Who here has reference design 780i mobo (evga or xfx) and QX9650?

[terentenet]

I am having a hard time getting my QX9650 to run stable at 4.0GHz on EVGA 780i. I've tried various settings, it won't stabilize until I give 1.513 Vcore, which seems a little high compared to other's voltages. I've seen many examples of QX9650 running 4GHz with way less, like 1.475v, some even less.

Here are my system specs:

QX9650 Retail
CPU cooled with Vapochill LS stock unit.
EVGA 780i P03 BIOS
8Gb 4x2048 A-Data Vitesta Extreme memory 5-5-5-12 at 800MHz
2 x EVGA 8800GTX KO ACS3 - watercooled
2 x Raptor X 150Gb RAID 0
2 x WDC RE2 500Gb RAID 0
2 x Plextor PX-760SA DVD-RW
Tagan 1100W PSU

Am I doing something wrong? Here are my current voltages and settings:
Vcore: 1.513v
FSB: 1.3v
Memory: 2v
SPP: 1.35v
MCP: 1.60v
GTLVREF Lane 1 & 4 : 00mv
GTLVREF Lane 2 & 3 : 05mv

All Spread spectrums are Disabled, All Processor features are also Disabled.
FSB: 1600MHz (400); Multi: 10. Memory runs Linked & Synced, 800MHz 5-5-5-12 2T.
Memory is not the issue, it passed 24hrs of Memtest.
FSB and MCP voltages are not the issue as well, I've dropped the FSB down to 1.1v and MCP at 1.5v and it's still stable, as long as I feed 1.513v to the CPU.
This weekend I will attempt a Vdroop mod on the motherboard. I have 1.513v set in BIOS, 1.488v at idle in Windows and 1.432v under load with Prime95 25.6 Small FFT's.

 

[JAG87]

*shy raise of hand*


I think your NB is not as stable as you think. You are running 8GB, and 4 DIMMS, it doesn't get much more stressful than that for the memory controller.

Another problem could be the GTLRef voltages at 400 FSB. You need a bit more. My 780i gives +65mv automatically to the second GTLRef in the bios (which is the data bus of Core 0 and Core 1), when I set 400 FSB.



This is what I would do:

put FSB back to stock, and all voltages back to stock. run the ram unlinked.

try 333x12 with 1.4v, all other voltages set to auto. if it's not stable, increase the vcore until it is. if you need 1.5125 in bios to be stable at 333x12, sorry to say but you have a bad chip.

if you find that its stable at a much lower vcore (like 1.44 or 1.45), then that proves that the instability was caused by the motherboard/NB/memory.

at this point, it is in your best interest to find what voltage you need to increase to find stability at 400 FSB. Could be that you need 1.4v vFSB, or 1.5v on the SPP.

You certainly don't need any voltage on the MCP, MCP Aux , and SPP>MCP (the last 3 voltages). That 1.60v on the MCP is useless, its just cooking it.



In conclusion, I wouldn't do the vdroop mod. It is not healthy for the motherboard and it provides no improvement in stability, because the mosfets will always have the idle>load dip, and the load>idle peak. FYI, I am running 400x10 with 1.45v set in the bios, 1.42v at idle, 1.35v with small FFT Prime95, 100% stable.

All motherboard voltages are set to auto, which at 400 FSB they become 1.3 vFSB, 1.4 SPP, and +65mv on the second GTLRef.

 

[aigomorla]

Welcome to the proud association of "Monster Rig Club"

JAG and i are members, Mark is another member, abit he's lacking on the GPU department.

Anyhow sorry for the OT, just happy to see were not the only one with bleeding edge machines..

But i also think your NB is lacking and the added ram is holding your back. Its harder to overclock a system with 4 dimm's vs 2 dimms.

I dont have a retail QX9650, but i understand quads enough to know whats wrong.


Try removing 2 ram sticks first and see if thats holding you back. Also im assuming your clocking the ram independantly? i would try to drop the speed on the ram to lowest possible and see if you can hold the cpu side stable.

Also since your on sub ambient, you shouldnt require that much voltage to get that OC. Generally when your below 0, you require about 75-80% of the normal voltage required by people on non sub ambient cooling.

