680i (eVGA, XFX, BFG...) A1 / D00 Rev. VCORE AND VDROP DONE!

[T_M]

couple questions though

Using variable resistors means that you can adjust the resistance to exactly what you want.
The "other" 50k to ground is vcore mod, adjustable vcore including above that which BIOS allows.

[dinos22]

man, that was easy! I did it while the computer was on and could watch the voltage go up as I kept rubbing the pencil across the terminals.

OMG you did that while the PC was on
you crazy man you crazy
errr don't do it again though

[treatmentx]

Works great here, no complaints. And you're covering up a resistor, so technically, you're decreasing the resistance, not increasing.

[Micutzu]

The mod is useful, but not as much as you'd think. I have ran a couple of tests with high loads and i came to a few conclusions.
The setup was comprised of Q6600 B3 on Vapochill LS, Seventeam 600EAZ (very strong +12V, integrated power meter) and an EVGA 680i SLI A1 with 4x2200UF low-ESR caps added on Vcore PWM +12V input and 4x 1500uF low-ESR caps added on the Vcore output; it also had Vcore mod and removed Ilimit, and i pushed the voltage up to 1.9V.
Most important thing is that the multimeter only shows the average voltage and you can't see the ripple and voltage spikes, wich usually are the cause of instability. The Vdroop mod only keeps the average voltage stable, but it actually makes the ripple and spikes worse; for these you would need to rework the LC part of the PWM with stronger coils and better caps, as well as increase the pulse frequency of the PWM (if the mosfets cand handle it) and even tweak the feedback response in a way if possible.

Pushing the voltage very hard made the CPU put big strain on the PWM and showed alot of it's weaknesses; the system drew close to 500W with only the CPU loaded (Orthos + 2x SPi 32M). Tunning for minimal vdroop at 1.45V I couldn't use more than 1.6V stable, the transient load spikes were absolutely awful. Permitting the vdroop to swing about 0.03V allowed 1.7V+ with reasonable stability and when closing in to 1.9V i had to have a droop of at least 0.05V. Remember this motherboard has double the filtering capacity on the output compared to stock ones, wich would probably act mush worse in the same conditions.

Hope this info helps someone, the EVGA 680i SLI is not the greatest tool for hardcore CPU clocks but it can be improved and become competitive with the right tweaks.

[gaz_az]

The mod is useful, but not as much as you'd think. I have ran a couple of tests with high loads and i came to a few conclusions.
The setup was comprised of Q6600 B3 on Vapochill LS, Seventeam 600EAZ (very strong +12V, integrated power meter) and an EVGA 680i SLI A1 with 4x2200UF low-ESR caps added on Vcore PWM +12V input and 4x 1500uF low-ESR caps added on the Vcore output; it also had Vcore mod and removed Ilimit, and i pushed the voltage up to 1.9V.
Most important thing is that the multimeter only shows the average voltage and you can't see the ripple and voltage spikes, wich usually are the cause of instability. The Vdroop mod only keeps the average voltage stable, but it actually makes the ripple and spikes worse; for these you would need to rework the LC part of the PWM with stronger coils and better caps, as well as increase the pulse frequency of the PWM (if the mosfets cand handle it) and even tweak the feedback response in a way if possible.

Pushing the voltage very hard made the CPU put big strain on the PWM and showed alot of it's weaknesses; the system drew close to 500W with only the CPU loaded (Orthos + 2x SPi 32M). Tunning for minimal vdroop at 1.45V I couldn't use more than 1.6V stable, the transient load spikes were absolutely awful. Permitting the vdroop to swing about 0.03V allowed 1.7V+ with reasonable stability and when closing in to 1.9V i had to have a droop of at least 0.05V. Remember this motherboard has double the filtering capacity on the output compared to stock ones, wich would probably act mush worse in the same conditions.

Hope this info helps someone, the EVGA 680i SLI is not the greatest tool for hardcore CPU clocks but it can be improved and become competitive with the right tweaks.

Micutzu - can you elaborate more or give some pointers on the tyoes of tweaks we should be doing and how to do them ??

thanks

[Micutzu]

What I've said applies to high loads and hardcore benching, in 24/7 setups you might not notice any issues.

The most simple mod is upgrading filtering capacity for the circuit's output with capacitors that have very low ESR and can work at very high frequencies; capacity is not very important. Since yesterday i've changed the extra caps with 5x 3300uF Rubycon MBZ, wich are very good electrolytic caps; there are of course Alu ones that have even better performance, but i didn't have enough of them at hand.

But, the most important mod is the base frequency clock for the PWM, wich varies with the value of Rt accordingly to the attached table. Rt is the resistor from pin12 of the PWM controller (ADP3189) and the standard measured value is 118Kohm. I was using about 80Kohm when i finished tests today, and at 1.810V idle i got ~1.790V loaded (full, on all 4 cores) with pretty clean output. The mosfets are getting very hot in these conditions, so you will definetely need a strong fan over them to run a 3DM06 for example. This eliminated the shutdowns in high loads tho', so i'm happy .

