nForce 790i SLI/Ultra chipset mobos - Reviews/OC/Guides

[saaya]

what different SOFTWARE tools report has nothing to do with what the real voltages are at.
they ALL read their values from the same single resistor/register! and this register/resistor is never spot on, never...
and even if its reliable at lets say 1.3v real measuring 1.31v, at 1.4v real it might measure 1.35v and be way off...

so all software tools use a different formula to calculate their final values, which they get from a probe which is never spot on to begin with...
So if 2 tools show 1.3v and one shows 1.4 it doesnt mean you are most likely at 1.3v...
you might as well be at 1.5v and they all misinterpret the values.
Plus if the bios is updated the probe might report different values than before the bios update...
the decent tools like cpuz and everest read the values directly from the probe, but most tools from asus gigabyte etc only read the bios values
and those values change depending on the algorythm the bios uses to interpret the values from the probes.

if you REALLY want to know what voltages your using then theres only one way... meassure them with a DMM

[3oh6]

if you REALLY want to know what voltages your using then theres only one way... meassure them with a DMM

and that is pretty much guess work with variables because your DMM could be off and your readings will vary greatly from ground point to ground point.

your vCORE is what you want to believe it is

[Demon^]

Hi all,

I need some serious help with my EVGA 790i and Q6600 G0, with my old board P5N32-E SLI(680) I could run the chip stable at 3.6Ghz all day long but now I have swapped to the 790i board I cant even get it stable at 3.6Ghz
I have tried some of the settings that people have been using in this thread but all to no avail.

System is as follows

Q6600 G0
EVGA 790i Ultra SLI
Corsair Twin3X 2GB DDR3 (2x1024MB) TWIN3X2048-1333C9 DHX 1333MHz
Etasis 750W PSU
CPU and Gfx on water, cooled with a Thermochill 120.3

If anyone can help me out here I would be hugely grateful

Cheers all

Demon^

[hifiking]

Some new info on my freezing problem. I can run ortos/occt 3dmark etc as much as I want. CoD4 lock up in minutes.

So I tried 3dmark in full resolution, 1920x1200. Lockups in minutes, tried a few times now. So, is there a bandwith problem here? I can't tuch the pcie frequncy even 1 mhz, I get post code 11 if i trie, any value. I'm going to play cod4 at a lower resoluton now to see if there is anything to it.

EDIT: cod 4 at 1280x800 crashes in 1 min, so it's a dead end.

[Coupa]

Some new info on my freezing problem. I can run ortos/occt 3dmark etc as much as I want. CoD4 lock up in minutes.

So I tried 3dmark in full resolution, 1920x1200. Lockups in minutes, tried a few times now. So, is there a bandwith problem here? I can't tuch the pcie frequncy even 1 mhz, I get post code 11 if i trie, any value. I'm going to play cod4 at a lower resoluton now to see if there is anything to it.

I can play crysis @ medium quality @ 1024x768 to my hearts content. Move it to very high @ 1440x900 and ittl crash in about 20 minutes.

Cod4 @ highest quality @ 1680x1050 will crash in about 5 minutes. I think its definately a bandwith problem.

[Broly]

prime can DIAF for me because I'm having no problems gaming lots, but only in PRIME!

[Big Myke]

Prime is causing problems for me aswell.

Has anybody tested the NB GTL yet on the SIIE? Do we still go in the negatives if we own a wolfdale? Or does it not apply to the NB GTL.

[slim142]

Its in intervals (dunno of how much, ive never really been bothered to check).

And i have a Q9450, and ive tried literally every GTLref in the negatives from when it literally wont boot, to the minimum value closest to 0, and still i freeze 10 minutes in to Call of duty 4/Crysis.