One last question, what is your vapo tuned at? Is it a stock vapo? Or did you get a custom moder to refill it with different gas?

How well is it tuned? as in how much heat can it handle? Because Quads + Phase = tricky unless its tuned for a quad to begin with. You'll burn your vapo unit out in a matter of months on constant use. Ask Yoxxy.

 

[terentenet]

Jag, I just tried what you said. I dropped the Vcore to 1.4v, let the memory have 2v, set all others on Auto, GTLVREF on Auto.
I set the FSB at 1333, RAM at 667 & loose timings 6-6-6-18 2T.
Multi at 12 for 4GHz and I started testing.
1.4v would not load Windows
1.45v into windows, locked as soon as I started Prime95. Instant lock.
1.475v still, instant lock.
1.493v runs Prime95 for a couple of seconds, then Core#1 fails.
1.5v makes a first pass (1 - 1.5 minutes) then one core fails.
So, it still needs 1.513v for stable operation at 4GHz, regardless of loose memory timings and low FSB.
I also tried 1.5v on the Vcore and manually upped the other voltages. It still fails, just like I haven't touched the voltages.
Does this mean I have a bad chip? Well, not bad, as it will do 4GHz, but with more voltage than other CPU's need.

Originally posted by: aigomorla
Welcome to the proud association of "Monster Rig Club"
JAG and i are members, Mark is another member, abit he's lacking on the GPU department.
Anyhow sorry for the OT, just happy to see were not the only one with bleeding edge machines..
But i also think your NB is lacking and the added ram is holding your back. Its harder to overclock a system with 4 dimm's vs 2 dimms.
I dont have a retail QX9650, but i understand quads enough to know whats wrong.
Try removing 2 ram sticks first and see if thats holding you back. Also im assuming your clocking the ram independantly? i would try to drop the speed on the ram to lowest possible and see if you can hold the cpu side stable.
Also since your on sub ambient, you shouldnt require that much voltage to get that OC. Generally when your below 0, you require about 75-80% of the normal voltage required by people on non sub ambient cooling.
One last question, what is your vapo tuned at? Is it a stock vapo? Or did you get a custom moder to refill it with different gas?
How well is it tuned? as in how much heat can it handle? Because Quads + Phase = tricky unless its tuned for a quad to begin with. You'll burn your vapo unit out in a matter of months on constant use. Ask Yoxxy.

Aigomorla, my Vapo LS is stock. I have it for about a year now and it's still holding well. All this time I used it to cool the ex processor, QX6700 Engineering Sample @ 3.6GHz with 1.48Vcore. From the readings on the display, the QX9650 with 1.513Vcore runs cooler than the old processor.

I will try your suggestion as well. Perhaps the 780i is picky about it's memory. I will pull out 2 sticks, leave all voltages as they are right now, I will only lower the Vcore. Will start from 1.45v and see if I can get it stable at lower Vcore.
I will also loosen the timings on the remaining RAM and lower their frequency.

 

[aigomorla]

Originally posted by: terentenet
Jag, I just tried what you said. I dropped the Vcore to 1.4v, let the memory have 2v, set all others on Auto, GTLVREF on Auto.
I set the FSB at 1333, RAM at 667 & loose timings 6-6-6-18 2T.
Multi at 12 for 4GHz and I started testing.
1.4v would not load Windows
1.45v into windows, locked as soon as I started Prime95. Instant lock.
1.475v still, instant lock.
1.493v runs Prime95 for a couple of seconds, then Core#1 fails.
1.5v makes a first pass (1 - 1.5 minutes) then one core fails.
So, it still needs 1.513v for stable operation at 4GHz, regardless of loose memory timings and low FSB.
I also tried 1.5v on the Vcore and manually upped the other voltages. It still fails, just like I haven't touched the voltages.
Does this mean I have a bad chip? Well, not bad, as it will do 4GHz, but with more voltage than other CPU's need.

Load the latest edition cpu-z

Then set your processor on the setting where it would hold prime for 5 min.

Tell me what it says under voltage when its at load.

Maybe your board isnt liking the vdroop. Also tell me you took your stock nb sink off and replaced the tim underneither.

Also can you download ntune and open up all the monitoring software and let me see wat your settings are.

