Particle why only 2 memory sticks? Or are those two kits of 2x1 GB?
Printable View
Particle why only 2 memory sticks? Or are those two kits of 2x1 GB?
This is really quality stuff :up: Almost double of what I can do on my machine :shocked: (granted that you have twice as much processors, but we all know how things are scaling up ...)Quote:
Originally Posted by jcool
It is just two 2GB sticks of ECC registered DDR2-800. That provides the bare minimum required to operate should I find that there is an issue using unregistered RAM. I'll order more if I have to use ECC/Reg. If not, I have what I believe to be a full compliment of regular. :)Quote:
Originally Posted by Nedjo
PICTURES
Here they are!
As always, click for larger versions. Sorry for the unusually poor quality, but I didn't have a proper lighting setup this time.
Motherboard, test RAM, and CPUs:
http://www.pcrpg.org/pics/computer/o...7/group_sm.jpg
Opteron 2427s--serials are only two apart. :)
http://www.pcrpg.org/pics/computer/o...7/front_sm.jpg
http://www.pcrpg.org/pics/computer/opty2427/back_sm.jpg
Awesome pics particle, can't wait to see you fire it up :D :up:
Thanks mate, although they still get owned badly by my Harpertowns due to lack of clockspeed...Quote:
Originally Posted by Lightman
Anyway, "real life" performance the Gaisntowns are better even with more than 1Ghz lower frequency (max primestable on those X5450's is 4,06Ghz).
Can't wait to see Particle's Hexie :)
So is the BIOS on the mobo new enough for the hexies?
Not sure yet--getting ready to fire it as we speak. Even if not, this board may have an ace up its sleeve:Quote:
Originally Posted by Utroz
Looks like it can do a "recovery mode" flash without a working CPU. All the minimum requirement items can be controlled directly from the non-programmable flash and chipset only.Quote:
Originally Posted by Supermicro BIOS Flash Readme
LGA, how I love you. Too bad AMD won't adopt you for their desktops CPUs :(
Get this baby fired up, I'm really curious how it'll perform.
Wow! Those serial numbers = AMAZING. I love the Opteron LGA package. I can't wait until the desktop socket gets that treatment.
Promising..."CPU Mismatch Detected"
http://www.youtube.com/v/-EjZmGvrdrE
It's interesting you say this will work without a detected CPU - I guess boot block code runs before POST, so makes sense. Nice to know I have this option too on the K9ND. Thanks for the info :up:Quote:
Originally Posted by Particle
NOOOOOOOOOOO!!! LGA delayed my build no end - I accidentally dropped a CPU into one of my sockets from amout 20mm above and twisted a few pins beyond home repair - had to pay to send it to MSI USA and have the pins repaired, and lost about 3 months, IIRC waiting for the whole process. :shakes:. I say AMD should bring back pins, especially here in the server market, where it makes sense to have the weak point on the CPU, which is likely to be rplaced more often/earlier than the mobo...Quote:
Originally Posted by [XC] gomeler
It's not hard to bend CPU pins back into place when they bend, and that's a very rare ocurrence in itself anyway. LGA, on the other hand :mad:. I mean, how hard can it be to go from 940 pins --> 1207??
When I try the recovery method, it reads the first bit from the floppy and beeps twice then beeps another five times--all short. Then it appears to start over and try again and will loop as such indefinitely. Anyone know what this means?
Neato thread! Grats on the cool hardwarez. Can you try and boot with one cpu?
The bios recovery for that board sounds a lot like the procedure for the ami bios msi boards. Did you try to completely clear cmos before you forced the bios recovery?Quote:
Originally Posted by Particle
I know you have a server supermicro board but check out these instructions from the msi forum :
http://forum-en.msi.com/index.php?topic=119375.0
Maybe you renamed the wrong file? Or didnt get the exact name right? Or have the wrong bios?
Maybe there is a jumper that needs to be moved for the bios recovery to work and allow the bios to be reprogrammed without a CPU?
Oh bless you! The Supermicro documentation was just WRONG. Instead of requiring the rom to be named SUPER.ROM, it had to be the (what I assume is the generic AMI default) AMIBOOT.ROM. Bingo, flash successful. Now for further playing...Quote:
Originally Posted by shadow_419
Unfortunately, after the flash, the board appears deader than a doorknob. Won't even turn the fans now. *shrug*
Are you able to force flash to the original at least?
No, somehow it is superdead now. Even the non-rewritable part of the BIOS (the part that includes the BIOS recovery) doesn't appear to function. No power light, no CPU fan, etc. Only the chipset fan spins.