[gaz_az]

What I've said applies to high loads and hardcore benching, in 24/7 setups you might not notice any issues.

The most simple mod is upgrading filtering capacity for the circuit's output with capacitors that have very low ESR and can work at very high frequencies; capacity is not very important. Since yesterday i've changed the extra caps with 5x 3300uF Rubycon MBZ, wich are very good electrolytic caps; there are of course Alu ones that have even better performance, but i didn't have enough of them at hand.

But, the most important mod is the base frequency clock for the PWM, wich varies with the value of Rt accordingly to the attached table. Rt is the resistor from pin12 of the PWM controller (ADP3189) and the standard measured value is 118Kohm. I was using about 80Kohm when i finished tests today, and at 1.810V idle i got ~1.790V loaded (full, on all 4 cores) with pretty clean output. The mosfets are getting very hot in these conditions, so you will definetely need a strong fan over them to run a 3DM06 for example. This eliminated the shutdowns in high loads tho', so i'm happy .

Any photos of your mods with a few pointers ??

[louie101086]

will this mod really do anything big for performance

[paul_]

So, for the vdroop mod, would it be OK to just do a few passes with a 2B pencil (got one from my Dad's architecture days) without a multimeter to check my work? Could I seriously damage anything doing that?

[dinos22]

Micutzu
hit up some numbers man

[Micutzu]

No numbers so far, still some work to do plus the pot i'm using barely handles dualcore. The board looks like in the pics, don't mind it's so dirty, i've done alot of testing on it.









Details on the soldering points in a few minutes.

[Micutzu]

Mods :



Original Ilim. resistor is 82Kohm wich limits the current to approximatively 140A continuous / 160A peak; at 1.8V a 140A limit means 250W output wich is just not enough to give hell to a quad ... even on a hard pushed dualcore you might have problems. As I saw swings of ~300W (measured on the PSU output) between idle and fulll load just by putting the CPU to work, i would recommend using a current limit of 250A or more (50Kohm or even less for the pot that replaces the Ilimit resistor).

The most delicate mod though is the PWM freq. increase. The original value of the resistor is 120Kohm, wich means it has a base frequency of around 2Mhz (333Khz per phase); a good way to improve power handling and transient response (very very important in CPU PWM design, although many times disregarded) is to make the mosfets switch faster. By replacing the original resistor with a pot set at 80Kohm, i moved the base PWM frequency to about 3Mhz, wich means every pair of mosfets is switching at 500Khz now, wich is 50% more than the original 333Khz. 50% faster switching means in theory 50% faster transient response and 50% less ripple .

You'd have to take into consideration that these mods put ALOT of strain on the mosfets, as they remain the weakest link in the chain; i've replaced 3 pairs of mosfets until now on this mobo only. Cool them properly when pushing hard, they cannot really handle that much current and stress on the long run.

[dinos22]

holy that looks hard core hahahah

soo o how do you actually measure spikes in voltage....

[mhw100]

Using variable resistors means that you can adjust the resistance to exactly what you want.
The "other" 50k to ground is vcore mod, adjustable vcore including above that which BIOS allows.

Is that the only advantage to the vcore mod i.e. it allows for a higher setting?

I know this sounds naive, which it is but how many strokes of the 2B would this take e.g. if you are doing 6 or 7 on the vdroop, would 3 or 4 do it for the vcore mod? I don't have a meter to measure anything - just the bios and everest.

[gaz_az]

Micutzu

Just a few questions - I am going to try this mod, but want to be very clear about what I am doing first.

I think from your photos I can work out what to do with the pots, but I am less clear with the underside of the board. I have no real experience with electronics, and have some basic questions ....


1)When soldering the potentiometres which legs do I use, there are 3? Does it matter which end goes to which end of the resistor ??

2)When you say replace the Ilim and PWM you mean remove the very small resistor completely and solder the pot to the two points left on the board !?

3)Once connected up and the system running what way to I set the pots ?? Do I wind the screw tight in, then loosen it ?? I presume I use a metre on the two connections of each pot to set the current.

4) I have attached two photos of an empty board. Can you mark which caps need removed, and mark which connections to make where and list the caps. You said Alu caps would be better, since I will be ordering everything can you give the details of the best ones to use.


I know it is a lot, but thanks in advance ......

[Micutzu]

dinos22 > You can see the spikes on an O-scope .

gaz_az > first of all, a friendly advise . I want to make clear that the mods I've done would really make a difference only in extreme benching conditions; as you say you have no real experience with electronics, some of the mods are pretty delicate and would be better left untouched.

1. You need to use the middle pin and one of the other two; I always use the ones marked in red in the small photo so it will decrease resistance (thus increasing voltage, in most mods) when you turn it clockwise. It is VERY important that you will set the pot at max resistance (screw it counter-clockwise until you feell it "click") before installing it in the circuit, and remember to always use a measuring/monitoring device to see what you are doing.