Post all your voltages so we can help you and see whats not right. Probably too low or too high vtt?

was jk man

im going to update the first post of the 790 corruption thread now
http://www.xtremesystems.org/forums/...d.php?t=183119

BTW, I will do my best to appear on that first page

I got the version 2 board. I have been using cod4 as a stability test, and it freez in max 10 min regardless of gtrlf or vtt. So I continued to other tests. OCCT at least 20 min stable, stopped it myself
3DM06 gfx tests 20 min stable, stopped it myself
memtest86, 20 min stable, stopped it myself

The only way I can freez this board is CoD4, or hours of occt/orthos might do it, but as long as cod4 freez I know there is a problem. I am gonna try some OC now.

My 1600 Kingston memory does 2000 at 8-8-8-20-1T and 2.1v, but performance is best at 1600 7-6-6-12-1T

Ok, where did this version 2 came from?
Try to use 1.20vtt and 1.30vtt to see if that solves anything. Since you have an E8400, dont go higher than 1.38vcore and also make sure when you change your GTLREFS settings, you change them equally. For ex: if its REF0 -20mV and REF1 -10mV, change them to REF0 -30mV and REF1 to -20mV. I think 5010 mentioned this in a post.

Also, make sure you run memtest86 (from DOS) for your RAM, 2.1v is waaay too high for a 24/7. I wouldnt go higher than 1.95v and 2.00v if you have a nice heatspreader or fans on top of them.
You should also try to use 2T because many experience problems running at 1T.
Finally set your SPP to MCP frequency to auto and lock PCI-e frequencies to 100Mhz.

Hope that helps

[slim142]

Prime is causing problems for me aswell.

Has anybody tested the NB GTL yet on the SIIE? Do we still go in the negatives if we own a wolfdale? Or does it not apply to the NB GTL.

Another GTL to mess with? I didnt know there was such thing as NB GTL. Well, since we have 2 people so far who are stable and didnt mess with it, I wouldnt touch it at all.

[hifiking]

Thx slim142:

I'v been trough most vtt from 1.1 to 1.35 with gtlvrefs from 0 to -0.40. And then different values for ref 0 and ref 1. Best so far is -0.1 and -0.5 with vtt at 1.35

Using the mem at 1.95 for 24/7

Tried 8-8-8-25-2T and the freezing in cod4 wasn't changed.

And 3DMark vantage has been on loop for quite some time now, no problems at stock resolution. Is there a way to increase pcie stability?

[Big Myke]

Another GTL to mess with? I didnt know there was such thing as NB GTL. Well, since we have 2 people so far who are stable and didnt mess with it, I wouldnt touch it at all.

Yes it certainly exists.

NB GTL_REF Ratio - [Auto, Default, -126mV ~ +160mV] The NB (SPP) GTL reference voltage provides the same functionality as those for the CPU, the only difference being that the Northbridge uses only a single value. Tuning in the right GTL reference voltage (usually near the nominal 67% value) can sometimes lead to lower stable VDIMM requirements.

From Anantech

[Broly]

Yeah but that needs HIGHER values for quad 45nms. Not LOWER if you read.
I used this, it's useful sometimes ;P

[Coupa]

Post all your voltages so we can help you and see whats not right. Probably too low or too high vtt?

Tried between 1.20 and 1.36, all go and freeze mid game.
Although later today i might go start playing even more with them. Someone said lower VTT voltages for 45nm help so i might do the whole range of tests - every gtl ref voltage that will boot with every available VTT between that range (wow thats gunna take some time, but its really all i can do).

[slim142]

Thx slim142:

I'v been trough most vtt from 1.1 to 1.35 with gtlvrefs from 0 to -0.40. And then different values for ref 0 and ref 1. Best so far is -0.1 and -0.5 with vtt at 1.35

Using the mem at 1.95 for 24/7

Tried 8-8-8-25-2T and the freezing in cod4 wasn't changed.

And 3DMark vantage has been on loop for quite some time now, no problems at stock resolution. Is there a way to increase pcie stability?

Try leaving VTT at 1.20 and start lowering your GTLrefs. I see a lot of prople with VTT at 1.20. You can also check what are the voltages 5010 and tekjunkie are using to see if that helps you in something.