Well that was anticlimactic.
did you test the floppy for errors?
Yes, I computed a CRC32 hash of the ROM on hard disk and checked the copy on the diskette. It matched.
Well that sucks :(
I've been watching this thread to see your results. :(Quote:
Originally Posted by Particle
Hopefully you'll be able to bring it back to life. I have confidence in you man!
I'm assuming you did all the generic stuff, unpluging, pulling battery, checking for shorts etc...
That really does suck, I wanted to see what that puppy could do...
Try running with a single processor. Whenever you're trying out a new CPU with an old Mobo, always keep an older opteron 2x00 handy. It helps with booting into the basic systems so you can update the board and then run the newer CPUs.
Other than that, I'm guessing its the Open Box board, not the chips. Also, ty for the pics as promised :D :D :D
Now we just need a close up of the ram since you have so much free time now :stick:
Perkam
I think the video shows him running it on a single processor. Damn I've been looking forward to this :mad: (nothing compared to you).
:censored: *cries* NOOOOOOOnOOOOOnnnOOOOOO! :(
Pics, awesome, video, awesome, dead board :down:
Bummer, this dose not put a dent into your original plan of getting bios chip right?
Maybe you can do a replacement and ask for newest version of the board with newest bios.
Tough luck mate... RMA time? :shrug: :(
Unless the bios chip is replacable and you can get one with the latest bios on it quickly.
Yes. All this really does it put me back where I expected to be in the first place after today. Remember when I said this would be a multi-part adventure where I'd get all the stuff and then order a BIOS chip since I'd be able to see the physical markings on the BIOS chip itself? Well, that's all we're back to. It was just teasing me in the mean time like I *might* have been able to bypass that since it had the BIOS recovery method and appeared "functional". What a cruel mistress.
My best guess is that during the BIOS recovery action, the boot block of the BIOS itself became corrupted. I would be surprised to find actual damage to the board somehow. All I'll need is a new BIOS chip with the latest version flashed to it. An RMA would be overkill for this kind of situation (and take a lot longer). heh
I already ordered the new BIOS last night, so we should be back in business next week.
Some more images that were requested:
http://www.pcrpg.org/pics/computer/opty2427/ram_sm.jpg
http://www.pcrpg.org/pics/computer/opty2427/bios_sm.jpg
This might be crazy, but I've found another motherboard that also uses a PLCC 32-pin EEPROM BIOS that just so happens to be LPC 3.3V. 33MHz. It is 4Mbit versus my board's 8Mbit, but I think I'm going to try and hot flash. I've got little to lose at this point. It's an old unused board anyway.
Unsuccessful (Sgt.McRuff is guessing)?
Well, the board can't even seem to flash its OWN BIOS, yet alone do my hotflash. Trying to get it to successfully update itself first.
AFUDOS fails with "Error: Problem allocating memory"
Look for BIOS Protect jumper on the board (or this option in BIOS itself).
I'm waiting patiently for your system to boot properly! Nice batch of CPU's you have there! :up:
Seems to be an issue with that particular motherboard. It can't even flash itself, yet alone do a hot flash. Others have had the same problem, and there isn't a protection jumper or setting to be found. I'm using a BIOSTAR motherboard now instead and trying to use UniFlash to write to my ROM. It's promising, but now my floppy drive itself could be bad. heh...it never stops. Neither do I.
Use an USB stick for flashing. ;)
Floppies are dead.
http://files.extremeoverclocking.com/file.php?f=197
Saves time and headaches of corruped files.
Flashing with uniflash in the BIOSTAR was problematic. Got it to flash, but like three cells failed to verify. Tried it in the Supermicro and the boot block is apparently still fubar.
--
I've never had that tool actually work, but I've got a spare 128 MB at the moment that's worth trying. I'll update you.
It works for me on 512MB SD card. Boots perfectly! I'm using that card to flash my ATi cards.Quote:
Originally Posted by Particle
Keep trying! :up:
Scavanaged another SST 49LF080A 8-Mbit flash from my old DFI ICFX3200-T2R/G that died after a BIOS flash. It fails as well, but in different blocks. Thus, the problem is likely physical instead of my flashing setup. I'll try flashing in another board anyway just to be thorough.
BWAAAAHAHAHAHAHAHAHA
http://www.pcrpg.org/pics/misc/horlyshoot_sm.png
She lives.
That pic is like right after the ads and the starting of a movie that you herd was awesome.
Having some trouble getting fire when two CPUs are installed now though...working on it.