 

[Pantlegz]

terentenet, I'm having similar issues with my evga 780i & E8400 OC. I just can't get the damn thing stable below 1.45v(1.43 vdroop) @ 3.8 even then it's onlty good for 4-5 hours of prime and will get an occasional BSOD, RAM passed 24 hour memtest as well. And on air it gets a little toasty, 60c is a little warmer than I'm comfterable with.

I have a felling it's the NB or memory controller on some of the evga boards but really don't have any proof to back this theory up.

 

[terentenet]

Per aigomorla's request, here are 2 snapshots I took. One at idle, one during Prime95 Small FFT test.

Idle shot: http://img352.imageshack.us/img352/7162/idleiv7.jpg
Load: http://img352.imageshack.us/img352/3696/loadxv4.jpg

I am getting ready to pull the 2 sticks out and test for OC stability at lower Vcore.
Yes, I took off the heatpipe thingy and replaced the stock goo with Arctic Silver Ceramique.

Pantlegz1, stay tuned as I will take out 2 sticks of RAM and check for stability. If these stability issues are RAM/Memory controller realted, I should be able to lower my Vcore for a stable overclock.

 

[terentenet]

Done. Pulled 2 sticks, same problem. 1.513v seems to be the only voltage that works for this CPU at 4GHz.
I will call some friends and perhaps I will get my hands on another QX9650, maybe I get lucky and I get a better sample....
Thanks all for your help. Greatly appreciated.

 

[aigomorla]

you definitely got a vdroop problem

1.513V bios, and loaded voltage is 1.424Vcore

Use the 680i pencil mod technique. Same thing on the 780i.

Try to lowe that vdroop so its within .01 As it is right now your near .1 vdroop!

With the vdroop pencil mod, you should be able to get it close, so you can set your bios at 1.43 and have it load @ 1.42Vcore and that should make you happy.


OT: You need to pencil mod it. I ran into this problem with quads a very long time ago on B3's. Needed 1.55Vcore for stable, however needed near 1.62 in bios. Had to pencil mod it and that fixed it.

Also make sure your mosfet sinks are securly attached. Make sure all contact points are contacted. A lose mosfet sink can wack out your overclocking on that board.

 

[terentenet]

As I have never done a pencil mod before, I need to ask... the main question; what pencil does someone use to make such mod?
Where do I get it from? Is it a regular pencil that you can buy in a library? I saw that the Vdroop mod for the EVGA 780i board is done with a 100kohm resistor. Does the pencil do the same thing? Is it safe and can I do it myself?
Damn... lots'a questions

 

[Pantlegz]

I've never done it either, I think any regular pencil should work. and using the 100Kohm resistor would have a similar effect.

 

[Rubycon]

Wow that's high VCORE on a Yorkie. They do like a lot of VCORE to get over the 4 gig hump unless you have a golden egg that does 10x400 at 1.2V. :Q

P5E WS Pro here - 1.4 VCORE idle, 1.4105 VCORE load. Yep it goes UP when loaded.

The pencil mod is the same thing as placing a high value resistor across the pads. The resistor will give stable long term results with minimal drift. The pencil mod is good for those that wish not to take a soldering iron to their motherboard.

EDIT: You have a vapochill - :Q How does it handle the full load? Linpack/SuperPrime will get the highest peak temps - many times 10 degrees Centigrade or even higher than Prime95 small FFT! :Q

 

[aigomorla]

Originally posted by: Rubycon

EDIT: You have a vapochill - :Q How does it handle the full load? Linpack/SuperPrime will get the highest peak temps - many times 10 degrees Centigrade or even higher than Prime95 small FFT! :Q

i asked first!!!!

Pencil mod is easy, and fairly painless. If you take it slowly 1 stroke at a time, it can be safe.

the best pencil is a 2B pencil to use. Its fairly easy, and follow the guides.

Basically after the first 3 strokes, you load up prime to see how close to the vcore you actually are. If your still a bit far away do 1 stroke and then aproximate how many more you need.

If you overstroked, your loaded vcore will be higher then whats reported in bios. So start at a smaller overclock and work from there.