If you say you are still having trouble, then I recommend you run memtest86 in DOS and run 2 standard tests and 10 times test#5 to see if there is a problem with your RAM sticks, if there is, RMA them or try another kit?

Best estability for pcie is to leave them at 100.

EDIT: Hey hifiking, try this

QDR/RAM=(your RAM speed) Unlinked
PCIE1/PCIE2 = 100
SPP<-> MCP= Auto
vCORE= (your vcore)
Loadline =Enabled
vTT = 1.28
PLL=Auto
vDIMM=(Manually put the stock voltage and timings for your ram)
vSPP =1.46
vMCP = 1.55
CPU GTL_REF0 = -29mV (this is from a qx9650)
CPU GTL_REF1=-32mV (this is from a qx9650)

You might also want to try different RAM speeds like 1650 instead of 1600, or 1605, 1610 etc
Also, I would try increasing your vcore (REMEMBER, DONT go higher than 1.38-1.40 ok?)

Yes it certainly exists.

From Anantech

Didnt really notice that, but are you guys messing with it?

Tried between 1.20 and 1.36, all go and freeze mid game.
Although later today i might go start playing even more with them. Someone said lower VTT voltages for 45nm help so i might do the whole range of tests - every gtl ref voltage that will boot with every available VTT between that range (wow thats gunna take some time, but its really all i can do).

It certainly does help in many cases.

@ everybody

I recommend you check your RAM sticks with memtest86. I know most of you dont use floppies anymore but this utility is really helpfull so you dont run the risk to BSoD or corrupt anything on windows.

[tekjunkie]

I tried your settings as posted here, but there's one concern.

While cpuz reports 1.344v vcore, everest and Asus probe II both report 1.42v.

LLC enabled overvolts ~.05v on my board, and 1.42v is kinda the scary region for 45nm, (at least that's what I've read so far).

I'd be careful in that voltage region, I like to stay below 1.37 with LLC enabled which will give me a rock solid 3.8 which I'm happy with.

I usually go with the temps of my core. If I am atleast 30 below tjmax I am happy. I also don't go with the actual temps reported by any of the utilities either. Since there is no documentation for the tjmax on the desktop chips, I rely on the direct reading from the DTS.

[tekjunkie]

@ everybody

I recommend you check your RAM sticks with memtest86. I know most of you dont use floppies anymore but this utility is really helpfull so you dont run the risk to BSoD or corrupt anything on windows.

Memtest is the very first thing I do, when I start changing anything in BIOS.
I start with test #6. I have noticed that with low VTT or low SPP voltages, memtest just freezes in test #6, with low vDIMM or higher VTT than required Test #6 shows errors.If I get atleast 3 good passes for test #6, I run Memtest for 3 full passes.

Also, I am getting my RMAed OCZ EB 4Gb sticks tomorrow. Will keep you guys posted.

[OBR]

Voltages at 450MHz + Xeon 3350 (Q9450)
I need 2.0V DDR3 because unstable 3D games ... now is everything fine (using Ultra SLI)



Is 24 hour Prime95 Blend stable, 10+ hours OCCT MEM + CPU test ...
Temp of CPU with watercool in OCCT 55C ...



Here is Xeon + memory settings: 20-7-7-7 P1,P2 Enabled ...

PS. Something new in 0603 BIOS?

[Coupa]

OBR - Tried your settings with my Q9450, freeze in 2 minutes on Cod4.

Im using BIOS 0402 which may be why, i dont have P1 or P2.

What BIOS are you running?

[tekjunkie]

PS. Something new in 0603 BIOS?

What BIOS version are you on??. By the looks of it, it's below 0603. I say this
as I noticed something different in 0603. As soon as you go above 1.30V VTT, the color changes to yellow.Something to take note of. I guess the engineers @nvidia or asus, determined that anything above 1.3 VTT is stepping into danger area given that the people @ AT blew their QX9650 @1.36VTT.

