SuperMicro X7DAL-E vmods

[STEvil]

Anyone per chance done anything to this board? I've been spelunking on mine but work has limited me from actually doing anything other than tracing a few circuits so far... :(

Figured it might speed up the process if I asked you guys ;)


Uploading pics now.. thanks to serlv for pics btw, my camera sucks ;)

edit, pics.. copy/paste into browser bar.

http://www.img.areyouserios.com/files/IMG_0061.JPG
http://www.img.areyouserios.com/files/IMG_0062.JPG
http://www.img.areyouserios.com/files/IMG_0063.JPG
http://www.img.areyouserios.com/files/IMG_0064.JPG
http://www.img.areyouserios.com/files/IMG_0065.JPG

[celemine1Gig]

Just wanted to take a look, but none of the links seems to be working. I always get the "REMOTE LINKING PROHIBITED" message. :(

[STEvil]

Looks like i'll have to use a different image host.. I need to chop one of those pics to point out what i've identified so far anyways.

[celemine1Gig]

Great. Perhaps I can help you. Who knows? :)

[STEvil]

Edit - got the rest except one, cant get it because it is under that stupid latch thing for the northbridge..

Edit - updated to include IC names, added a few I left out last time..

Edit - updated again.

[celemine1Gig]

IC names? ;)

[STEvil]

Im blind :(

I'll get a magnifying glass and add more when I get back in a few minutes.. left out the 14 legged ic by mistake, and mosfet next to it.

edit - done. Cant read some of them even with magnifying glass. Almost lost my x1600xt to flashlight (complete with sparks!) oops :eek:

[STEvil]

Pics in first post are working again btw.

[celemine1Gig]

BTW, a little trick to be able to read the printing on the chips:

Take a q-tip and put a small amount of white thermal paste onto it. Then smear it onto the chips surface and then rub it off. Part of the paste should remain in the surface and improve the readability. Digicams also help when it comes to identifying chips. Taking photos from and angle with flash helps alot.

Edit: Links still don't work for me. :(

[Nosfer@tu]

SOOOO nice to OC 2 Quard cores :D

Higest I saw on Hwbot.org was 3.7 Ghz I belive.
Im DEFENETLY in for a watch on this one :D

VERY COOL STEvil

[STEvil]

Wish I had a pair of quads Nos, all i've got are a pair of 5120's...... for now ;)

[rgathright]

I see you like your X7DAL-E! I have mine running at 1590 FSB stable, for months now. I used the tape mod and Systool to get it there.

Visit my review to get an idea of what else you can do to the board.
http://www.epinions.com/content_304338931332

[celemine1Gig]

OK, judging from the one pic that is working for me, the three important chips for vmods are the two lm358's and the one intersil 6532.

This should boil down to a vdimm, vtt and perhaps some other mods. Remember, it's just the one pic I'm seeing.

Edit:

VDimm and VTT --> ISL6522 Datasheet:

http://www.intersil.com/data/fn/FN9030.pdf

Sorry for the wrong link beefore. Had read it as 6532.




LM358M Datasheet:

http://www.national.com/ds.cgi/LM/LM158.pdf

Edit2: Just saw your other pics.

Edit3: VCore and VDroop are the two intersil chips right next to the CPU sockets. I can't really read the markings, but I'd guess that it says "ISL6306". ;)
http://www.intersil.com/products/pt/parametric_table_10489.asp

[STEvil]

rga: been to 1600 and your link is broken.

Celemine - 6302 next to top socket 6311 below and to right of northbridge (directly across from top pci-e 16x slot)

What would you guess the 6522 to do?

[celemine1Gig]

I'm sorry, but right at the moment I don't have the time to go into detail and have a thorough look at the datasheets.

I'll do so when I find the time to. ;) Perhaps today, but perhaps it'll take some days.

Edit: Didn't even find a ISL6302.
Are you sure it's 6302 ?

[Nosfer@tu]

Wish I had a pair of quads Nos, all i've got are a pair of 5120's...... for now ;)

Still very EVIL performance

[STEvil]

Oops, might have meant 6306. I'll check and update post in a few mins.

Nope, 6302.

[STEvil]

Updated pic with voltages.

ICS6522 controls northbridge voltage, pins 3 (0.81v) and 4 (1.52v) are FB and COMP, respectively.

COMP and FB
COMP and FB are the available external pins of the error
amplifier. The FB pin is the inverting input of the error
amplifier and the COMP pin is the error amplifier output.
These pins are used to compensate the voltage-control
feedback loop of the converter.

FB to GND will raise voltage? correct? :D

Now to figure out VDIMM..

[STEvil]

An important update.

It seems that the AMB is fed 1.5v and the memory chips 1.8v... hmmm.

