Vista 64 Locking up UNDER 4.0GHz

[Viscosity]

Howdy folks!

I've just embarked on my first custom PC building adventure. So far it has been a rewarding and wonderful experience (even despite the 20 or so hours I've put in trying to tweak settings so far).

Before I describe my problem, here are my specs:

eVGA 790i Ultra
C2D E8400
Zalman 9700LED CPU Cooler
eVGA GTX 280 SLI
Corsair 750TX PSU
2GB Patriot Viper DDR3 2000MHz (SLI ready, but I don't use EPP)
Vista 64 Ultimate

Okay.

When I put everything together and fired it up, POST went great, as well as the BIOS update (P06). Installation of Vista 64 went off without a hitch (about 20 min for a full install!).

After I installed Vista, I updated all the drivers I needed, and installed a few programs (Crysis, obviously). After I finished with that, I headed back to the BIOS to try out some CPU overclocking. I worked incrementally, eventually hitting an 1800 FSB (4050MHz). Everything booted fine.

Here is some background about running at 4GHz:

When I started to run torture tests, I realized that this CPU demands a REALLY high voltage in the BIOS over 4.0GHz. The memory ran stable in memtest, and I have it sitting at 2.0V. So I narrowed down the problem to VCORE, and I had to bump it up all the way to 1.49 in the BIOS in order to get Prime95 to run stable (it ran stable for 19 hours and then had an error -- if someone can tell me if that's acceptable or not, that would be great).

CPU-Z and CPUID Hardware monitor said the VCORE never went above 1.46V, so I understood this as the baseline (4.49V in the BIOS is meaningless if the CPU never goes above 1.46, right?) Anyway, it looks like my vdroop is about .3, and the core voltage reported hovered around 1.434 or 1.440 for the duration of the test (though I did notice it drop as far as 1.416 a couple times). Temps (using Core Temp) seemed acceptable -- 49C in idle, 68/70 in full load (never going above 70).

Now here is the problem:

I decided to start pulling the clock speed back a bit, because I didn't want the processor idling at 1.464V (if this is not a problem, it would be good to know that too). I dropped the FSB to 1778 (4000MHz) with the same VCORE, and got about the same Prime95 result (an error after about 20H). Lowering the VCORE gave errors much quicker.

Then I decided to dial WAY back and try a 1600FSB (3600MHz). I dialed down the VCORE in the BIOS to about 1.45, which should be more than enough with a slightly smaller vdroop. Booted to desktop, Vista locked in about 5 seconds. So I bumped up VCORE again to about 1.475 in BIOS, same thing happened.

Now I started scratching my head. I switched on EPP and let everything go into auto mode, which is what I originally worked around in Vista with -- 3.0GHz, 1333FSB, but kept VCORE around 1.45 (which is higher than it needs). Same Vista lockup on desktop after about 5 seconds.

The only thing that solves the lockup? Boosting to 4.0+ GHz!!

In addition to the questions I had lurking throughout the post, anybody have any ideas on why this might be the case? Also, this is my 1st time building a rig and tweaking, so if I left out any information, I'd be happy to follow up with it.

Thanks in advance!

-T.C.

 

[JAG87]

This is easy. Point finger at northbridge.

When you set the FSB down to 1600 and 1333, the NB voltage goes down as well, and I think that in combination with 2000 Mhz DDR3 memory, the NB has trouble (or better yet the memory controller).

When you run 1800 FSB, the bios automatically juices up the NB voltage, and that helps it cope with 2000 Mhz DDR3.

That's my theory, try it out. Put FSB to 1333 or 1600, and manually set the NB voltage to 1.3 or 1.4v. I bet you wont get the lock up

 

[Viscosity]

I'm not sure what the NB voltage setting is on the 790i.