When I was running 0601, I had the GTLs on Auto and prime errored out on 1 my cores consistently @ a given point of time. But, with 0603, prime errored out immediately with GTL on Auto, with the same exact BIOS settings. Not that means anything, as there are two many parameters here to come to a conclusion, but I just want to throw it out there. It could very well be the different BIOS versions tune the GTLs differently on AUTO.

Also, 0603 gives the same memory bandwidth as what you have posted, but without P1/P2. The article @ AT said that they keep P1/P2 on Auto below 1900 to be stable. So, I guess P1/P2 does do something detremental to stability, albeit giving more memory performance.Also, Ryder and Tony of OCZ recommend P1/P2 on Auto if running below 1900.

[hifiking]

EDIT: Hey hifiking, try this

QDR/RAM=(your RAM speed) Unlinked
PCIE1/PCIE2 = 100
SPP<-> MCP= Auto
vCORE= (your vcore)
Loadline =Enabled
vTT = 1.28
PLL=Auto
vDIMM=(Manually put the stock voltage and timings for your ram)
vSPP =1.46
vMCP = 1.55
CPU GTL_REF0 = -29mV (this is from a qx9650)
CPU GTL_REF1=-32mV (this is from a qx9650)

With the evga this is the closest I get, and it still freez. I'm going to try memtest again later today when I am leaving.

vtt 1.25
vspp 1.45
vmcp 1.55
gtl0 -30
gtl1 -30

As you can se it increases by 5.

[saaya]

and that is pretty much guess work with variables because your DMM could be off and your readings will vary greatly from ground point to ground point.

your vCORE is what you want to believe it is

Well, ive never seen a DMM beeing really off... whats the highest youve ever seen a DMM to be off? 0.02?

And yeah, i also had trouble with different ground spots in the past... but you just gotta check all the ground spots against each other to find a propper one that always worked for me at least...

so do all you guys with 3d instability and pciE ocing problems use the S2E?
Iirc the asus 790 board has a different pciE clock signal than the reference board wich made pciE ocing very unstable or impossible.

Could it be that bad that it actually crashes at stock speed when the pciE bus is really stressed with high bw? hmmm

[Coupa]

Well, ive never seen a DMM beeing really off... whats the highest youve ever seen a DMM to be off? 0.02?

And yeah, i also had trouble with different ground spots in the past... but you just gotta check all the ground spots against each other to find a propper one that always worked for me at least...

so do all you guys with 3d instability and pciE ocing problems use the S2E?
Iirc the asus 790 board has a different pciE clock signal than the reference board wich made pciE ocing very unstable or impossible.

Could it be that bad that it actually crashes at stock speed when the pciE bus is really stressed with high bw? hmmm

Evga forums are full of freezing in 3d games too arnt they?

But the PCIe bus being the problem does make alot of sense, in that i havnt frozen when ive just been using CPU, RAM and HDDs, but as soon as you start using the Graphics cards, in the PCIe slots, then you start running into problems.

Id plug in a PCI graphics card and have some fun on...erm...yeah. But the problem of course is that i can run some of the lower powered games for as long as i want without crashing. Is there a way to just stress the PCIe slots?

[OBR]

Coupa - i am on 0601, and no freezing in games ... i played this weekend whole Rainbow Six 2 without any crash ... game completed successfully with no problem.

Most important thing in stability of 3D Games is timing and voltage of MEMs ... my Corsairs are crap and it is very hard to do them stable ... with another TeamGroup modules i am happier.

[hifiking]

My memory passes memtest fine, 5 passes and test 5 a few more times. No errors.

Two times on fresh installs the corruption occured when using windows update. So now I'm going to reinstall, enable sli but no updates. This is a desperate test, but I feel I have to test everything.

Singel card, no updates = fine, for a limited test period
Singel card, updates = corruption
Sli, updates = coruption | but singel after is good
Sli, no updates = testing soon

[Broly]

I hate my patriot ram.
Atleast corsair's dominators were overclocked from JEDEC 1333 to 1600 7-7-7-20.
Patriot overclocked 1066 jedec to 1600 7-7-7-20. Absolute BS