2. Yes, you need to remove the original resistors and it's better to use a hot-air soldering tool for this; you can do it with a regular soldering iron also but you need to practise on some dead hardware first because it's rather delicate and you can touch nearby parts and unsolder them by mistake too. The mods can also be done without removing the original resistors but there are some reasons i chose to do it this way: more linear and precise adjustment, the posibility to use resistances lower or HIGHER than original ones for testing purposes etc. If you want to do it without removing the original resistors, you need to have pots with 10x times the original value (1Mohm instead of 100Kohm in this case).

3. If you connected the pots as described above, you will turn the screws clockwise to start making adjustments. Always monitor what you're doing (measure the resistances for Ilim and PWM freq. mods with the PC turned OFF), and push it little by little.

[Micutzu]

4. You don't need to remove any caps, just add extra on the backside. Red dots for (+), blue dots for (-); try to have as strong solderings as possible, because trough those cap's pins there is alot of current flowing. There are only the Vcore's output caps marked, the +12V extra ones are not really necessary. Let me know where you want to order from and i will give you a few pointers, generally any solid Aluminium capacitor with about 500uF or so will do.

[evoic]

I am nowhere near as extreme as I thought I was before I read this thread.

If what you are doing is the equivalent of riding a bike, I am still about 1 year away from getting my first set of training wheels.

I'll stick with the pencil-mods until I have an extra few years to get my degree in electrical engineering.

Wow.
.
.

[gaz_az]

4. You don't need to remove any caps, just add extra on the backside. Red dots for (+), blue dots for (-); try to have as strong solderings as possible, because trough those cap's pins there is alot of current flowing. There are only the Vcore's output caps marked, the +12V extra ones are not really necessary. Let me know where you want to order from and i will give you a few pointers, generally any solid Aluminium capacitor with about 500uF or so will do.

Thanks for the detailed answers, some very good advice about the pots and I think I would be confident to give it a go ! I do take your point about being a beginner, but I am planning phase cooling for my quad, and in any case I like to jump in the deep end !! I can get my hands on an air solder station, so hopefully with a magnifying glass I would be able to get the job done.

Still a bit confused about the other caps though ......

The four +12v ones you have not marked look a bit more straigtforward. You had origonally said these 4 were 2200UF low-ESR , but in your picture they look like 1200uF. It wouldn't be too much trouble to do them. How un-necessary are they ?? I have marked them in green if you want to check them.

My main proiblem is, I can't quite match the red and blue marks you have made on my photo, to the caps on your photo !! Do I need 5 3300uF Rubycon MBZ's or more?? Can you explain exactly how they are wired . From your photo it looks like you have joined the red and blue dots together, then connected one end to the cap, and the other leg of the cap to where the red line is marked!!?? Is that correct ??

Also you have a red and black wire going from one of the points, and there is no cap connected to 2nd set of red/blue dots down on the right hand side.

I have marked the areas in red I am not sure about, on the photo below .....??



Lastly - I am waiting for an RMA board to arrive, when it does I will use it as my test board, so it might be a few weeks before I get started. I will look out the stuff I need first and run it by you before I order !

[mhw100]

Setting in BIOS = 1.5375

Before After
idle 1.47 1.53
load 1.45 1.54

Looks great so far!! Should I take some graphite off since it is overvolting on load by a little or is this ok?

Thanks a million for this easy 2 second mod!!!!

[Micutzu]

gaz_az >

The 4 extra caps filter the +12V input of the PWM, for clean and stable input. I originally used 4x2200uF for this at first, but when i replaced the 1500uF's with 3300uF i also used these 1200uF's instead of the 2200uF's as they were scavenged from an old mobo and i didn't trust them too much. Their importance is minimal, the board already had good enough filtering and the PSU offers very clean voltage, maybe with a worse PSU they would become more important.

I have not connected the blue and red dots together, those are small SMD multilayer ceramic capacitors for extra filtering of the very high-frequency ripple.

You can basically connect a capacitor with the (+) at and red dot and the (-) at any other blue dot, it doesn't matter too much; i have used the thicker red dots for the (+) of the caps, and then with short but thick wires i connected the (-) to the closest blue dot. I haven't used all the dots, i just marked them so you could choose your best option for arranging the caps. The red+black wires are there for precise Vcore measurement .

Pay attention how you solder and arrange the caps, because you will need to use a backplate for the phasechange unit.

[gaz_az]

Micutzu - Cool got it !!

What is the capacitance value of the smd multilayer capacitors, or is it important ??

Did you just connect the Vcore wires to a molex plug for measuring ??


will drop you a p.m. when I am ordering !!!

[Micutzu]

I think they're all 100nF .

[JAG87]

Micutzu, thats dirty... must go shower now.

[gaz_az]

Micutzu - sent u a PM