:party:Quote:
Originally Posted by Particle
congratulations this are looking up now
Ok, so it fires with two CPUs installed now--it was a memory slot issue I guess.
However, the BIOS only detects one CPU. *scratches head*
This has been the strangest build I've ever done.
This is getting stranger by the minute... hope you can sort it out. I hope it's not a damaged mobo or one of the CPUs...
Either CPU works great by itself when in CPU1. On the rare occasions that I can get it to actually boot with two CPUs installed, however, CPU2 doesn't even seem to receive power. The heatsink never gets above room temperature while CPU1's always does. I have to wonder if it is memory. The manual says I can run two CPUs with 2 DIMMs installed into CPU1's banks and none in CPU2's, but who knows. The manual was wrong about many other things regarding memory already. It seems like an odd condition to withhold power on, however.
Usually each CPU with integrated IMC has to have its own ram.. I would say screw the manual and try a different mem slot configuration?
Atleast she's breathing! Can't wait to see this build completed, it's been a sort of inspiration for my next build.
I had similar behaviour on old Athlon MP setup. Problem was unregistered memory. After switching to Reg. DDR all worked fine.Quote:
Originally Posted by Particle
Hope it helps ...
Have one DIMM on each CPU right now and it booted for some reason. Only one CPU is detected and I only get half my RAM, but here's something interesting. CPU-Z can detect both memory sticks. I can also see CPU2's temperature sensor. Something really weird is going on.
12,630 in CB10 with one CPU at 2200MHz and single channel memory btw
Try swapping cpus, to rule out one bad one? Any other memory you can test with it? Cmos clear?
He already swapped the CPUs around.. see #145
Gawd how great it would be if ppl would actually READ a thread before posting.. not directed at you Sgt. McRuff, just a trend I noticed on here that's getting more and more evident every day.
:sofa:
Sorry dude, I must of misread it, dyslexia is a *****.
Nothing to worry about, like I said I did not mean to attack anyone, just an observation I've made recently. And most people REALLY don't read properly, unlike you (sorry bud...)
I've got six 1GB ECC reg DDR2 DIMMs coming tomorrow from some family in KC. I should be able to confirm or eliminate memory configuration as a possible cause at that time. In the mean time, I need to figure out a way to jack the HT link up so I can overclock things. There are zero settings for the speed of anything but RAM. nTune also just crashes on load--before it even does anything.
SetFSB? Can you tell us which PLL?
It's integrated into the MCP55 southbridge I believe. I don't have the experience to try and pick out a chip on the mobo, either. Don't know what I'd be looking for. If you have some pointers, I'd be willing to try. :)
Will be a small to medium-sized black chip, should have 3 lines of writing on it.Quote:
Originally Posted by Particle
First should say something like ICS932S421BGLF (that's my Tyan S7002's PLL) generally something starting with ICSxxxxxxxx.
Check here for a complete list of supported PLLs. The chip should be located near the PCIe/PCI slot area.
Particle do you know how to tweak with R/W everything tool ??
u can overclock mcp55 with that tool and tweak ram timings thru registers
I'll start looking immediately.Quote:
Originally Posted by jcool
No, I wouldn't know what to do with it to get the desired effect. Is there a guide and a download link?Quote:
Originally Posted by msimax
I use to have problems with systool on my Tyan 5000i board.
It would crash the system everytime when starting the application.
as a workaround I did :
go start / run
type "SysTool.exe" -fsb ICS932S421:360
(360 is the target FSB )
I don't think that applies to me as a fix yet--at least not until we can identify the PLL. I likely don't have the ICS932S421. I can't find any black chip anywhere on the board starting with ICS*. Could it be under a heatsink?
particle i can send u a guide send me your email in pm
Could also be starting with CV or CY or RTM or SLG or W (see setFSB list)Quote:
Originally Posted by Particle
Usually PLL chips aren't under any heatsinks. Got a high res pic of that mobo maybe?
Glad to hear the board isn't dead, hopefully you can figure out what's up with the memory issue.
Get a high resolution pictures and we can find the PLL for you :up:Quote:
Originally Posted by jcool
I'll get some high-res shots of the board as soon as I can. I'll take off the oversized chipset heatsinks too for good measure. Thanks for offering to take a look :up:
---
I've been following the guide that was emailed to me, but the register doesn't appear to make a difference like it does on other MCP55 systems. Perhaps this supports the idea of an external PLL?
---
Almost forgot to take an initial CPU-Z shot for you. Here's the default stuff. CPU-Z appears baffled by this chip so far. :)
Click for details.
http://valid.canardpc.com/cache/screenshot/598764.png
I love the cpu-z it shows 0nm !!! LOL!Quote:
Originally Posted by Particle
SUCCESS!