 

[aigomorla]

Originally posted by: Pantlegz1
terentenet, I'm having similar issues with my evga 780i & E8400 OC. I just can't get the damn thing stable below 1.45v(1.43 vdroop) @ 3.8 even then it's onlty good for 4-5 hours of prime and will get an occasional BSOD, RAM passed 24 hour memtest as well. And on air it gets a little toasty, 60c is a little warmer than I'm comfterable with.

I have a felling it's the NB or memory controller on some of the evga boards but really don't have any proof to back this theory up.

your problem sounds like a couple. You dont have a vdroop problem. .02 isnt bad. The OP had near .1!!!!!

first one im guessing is bad mosfet contact.

Second one is correct, a bad northbridge cooling contact.

You need to pull the board out, and remove the stock cooling. Clean that thing up with some iso alcohol. And then reapply some fresh tim.

Then on the mosfet, you need to clean that gooey nasty kinda feels like gum pad. And apply some fresh tim on top of the black mosfets and then reapply the mosfet sink.

As i said an overheating mosfet can cause your overclock to become serious WACK.


Try doing those two and then come back, or pm me.


EDIT: dude ive been noticing some very disturbing 780i post, and the vdroop on the quad isnt making me happy. I have a 780i but its currently undergoing some custom work on it. Waterblocks to ensure that thing never see's stress.

Maybe i'll go this route as my main rig instead of the 780i route.
http://i125.photobucket.com/al...aigomorla/IMG_0882.jpg

But i wanted to play with a E8500 on that setup. :\

 

[JAG87]

First of all @ aigomorla

don't give advice you are not entitled to give, at least yet. just like you say, don't talk if you don't own, right?

doing the vdroop mod is a bad idea, and fixing up the TIM on the motherboard heatsinks wont help him either since we know its the cpu.

obviously fixing up the TIM will be a positive thing, if you remember I even specifically suggested you do that on your 780i like I did on mine, but unfortunately it wont help him get 4 GHz with lower vcore.



@ terente

don't do the vdroop mod. people do not understand that vdroop is there for a reason, vdroop is there to fix what the motherboard does wrong. mosfets have oscillation (idle>load and load>idle), and vdroop is there to compensate.

780i has very poor power regulation, and therefore it needs a lot of vdroop to compensate, while higher quality boards like asus and gigabyte have very low vdroop since their power regulation is 100x times better.

read this article, pictures are 1000 times better than how I can explain in words: http://www.anandtech.com/cpuch...howdoc.aspx?i=3184&p=5

you gotta face it, you have a cpu that needs 1.424v to be stable at 4 GHz, and that is pretty lousy for a yorkfield. I have the same vdroop as you, my voltage is 1.45 in bios, 1.416 in windows, and 1.352 under Small FFT load, and I am perfectly stable.

time to sell the chip and buy a new one my friend. don't worry, it happens to the best of us

 

[terentenet]

Jag, I allready spoke with some friends about testing a new CPU. I will get to the bottom of this somehow. I will also test my skills on a pencil mod. I will test with my old combo; 680i and qx6700. That will help me to better understand the process. Then I can make my choice, regarding wether or not to mod the 780i.

 

[terentenet]

Okay, I've done the Vdroop pencil mod today. Voltage is stable, I set 1.438v in BIOS. I have 1.424v in Windows idle and 1.432v under load.
System seems stable so far, temperatures are the same. The good side is my peace of mind, knowing that I'm in the green zone with the voltages.
I will still try to get ahold of a better CPU, perhaps I can have under 1.4v for 4GHz.

EDIT: Priming with 1.431v in BIOS, 1.424v under load... If it passes 3-4hours, I'm happy with that and I will try a step lower. It's better with the Vdroop mod so far, the load voltage seems stable. Before, it used to fluctuate between 1.416v and 1.424v.

 

[Apocalypse23]

Whoa, I was planning to build a 3 way SLI system in the near future myself but was looking to get a Qx9650 and a eVga 780i board and overclocking it to atleast 4.0 GHz like I did to my E8400. With my Asus P5K-e/WIFI I initially had vdroop issues, well I wouldn't say big time issues, it was just the stock BIOS was reporting me a .10 voltage increment in windows CPU-Z and all other monitoring programs. For example I'd set my CPU voltage from the mainboard to 1.43 or something and it would show as 1.33 in CPU-Z or 1.32 V. I "fixed" this if you may by downloading Asus's latest BIOS that shrinked this Vdroop to .01 V which was better. Like Jag said, Vdroop is a good thing, but as long as your temps are okay don't worry.

