i7 920 OC: VTT higher than VCORE OK?

[TJCS]

My question is: Will my current settings lead to something catastrophic (ie. dead memory controller)?

Rig in my sig is in question: I managed to run(stable) 4Ghz with:

• 21 x 192
• VCore: 1.275 (Without VDroop)
• VTT: 1.41V
• PLL: 1.86V
• VDimm: 1.65 (Stock)
• LinX: 10-Pass Load-Temps 63~65
• All other voltages are set at AUTO (Asrock OC Tuner Utility Reports standard voltages)

VTT Voltage options in bios are: 1.27V → 1.34V → 1.41V
(Nothing in between and system hard-locks at anything less than 1.41V VTT).

I know that you should keep VTT within .5V of VDimm and I have also heard some recommended VTT should be 0.05 lower than VCore?

I have tried a lot of VCore and VTT combinations(VCore up to 1.35 and VTT up to 1.48) but this by far seems to be the most stable.

 

[Shmee]

weird that you dont have any other VTT options. 1.41 seems a bit high, what is your board? There might be a bios update to fix it. If not, you might be sol. I would prefer to have VTT under 1.375v if possible. Is your chip a C0 or D0?

 

[TJCS]

weird that you dont have any other VTT options. 1.41 seems a bit high, what is your board? There might be a bios update to fix it. If not, you might be sol. I would prefer to have VTT under 1.375v if possible. Is your chip a C0 or D0?

The motherboard is an Asrock x58 Extreme running latest bios P2.90 and the i7-920 is a D0.

I will try to dig around some more but I don think there are other values in between. :\

 

[TJCS]

Will my settings be ok as long as I keep temps in check? Or will is there something else I need to be worried about with my current setup.

 

[bryanW1995]

1.41 is a bit high, but probably nothing to be concerned about as long as temps remain in check.

 

[TJCS]

I just tried lowering the VTT to 1.34V and was met with a BSOD 0x000009 within 5mins of LinX.

 

[Shmee]

hmm pitty. Have you tried relaxing memory settings?

Though honestly, I wouldn't worry about it too much. If your settings were stable and you were happy, seems like a good sweetspot. Though a bit high, and unneeded, its what you have to work with. Only thing that might help is raising vcore a bit. I do recall, AFAIK, 1.4 vcore is preferable to 1.4 VTT with your chip. High VTT will kill it quicker.

Not positive on this though.

 

[TJCS]

I tried upping the VCORE a few notches from 1.275V to 1.325V (while isolating the VTT at 1.34V) but was all met with 0x000009. And in regard to the timings, I tried loosen them to 888 but seems to result with same BSOD results. Increasing TRFC from 74 to 88 seems to notably reduce system hard-locking when pushing BCLK near the 4Ghz zone.

I forgot to mention that before attempting 4Ghz I was stable at 3.9Ghz (21x186, 1.275V, 1.27V VTT), and there seems to be a wall at 188 BCLK with VTT less than 1.41.

So far, with the settings in the OP, the results are:

• LinX 20x no errors
• 3DMark Vantage 5x (low-res graphic and cpu tests)
• Stable testing in games (L4D2, CS)
• 2hrs of prime95 no errors (currently ongoing)

The benchmark scores are higher (and more consistent) across the board with:

4.0Ghz 777-24-88-1T (1.27VCore, 1.41VTT)
than
3.9Ghz 777-18-59-1T (1.27Vcore, 1.27VTT)


I have only had the 4.0ghz settings for less than two days, but I am really looking forward to keeping them based on the results. Just hope the low vcore and high vtt doesn't pose any dangers.

 

[Spikesoldier]

i dont think a high vtt and low vcore should give you issues.

the deal with vtt and vdimm is: vdimm MUST be lower than VTT+.5v. otherwise you will get the fried memory controller.

i.e.

vtt 1.2 , vdimm 1.65 OK
vtt 1.3 , vdimm 1.50 OK
vtt 1.4 , vdimm 1.65 OK
vtt 1.2 , vdimm 1.75 BAD
vtt 1.4 , vdimm 1.75 OK

 

[rgallant]

don't have that board and after 2 years with the same bios setting I forget all the pro's and con's but remember 1.35v is the max[intel] for vtt with a good multimeter on test points - which you won't have either so play it safe,
for 4.0
-your problem is with

"All other voltages are set at AUTO "

I think on a updated BIOS.

just saw this
(seems to be set at standard using Asrock OC Tuner Utility)*************************throw every setting tweak out the window
I would NOT use ANY of that stuff --but that's ME

 

[TJCS]

I think i was unclear in my OP, but I meant to say that I used OC Tuner to check the voltages, but I do not use it to do any tweaks (But I agree with not using it to tweak settings etc. ^_^).

 

[spinejam]

This is what I've been using for over a year:



Uploaded with ImageShack.us

 

[TJCS]

@spinejam thank you for taking the time to post it, it is definitely a helpful reference especially how similar our setup is.

I think I might have caught a break/mistake T_T. My friend was helping me test the setup and it turns out he only tested VTT 1.34 at BCLK 196 earlier. He said he later changed BCLK to 192 and VTT 1.34V and benchmarks seems to be running stable atm.

 

[coffeejunkee]

I would advise to take a 2-step approach. First lower the cpu multiplier and see how much vtt you need for a certain bclk (at least 20 runs Linx stable preferably). Then increase the cpu multiplier and see how much vcore you need. It's a pity so few vtt options are available on your mobo though.

Personally I would step down a bit though, go bclk 190x20 for 3.8GHz and see if you remain stable with c-states enabled, they do wonders for idle usage.

 

[aigomorla]

920 is fine.. lol...

These ARENT glass slippers like the 32nm are... 45nm is solid... the first gen i7 were SOLID.

D0's are tanks... i have yet to pop one at 1.45Vcore 1.45VTT.

Of course i kept heat majorly in check tho.

i even know cases where guys run them at 1.5V.... key thing is keep your heat in check... the D0's are not glass slippers but at the same time, they would rank as space heaters compared to a SandyB.

 

[TJCS]

1.34V VTT seems to be stable after testing :thumbsup:

@ cofeejunkee

Before testing 4.0Ghz I have found the optimal VCore for 21x186 to be 1.275V on this MB. At 3.9, anything less dropped GFLOPS, too much made GFLOP jump around...weird...

Thanks for noting the c-states, I will run the machine for another day or two to verify stability and try out those options.

 

[TJCS]

920 is fine.. lol...

These ARENT glass slippers like the 32nm are... 45nm is solid... the first gen i7 were SOLID.

D0's are tanks... i have yet to pop one at 1.45Vcore 1.45VTT.

Of course i kept heat majorly in check tho.

i even know cases where guys run them at 1.5V.... key thing is keep your heat in check... the D0's are not glass slippers but at the same time, they would rank as space heaters compared to a SandyB.

Thanks for sharing those experiences, it's good to know that I am not in the red zone.