[[XC]Atomicpineapple]

STEVil, just had thought. I looked at the thread on 2CPU.com and the AMB running out of steam first as you suggest would explain why Cronos had so little success with gaining FSB when he reprogrammed the SPD chips of his memory for lower frequencies/timings. Based on that an AMB and NB voltage mod might help more than a vmod for the actual RAM chips themselves.

[STEvil]

There's so much stuff recieving 1.5v that modding the ICS6522 chip might raise AMB, Northbrige, and PCI-E voltage all in one go.. which isnt exactly bad :D

Kind of missing the simpler circuits on "average" hardware now.. this is a mess lmao :D

[[XC]Atomicpineapple]

Well AMB and NB in 1 might be alright, not so sure about PCI-E, dont want to kill my 8800GTS

[STEvil]

bumping up PCI-E a bit should be fine as long as its not overdone... 1.65v tops i'd say, which given the heat the AMB's and NB produce is close to tops for them anyways.

[[XC]Atomicpineapple]

I think if voltages go up a look at some better AMB coolign might be needed. NB is easy, WC it or heatsink it, plenty of purpose built products there, but the AMBs arnt so easy.

[STEvil]

Found the resistor to shade with pencil, but no effect. Might be my pencil sucked or there's a protective coating on the SMD.

4th pin up on right hand side of 6522 in the picture of post #5 is FB. SMD resistor (15.2K) next to it is the right one I think.

Gonna have to break out the soldering iron.. ;)

[Nosfer@tu]

or you could use a Glasspen to remove the protective coating ?

[STEvil]

If I had one I guess.

[Nosfer@tu]

If I had one I guess.

You could also use acetone or sand paper.

Acetone should work okay, BUT Only put it on a cotton stick.
And dont use 2 much, it is not so healthy for the PCB.
But you propperly already knew that :)
I think that would work.

[STEvil]

Will check into it later.

Managed to kick in the OVP protection already.. oops :D

[Stoolman]

I just got a Intel V8 system tossed my way... but the intel board is no good for OC.. I just ordered the Supermicro X7DAL-E board today, and as i understand, systool allows for overclocking this board.. although i could not find any spesified clockgenerator that stated supermicro anywhere..

If there is a chance.. could any of you workstations class gurus post a detailed mod picture of the Vdimm and the Vcore mods here?

I sure would appreciate it..

It is gonna be fun go see how these Xeon 5365 quads will like some heavy duty water cooling.. I am gonna pair the system up with a 8800 ultra card, and the system allready came with 4GB of FB dimm...

So, if I only could get this OC thing going now.. Looking to take down Wprime WR...

[STEvil]

Its the very top IC in systool if you have version 728 or 730. ICS69401 or something iirc...

[STEvil]

Attached some stuff to it today. Took the middle pin on right side of ICS6522 (1.52v) down to 1k ohms with no effect on any voltages.. Pin to ground reads 626 ohms stock while running.. should have had an effect by 1k ohms no?

Suggestions anyone?

[STEvil]

ICS6563 controls AMB voltage.

Here is the datasheet: http://www.intersil.com/data/fn/fn9126.pdf

It appears that if we set it to VRM9 mode we can run up to 1.850v. K9 is limited to 1.550v and VRM10 is limited to 1.600v

The "OFS" pin may allow further adjustment as well as the COMP and FB pins.

I have not mapped out where the pins go yet, the contact points on the package of the chip are too small for me to solder to.

[[XC]Atomicpineapple]

Thats some good news! Now can we force it to work in this VRM 9 mode? And do we know what mode its working in by default? Raising the voltage to 1.6V might get us somewhere if its the AMB thats holding us up.

Was your theory that NB,AMB and PCI-E voltage are all controlled from the same point correct?

[STEvil]

Not sure, havent raised the voltage yet.

I assume we might be able to use the SMD components around the ISL6563 to control it, but I need to get a magnifying glass to see there and trace out the circuits.

[[XC]Atomicpineapple]

I think we need to take a good look and AMB cooling before we give them more volts. SANDRA gives me a max temp of 95 degrees for my AMBs and theyre running at 55-65 with some hacked up P3 heatsink pieces stuck on the AMBs with a ceramique/superglue mix and a 92mm fan blowing on them.

[STEvil]

I will see what i've got in my spare heatink box tomarrow maybe. I can bench at 400mhz, 386 is 24/7 stable. Use anything from 60mm to 120mm fan over the FB-DIMM's.

Its significantly lower with no cooling ;)

[boblemagnifique]

I am interest for the vmod vcore , Vnb for the supermicro :D

[uOpt]

So did any of this lead to higher stable clocks?

Over at 2cpu people didn't have too much luck.

[Nosfer@tu]

I am interest for the vmod vcore , Vnb for the supermicro :D

Dammm, Watch out. Now boblemagnifique is joining the game :p

[STEvil]

sorry uOpt, but work is keeping me too busy to play :(