Found a version of the nVidia control panel that -doesn't- crash me. It can do virtually nothing and thinks my PCIe clock is 2.5GHz, but it does what I need. HTT multiplier and HT base. :) From here she be pure gravy.
Demo:
http://valid.canardpc.com/cache/screenshot/598769.png
:clap: Awesome! Now let's see 3GHz :D
You teaser! :D
Show me more!!
Also upgrade CPU-Z;)
:shocked:
Awesome. Can't wait to see how far it'll go!
At 2.5GHz right now--haven't tried higher. Stable enough for Cinebench, but I haven't tested further. I'm going to wait until I know this board is good or not with CPU2 before I risk blowing it up. Hard to ask for a replacement if I melt the VRMs. Granted, I'm at default voltages and at frequencies less than other retail models.
http://valid.canardpc.com/cache/screenshot/598771.png
Jesus that would be a fun rig on LN2. Doubt they'll ever release a 6-core black edition for Socket F though :(
what sort of score did it get at 2.5GHz on cinebench with 6 cores?Quote:
Originally Posted by Particle
Found the CPU-Z 1.52 beta...much better!
Ignore voltage. Everything is still at the default vCore of 1.25V and vNB of 1.175V.
http://www.pcrpg.org/pics/computer/o...b10x64-001.png
Gratz on getting it going man! Nice work.... :clap:
I would assume your running a 32bit OS based on the Cinebench score, good score for a 32bit OS!
Good luck getting the other chip cranking, that sure looks like a fun rig to play with... :D
So you think it is the non-ECC ram that is stalling you with both CPUs?
Does ntune give you full adjustment of the processor like vcore/vnb? Can you adjust the memory divider in BIOS? HTT divider? I'm really curious how high that dual-socket board will be able to clock up the HTT, most AM3 boards can do up to 350HTT, 272HTT on that server board would be awesome for a solid 3GHz.
There's some potential I would imagine. It is even an HT3 capable board, and that's fairly new for Socket F.
For clarification, I'm using ECC registered DDR2-800 memory already. I just need more of it before I can confirm if the board is bad or the second CPU is a memory issue. The manual states that with 2 CPUs, two DIMMs are required on CPU1. I only have two DIMMs total right now--I thought I could use 1 DIMM on each (minimum required to run in case it didn't work with unregistered, non-ECC memory that I wanted to test against it). However, it may be the case that 2 DIMMs are required on CPU1 -and- memory must be present on CPU2 before it will initialize. As such, I need more memory before I can test that theory.
Cinebench was ran on a very much not fresh install of Windows Server 2008 x64 with Arma II open in the background. At 2500MHz though, it's still not a bad score. Some tweaking is still required, however, since at 2200MHz I scored virtually the same.
Hey, Particle been really sick for past few weeks and not on the forum.
See you have the board and had problems flashing. You need to put a the Boot Block Jumper in reverse position in order to Flash Bios on the Board:)
oh man i wonder how well she'll scale on 2 socket with cinebench!!
also, does it REQUIRE ecc ram modules?
i think u might be able to get away with using ddr2 regular modules becuz it's not a chipset limitation it's more of the imc.
:rofl: Games b4 benchmarks :up:Quote:
Originally Posted by Particle
Damn good score any ways. I bet you have arma 2 open, tf2 , bf2, fear2 and hl2 open all at same time to alt tab with no lag between them all :ROTF:
Erm, but I guess you would need more memory for that.
Glad to see you around again, Esau. I was wondering where you were in all of this. :) Is there really a jumper on these Supermicro boards? I think you're just pulling my chain here, as I recall re-flashing it again just to be safe once I actually managed to get booted up.Quote:
Originally Posted by Brother Esau
Ninja Edit: Ohhhhhhhhh, I get it now. You mean like rotate the jumper. :) I thought you were getting at the potential fact that there's a 3-position jumper for BIOS protection and that it needed to be moved from like 1-2 to 2-3 to flash properly. heh
I don't think ECC is strictly required--you can turn off all ECC options in the BIOS. However, I'm pretty sure that registered memory is required. The two aren't the same thing. You can have ECC memory that is unregistered or registered memory that isn't ECC as well as ECC + registered. Normal desktop memory is typically non-ECC and unregistered. Some exotic desktop systems and Macs have used ECC unregistered.Quote:
Originally Posted by i found nemo
The "registered" thing indicates that there's an extra chip onboard that serves as a buffer. It's either for TX or RX (only one) but I don't remember which. That's as opposed to those FB-DIMMs you've probably heard about. They buffer in both directions. A system requiring ECC or non-ECC, registered memory won't work with desktop memory. Besides, I tested it. :) Doesn't work.