Most probably I am wondering if there is a 780i BIOS UPGRADE available that resolves this issue?

And also, would a 780i board from ASUS or XFX be better?

For example: http://www.ncix.com/products/i...eluxe&manufacture=ASUS
that's the ASUS ASUS P5N-T Deluxe 780I SLI . I am considering getting this one or looking into it if it is a more suitable 780i system than the evga or xfx...mainly since ASUS has a good BIOS upgrade rate.

Anyhow try an upgrade for the BIOS perhaps and see if it resolves the vdroop or minimizes it. I need to hear abuot others with comments about the ASUS 780i board vs the XFX and evga though.

PS - I personally would never bother with any physical alteration of the motherboard , i.e using pencil mods to get by, look to BIOS upgrades as the only available solution.

Thanks!

 

[Apocalypse23]

yea! i love the thought of getting a 3 way SLI system, probably a future 9800 GX or Ultra, but I would love to see if ATI has it's 4 way crossfire boards out yet.

 

[terentenet]

As far as I know, the ASUS Striker Extreme II doesn't work with QX9650. Yet. I have a friend that has tried that combo and it didn't work out for him.
Between ASUS and EVGA I'd get... EVGA. I don't like ASUS's customer support. I owned a few ASUS motherboards but I was not impressed by any of them (performance wise).
I don't see the pencil mod as a physical alteration to the board. It can be undone with a tissue and rubbing alcohol.
Regarding the video card business, I am dissapointed by both ATI and nVidia. I hate the new route they've chosen, with dual GPU's on a single card or sandwitched. I want a new HIGH END. Bring out R700 and 9800GTX allready!!!

 

[krnmastersgt]

Between EVGA and XFX, the only difference would be.....none that I know of, both are reference models from nVidia so aside from different logos they are the exact same boards, same BIOS too. Also when and if I have problems with my 780i I just head over to the EVGA forums if I want to find a solution. Aside from warranties they offer theres no difference between XFX and EVGA. (I think I'm the only one on the AT forums with an XFX 780i )

 

[aigomorla]

EDIT:

okey after reading JAG's article.

He was worried about the voltage shoot when your processor comes off full load.

Basically its this, your voltage can spike around .07 according to the article after it pulls off full load.

Now i still think the pencil mod is safe because it required 1.513 voltage on bios for you to hold stable.

Now your pulling 1.43v in bios to hold that stable. So when it peaks out of full load you'll be peaking 1.5v for no more then a few sec, instead of sitting there @ idle all the time.

So yeah... i think your better off with the pencil mod, expecially if its allowing you to pull that voltage down lower.

Originally posted by: krnmastersgt
Between EVGA and XFX, the only difference would be.....none that I know of, both are reference models from nVidia so aside from different logos they are the exact same boards, same BIOS too. Also when and if I have problems with my 780i I just head over to the EVGA forums if I want to find a solution. Aside from warranties they offer theres no difference between XFX and EVGA. (I think I'm the only one on the AT forums with an XFX 780i )

Thats because most of us bought evga's when they were 680i. Then we stepped up, to a 780i. Also back then eVGA was faster with bios upgrades then every other company which was a big plus.

but the sorry %#$$@# foxconn board just cant hold out to a lot of stress.

 

[JAG87]

as I explained to you through PM aigomorla, its not the overshoot you have worry about, but the undershoot when going into load.



terente, you will not notice instability with prime95 and the vdroop mod. there is only 2 changes of state with prime95, when you start it, and when you end it. unfortunately programs don't work that way, load and idle cycles fluctuate all the time. games are a huge example.

you will see that although you are 100% prime95 stable, and your PC seems rock solid in all applications, you will get a BSOD once in a blue moon while doing some random thing. and you are going to be like hmm, what the hell was that.

the vdroop mod is not good, the only way it would be good is if your PC is either constantly idle, or constantly under load, because then the voltage is stabilized. but normal usage is totally opposite of that. beware.

