[sabregen]

do you have a decent magnifying glass? I don't know if I can let this rest.

 

[TechFre@k]

do you have a decent magnifying glass? I don't know if I can let this rest.

I have a great one at work, but not here. I have a pair of really strong reading glasses though, it might work.

 

[sabregen]

I'm watching top gear Season 11 and straightening pins with the naked eye. To make this exercise completely pointless....I should start drinking.

 

[TechFre@k]

I'm watching top gear Season 11 and straightening pins with the naked eye. To make this exercise completely pointless....I should start drinking.

LMK how it goes.

 

[sabregen]

Okay, well....the pins got straightened, I got the BIOS chip for the KSN32-D hotflashed on the H8DC8, without having to tell AFUDOS to ignore the ROMBIOS mismatch. I literally walked the BIOS chip flashes up from the very first one to 0403 (barcelona support came with 0380). I still have no video, and the board doesn't beep when it has no RAM installed. It does however, show a CPU warning light when you only install in Socket #2 instead of CPU #1.

I think it's a moot point. The board doesn't complain when you attepmt to boot it with no RAM, CPU, or video card. It just turns on. Reeks of dead board. At least I've gotten very good at the hotflash process.

 

[Farva]

+1 for Gibson

 

[sabregen]

+1 for Gibson

Call it what you want, but I do not understand this reference at all.

 

[TechFre@k]

Lame, are you gonna RMA to Newegg or Asus?

On a side note, my Zalman case has shipped from Cali, should be here Tuesday

 

[sabregen]

don't know. you should come over and give me another set of eyes maybe a fesh perspective would help.

 

[nitrobass24]

Shit d00d that sux...i just got home and read through...I would RMA to Newegg and have them send another one ASUS is a little slow for my liking.

 

[sabregen]

nitrobass24 - good timing, I was just lookign to see if anyone had any ideas or had posted anything since my last post yesterday.

I have already generated the Newegg RMA, and it's all boxed and ready to roll out tomorrow. It's very frustrating, but on the upshot, by the time that I get the board back, I should have the Athena Power enclosure, and the Dell SAS controller. Obviously, I have the rest of it, including the initial 1x Fujitsu 10k 2.5" drive that will serve as the boot volume. I figured that Newegg would also have faster turn around on the RMA. If they screw me again though, it'll be the proverbial slow boat for me to RMA through ASUS. There may be a tertiary option though. I called ASUS on Saturday morning, and the tech support tech suggested that I call back on Monday and ask to speak with the BIOS department to discuss my issue, and potentially get a secondary BIOS chip (even if I have to pay for it, I imagine it wouldn't be more than $20) that's already flashed to the latest revision, by ASUS themselves. He also mentioned that when an end user has experience, and is confident that they can replace the BIOS themselves, ASUS considers this route as a potential resolution, instead of using time/resources on both ends to handle the RMA of an entire board. He suggested I try this option as well, if I have legitimate concern (and I do...hotflash in general just doesn't sit well with me as a reall good/solid solution) that the BIOS chip may be corrupted, or at a level that will not support my CPUs.

In short, I think my plan is to RMA the board nack to Newegg (maybe I'll get one with proper brackets for Socket F, and a few extra accessories), and attempt to obtain the newest revision BIOS on a replacement chip directly from ASUS. If that all goes awash, then the board will be going directly to ASUS RMA Dept.

 

[nitrobass24]

Well just because the cpu pins were bent I dont have confidence in that board regardless of the BIOS issues. So I would RMA it to newegg and get another one. At the same time IDK how comfortable you are with hotflashing but you could on monday spend the $20 and get a new bios chip with latest bios and just plop it in when the RMA board shows up.

Also YGPM

 

[sabregen]

Well just because the cpu pins were bent I dont have confidence in that board regardless of the BIOS issues. So I would RMA it to newegg and get another one. At the same time IDK how comfortable you are with hotflashing but you could on monday spend the $20 and get a new bios chip with latest bios and just plop it in when the RMA board shows up.

Also YGPM

PM responded to. As part of this weekends activities, I have gotten very good at doing the hotflashes, it's just that I'm not sure that I trust them so much as I trust having a functional board that I can do the update on, without the involvement of another board.

 

[Hilmarhd]

You should probably kick someone in the nuts!

 

[wardour]

I am still shopping for a board for my 8-core build. I'll be watching this one, good luck with the RMA.