Here are the available voltage parameters for the board:

CPU Core: xx
CPU FSB: xx
Memory: xx
nForce SPP: xx
nForce MCP: xx

Unfortunately, the manual isn't very clear on what the SPP or MCP do (and I don't even really know what the CPU FSB is for). I have had these on auto for the most part.

I'll try to boost all of those to the max green settings and see if it changes anything.

brb!

 

[JAG87]

SPP = northbridge
MCP = southbridge

SPP is what you want to increase, leave MCP alone.

 

[Viscosity]

Nope!

FSBs 1333, 1400, 1600, 1700 didn't work with SPP set at 1.3- 1.45.

I do have it dialed down now to 1750, that seems to work with SPP and FSB set on auto voltages. Still want to know what the deal is with the lower clocks.

And thanks for the help so far JAG87.

EDIT:

Just as a note: The lockup happens at the same time, in the same place, across FSB and voltage settings under FSB 1750.

Vista boots to the desktop, and when the little loading animation attached to the cursor finishes, the system locks (about 2-3 seconds on the desktop). Requires a manual reset. Hope that helps.

 

[Team42]

In reply to JAG87:

Sorry mate, but I'm fairly positive (around 99.99%) that's wrong.

MCH = Memory Controller Hub aka Northbridge.

That's the setting that needs to be tweaked. The Southbridge can generally be left alone.

 

[Viscosity]

According to this diagram, I think JAG is right:

It looks like SPP refers to the northbridge on the 790i chipset:

http://www.xbitlabs.com/articl...-striker-ii-nse_2.html

The problem persists. I still have no idea how to get the slower FSB to work. Getting a little frustrated / worried at this point.

 

[TC91]

spp is the northbridge on nvidia chipsets

 

[Viscosity]

I don't know if I will have this problem fixed anytime soon, so:

Is it okay to leave it at 4GHz, with an idle Vcore of 1.46? When under load, the vcore drops to 1.432 stable.

 

[Team42]

My mistake: sorry for that (that 0.01% error can be a real bummer, can't it?)

 

[CurseTheSky]

I didn't read through your entire post (low on time) but here's a few quick suggestions:

-Vista (especially x64) seems to be MUCH pickier about memory timing and subtimings than XP. If you haven't already, try pulling the battery / using the jumper to reset the BIOS. Using the factory / fail-safe defaults option in the BIOS wasn't enough for my situation. If your original Vista install seemed to take FOREVER compared to XP, this is probably the culprit. I had to reset my BIOS, then reinstall Vista for everything to work correctly (and the install took about 1/5th of the time it did originally).

-Try taking out one graphics card and anything else that isn't necessary to see if it still locks up. If it doesn't, it's either a bad driver, some kind of conflict (IRQ?), or a power supply issue.

-I heard 1.45v was a bit high to run a Peryn (45nm) chip constantly, probably equivilent to 1.50-1.55v on a Conroe (65nm). If you're not worried about replacing the chip, I'd go for it (with the specs of that system, money doesn't seem to be a large issue...), otherwise you may want to back down a bit.

 

[Viscosity]

Cursethesky,

Thanks for the tips. Before I got to your message, I had already done a full format / reinstall of Vista 64, and it started up beautifully. Installed in 20 min like it did the first time.

I think somewhere in my excited-first-time-overclocking state I switched a lot of the memory stuff without thinking about it. When I disabled EPP and unlinked the memory, I manually put in the timings (which worked fine).

Now when I think back, I'm pretty sure I switched between Auto ("optimal") memory timings and manual, and the subtimings from the auto got screwed up from the FSB switching and never reset when I put it back to manual. I'll have to record what they are.

As of now, I have the FSB pushed to 1600 (3.6GHz), and have an idle vcore of 1.448. I'm gonna run some torture tests and see if its stable, and then pull back a bit on the voltage if I can.

I'll be weary of pushing for 4.0Ghz so fast next time.

Thanks for the help everyone! I hope to be more active in the forums from now on since I ditched the pre-built Dell habit for custom hardware.

- T.C.