Thanks. heh I was pretty curious to see how well Arma II would hit all the cores. It is one of the best-threaded games I've ever seen in my life. At 3.5GHz on my Phenom II X4 940, it would go back and forth between about 50% and 90%. Most of the time it was in the 60s and low-to-mid 70s, but that's still damn impressive.Quote:
Originally Posted by Sgt.McRuff
Coincidentally, I received about the same framerate at 6x2500 with a single Radeon 3870 as I do on my "gaming" desktop at 4x3500 with dual Radeon 4850s when everything was maxed out. Slightly lower resolution (1600x1200 versus 1920x1200).
i forgot about the whole registered unregistered thing do'h!!
well i hope you get 'er running good
Looking forward to 11x270 :D :D :D
Perkam
I've been stalking this thread since you started talking dual socket.
I hope you get the extra ram soon!
I want to see how this baby rocks and clocks!
Also seems very very good that you run that well on the system with only a single 3870.
Can't wait to see what it does with 12x2500 or possibly 12x3000 :D
Now that's classy. The whole system only pulls 109W idle, including PSU losses, GPU, and everything.
185-190W in 3DMark06 demo mode
165W in Cinebench 10 x64
:shocked::shocked:YOUR MY HERO!!!:shocked::shocked:Quote:
Originally Posted by Particle
You serious?Quote:
Originally Posted by Particle
That's impressive.
What PSU are you using also?
All 12 cores at fulload?:shocked:Quote:
Originally Posted by Particle
Guys.. ONE CPU. Read etc. ;)
Still nice particle, those new chips certainly aren't power hogs :up:
My Gainsy takes 95W idle (with onboard vga) and 275W BOINC with 16 threads running, for comparison. :)
Yes, just one CPU so far. Default vCore on these things is apparently 1.225V. It's still pretty cool considering that's for six cores and a Radeon 3870. It's hard to find exact numbers, but it looks like the idle power of that GPU is supposed to be somewhere between 15 and 25W. It's pretty cool though when idle (with PowerNow turned off even), your CPU's heatsink is cold/room temperature. BIOS has to be off a little bit since it says exactly room temp and there have to be some losses in there (core temp of 66F in a 66F room? nah). No program seems to read it right either, and they're usually even more off than the BIOS since CoreTemp and the like says my CPU is forming ice. I had a box cooler on it for a while there, but last night I swapped it for a Freezer 64 Pro temporarily. The eventual plan is to use a Xigmatek HDT-1283 and a Sunbeam Core Contact Freezer if they both end up fitting.
In other news, last night (technically this morning) I had a 12:45 AM delivery of six 1GB ECC/reg DDR2-400 memory. Sure, it's slower than crud, but it should let me test the CPU2 theory after work. :) Maybe during lunch.
I hope your lunch is before I will need to go asleep :)
I really want to see some scaling of the software with core count. We can also run some clock for clock, but not core for cores comparisons.
Good luck and let the force will be with you! :lsfight:
we want results now, lunch can wait!
Let him eat lunch... Don't want him tired or hungry and kill the thing.Quote:
Originally Posted by [XC] gomeler
:rolleyes:
Now if you could get 12 cores at 3ghz and 4 4890s it might be able to run crysis :ROTF:
Also I would like to see screen shot of arma 2 and how much its using on cores.
Hmm. Tried briefly after lunch as I had a bit of spare time. It didn't POST. I'll verify the DIMMs all on CPU1 after work, but ultimately I think it's going to come down to a bad motherboard.
nooooooooo :(
Ok, time to get a refund for this board. I'll order a new one and take a loss so there isn't a two month RMA delay. :(
so what motherboard now? It seems like amd is being slow about releasing a server chipset.
I'll just get another of the one I already have. It's unlikely that a brand new one is also defective. If it -still- doesn't work, I'd blame my CPU(s) for a dead HT link. Of course, I have no real way to test that.
Oh noes.. sorry to hear that particle. :(
Best of luck with the new motherboard.
Alright, it's official. Refund RMA approved. I'll ship it off tomorrow. In the mean time, I've already placed an order for a brand new H8DAE-2 with rush processing and next day shipping. It should be here Thursday if things go well. This is getting expensive, but I guess ultimately it's still not that bad. Newegg wants $35 to refund the defective board. Bleh!