GTX 680 PCIe 3.0 & Rampage Mobo

[Arne_Saknussemm]

How many of you with Rampage 4 Extremes have had no problems running your GTX 680 at PCIe 3.0 via the "hack"

Negative experience is often up on the forum but not so much when something works so I thought it would make an interesting stat. Plus I´ve just ordered a DCII top and wondered what the chances were....:)


EDIT:

Manual hack: http://www.evga.com/forums/tm.aspx?&m=1607292&mpage=1

Nvidia force enable: http://nvidia.custhelp.com/app/answers/detail/a_id/3135/session/L3RpbWUvMTM0MDIyMzU2OC9zaWQvaDEzbE45...

EDIT 2: One person has found that running two cards initially broke 3.0 but changing the order of the cards in the Motherboard slots got it back up and running. Suspicion is that some GPU chips may better handle the extra load that the first card has to bear in the relationship with the CPU.....

[Nodens]

If you are getting crashes with 3.0 on I'm afraid I have an answer that you won't like. There's nothing you can do about it. Not all SB-E chips are GEN3 capable unfortunately. Replacing the CPU could make it work but it is still a lottery as the new chip could also be unable to do it. I would take the free C2 upgrade if I were you anyhow. Also have in mind that the difference in performance with existing cards is largely unnoticeable..

[Arne_Saknussemm]

If you can do it, yeah, get another CPU....that will be the limiting factor if it's the 3.0 that is causing crashes. Have you had no problems with your Vengeance RAM before this?

[Lettuce]

When I tried overclocking the ram about 9 months ago I had some issues, but nothing since then. The system has been completely stable until 3.0. If I hibernate and don't shut down for 3 or so weeks it gets a bit unstable, so generally really good.

Thanks for the advice, I'm pretty apprehensive about moving to the c2 stepping, I'll never, ever use VT-d and I had read of some people having higher temps and worse overclocks with c2.
I'll try some other things then I guess, like enabling 3.0 on one card only and moving the 2nd card to a different slot etc to check where the fault is.

I'm also thinking of disabling power saving on the PCIE lanes.

Any helpful settings in bios I should be aware of?

[Arne_Saknussemm]

It is a bit of a lottery, as Nodens said, with this... Vengeance RAM has a terrible record on X79 CPUs and makes me nervous every time I see it. But!...If your CPU is running it fine then maybe, since 3.0 over 2.0 makes very little difference, you could stick with what you have. I would hate for you to swap CPU for one with a weaker IMC and have trouble with your RAM.

To eliminate the Vengeance from the equation could you beg/borrow a stick of Gskill z series to test 3.0 with that? or try with just one stick of RAM in and see if 3.0 will run on one card and then two...etc.

[Lettuce]

Sure, I'll give that a try first.

Thanks for the idea.

[Nodens]

Don't be afraid of C2. Unless your current CPU is a really good specimen, then moving to C2 is good. I seriously doubt Intel made a new stepping with just a VT-d fix (which you may actually be tempted to use with Win8's Hyper-V if you ever want to run a VM for anything). Usually each stepping revision includes several fixes which are undisclosed to the public for obvious reasons. You know your chip can't handle GEN3 for sure, so unless it overclocks better than the majority of C1 chips, go for the upgrade. So far the majority of C2s I've seen overlcock better than C1s.

EDIT: Since his system is perfectly stable at GEN2 I doubt the ram has anything to do with it. I just noticed on his setup that he's using a lot of PCIe cards. SLI+Revodrive+Highpoint controller. The use of lots of cards is far more suspicious than anything else. It could be that his CPU would be able to do it with just 2 cards in SLI.

[Lettuce]

I tried with different ram and I still have issues at 3.0 speeds. Have not tried 3.0 on one card only yet.




Yeah, I've played around with virtual machines, and I use one pretty frequently, but I'll never use VT-d. I did have a read through the 90 or so problems with this processor, and VT-d was the only one fixed by this stepping, since it was an advertised feature. All the other issues, including those relating to PCI 3.0 compatibility are not addressed.

Overclock better than c1 hey?... That is quite convincing, OK I'll keep it in mind, I am apprehensive about the memory controller being damaged by lower voltages than c1, I'm not approaching even near 1.2v though so it won't be an issue.


I have tried running with only 2 680s in sli 16x 3.0, and a revo drive at 4x 2.0 and I still have stability issues, I don't really want to pull the revo drive as it's my system drive, plus where I have games installed.

[Arne_Saknussemm]

Lettuce wrote:
I tried with different ram and I still have issues at 3.0 speeds. Have not tried 3.0 on one card only yet.

Interesting, so seems to be the CPU....you're not overclocked with a BCLK that's not 100 or anything like that are you?

C1 vs C2 I don't think anybody could say which is a better overclocker. the variation between CPUs is so great that it would hide the variation between steppings unless you sampled thousands of CPUs...I doubt anyone has done that...

[Lettuce]

Turbo multi is x40, non turbo is standard, BCLK is 100.

GPUs are standard clocks for Direct CUII Top 680.

I have the additional PCIe power plugged into the board, is there a way to adjust it?

[Nodens]

Lettuce wrote:

Yeah, I've played around with virtual machines, and I use one pretty frequently, but I'll never use VT-d.

Why wouldn't you use VT-d? It would allow you to run HyperV on Win8 and keep your performance to the maximum (including graphics for games) and you could load VMs natively on demand as well (I am using it that way myself-I used to use VMWare on my development rig before I got this new HyperV setup--I still use VMWare for some things though).

I did have a read through the 90 or so problems with this processor, and VT-d was the only one fixed by this stepping, since it was an advertised feature. All the other issues, including those relating to PCI 3.0 compatibility are not addressed.

Logic or manufacturing fixes etc are not disclosed unless they go public by a third party. The list you were reading was probably the disclosed stuff. (Do you have a url btw I'd like to see it:)). GEN3 compatibility won't be addressed with a stepping revision, you can be sure of that. Wait for IB-E for that.

Arne Saknussemm wrote:

C1 vs C2 I don't think anybody could say which is a better overclocker. the variation between CPUs is so great that it would hide the variation between steppings unless you sampled thousands of CPUs...I doubt anyone has done that...

I largely agree to this mate. My personal observation is that the C2 chips need less overall voltage but the actual overclockability does not depend only on that. I've seen about 20 samples which are very few indeed but all of them needed less voltage for mediocre overclocks (I've replaced most of these chips on friends' systems and overclocked them). There is the rumour about the C2 being more sensitive to VTT/VCSSA, degrading faster with high voltages, that I can not vefiry as I've never used anything more than 1.25V and I've never had a C2 chip degraded. I'll not risk going above 1.2-1.25 on the C2 for 24/7 anyhow 🙂

Lettuce wrote:

I have the additional PCIe power plugged into the board, is there a way to adjust it?

There is nothing you can do really 😞