 

[Tempered81]

I've also heard a lot of people saying vdroop is part of the design and not a bad thing.

 

[terentenet]

Originally posted by: JAG87
as I explained to you through PM aigomorla, its not the overshoot you have worry about, but the undershoot when going into load.
terente, you will not notice instability with prime95 and the vdroop mod. there is only 2 changes of state with prime95, when you start it, and when you end it. unfortunately programs don't work that way, load and idle cycles fluctuate all the time. games are a huge example.
you will see that although you are 100% prime95 stable, and your PC seems rock solid in all applications, you will get a BSOD once in a blue moon while doing some random thing. and you are going to be like hmm, what the hell was that.
the vdroop mod is not good, the only way it would be good is if your PC is either constantly idle, or constantly under load, because then the voltage is stabilized. but normal usage is totally opposite of that. beware.

Jag, it's good to know that. I will keep the Vdroop mod, it keeps my sanity by knowing I don't push 1.513v in the CPU. Well, it's the same voltage actually, it's just that what I set in BIOS I see in cpu-z.
If I run into stability problems, I will undo the mod. It's simple, I can just smear the pencil off the resistor.

Aigomorla, I'm in the process of upgrading my computer... well, not the computer, but the case actually and some of the cooling parts.
I know you're into water cooling, so please, I need some advice. I currently cool 2 8800GTX cards with EK full blocks and an underpowered radiator; BIX 240. Tubing is Tygon R 3603 and the pumps are 2 Laing Ultras with Dual Watercool top. To be able to cool, I am using 2 Akasa fans rated at 105cfm. Quite noisy. So noisy in fact that I can hear the computer running from outside the room, with the door closed

Now; I want to change all that. Because there's no place in one case (be it Lian Li V2000 even) for all the components, my next build will have 2 cases.
Not Lian Li V2000, but Aplus Black Pearl cases (basically a V2000 with same design, but not all meshed). I see now that these cases appear as being made by Tagan.
My plan is to nail the 2 cases together and obtain something bigger than a Mountain Mods case.
The case on the left side will hold all the mechanical parts and the cooling parts (HDD's, water pumps, DVD-RW), while the "system" case will hold only the system and the Vapochill LS unit (it has to be near the processor, otherwise the cooling head will not reach it). That should help by separating some of the main heat builders from the system.
I live in Romania and Thermochill radiators are nowhere to be found here. So I chose 2 Black Ice GTX 360's to do the job of cooling the 2 8800GTX's. I went for 2 360 radiators because I'm also futureproofing, for when I will run Tri or Quad SLI.

My question is: can you give me a fair comparison between the Black Ice GTX 360 radiator and a Thermochill PA120.3? When used with low CFM fans?
Are these fans any good? I like the design and they say these have big static pressure, which should make them perfectly suitable for water cooling.
Also, I will have to make the build using smaller diameter tubing. Will that be too restrictive? Or the 2 Laing Ultra's in series will push the water through ok?

When all is done, new system will be:
QX9650 - frozen by Vapochill LS (allready have it)
EVGA 780i (have it)
8Gb A-Data DDR2 memory (have it)
2 x WD Raptor X 150Gb RAID 0 for the system only (have them)
4 x Seagate Barracuda 7200.11 1Tb RAID 0 for storage (4Tb.... niiiice) (bought them allready)
1 x Plextor DVD-RW IDE drive (not decided on the model yet)
2 x EVGA 8800GTX KO ACS3 - properly watercooled by 2 x 360 radiators (waiting for the high-end 9 series to appear and Tri or Quad SLI those) (ordered the rads, waiting)
Tagan 1100W PSU (I might have a 2nd low power PSU (3-400W) strapped with the Tagan, to power the HDD's, DVD-RW, water pumps and radiator fans) (not decided what the 2nd PSU will be)
2 Aplus Black Pearl cases (bought them)
Creative X-Fi Elite Pro + Z5500 speakers + Sennheiser HD595 (have them)
Dell 3007WFP-HC (have it)

I have most of the components ready. As soon as the rads arrive, I will get my buddy to start on the build - nail the cases and figure a way to mount the rads inside (I'm not doing that, too much skill and patience required).