Maximum PC

Hi folks,

[If you don't want the backstory, skip ahead]

Those of you with eidetic memory might remember the last time I posted, about 3 years ago, when I built my first rig. I've been running that rig for a long time, but it was rocking XP forever and i decided I wanted some DX10 goodness and upgraded to 7 with a clean install (to paraphrase Dr. Evil: there's nothing quite like the sensation of a freshly formatted primary partition. It's exhilarating; I suggest you try it).

OK fine, so I installed 7 and spent a week or so getting everything back in the groove. And I notice that the 2 GB ram I initially had in there was starting to look a bit measly. Enter Newegg, got me 8 gigs of DDR3-1333 to max out my mobo for $75 all-in. Nice. Well, you know where this leads... All this nice ram, and just a measly 8800 GTS 512 to play with? No way. So on to ebay, got me a second identical EVGA card for $55 and now I'm running SLI. Yeah, I could have gotten a single card that's better than both of them, but not for less than $100 and I'm kinda poor right now. Sweet! Crysis, here I come! (I'm behind the times on games, too.) Well crap. Random hangs. What's up? The voltages on my 3-year-old 550W PSU that's been running 24-7 the whole time. Can't hack. No good. Well, PSUs I won't get used, and if I'm gonna get a new one I may as well get a bit more headroom and try OC'ing, right? So I got a CM GX 750W for $85 after rebate, and for good measure a CM Hyper 212+ for $25 to help with the thermals. All told I spent $240 but the rig should now last me a few more years, so I regard it as money well spent (my wife disagrees).

[END BACKSTORY]

So I'm trying to OC the processor now, but I'm having a hell of a time of it. All the forums I've seen suggest 3.6 GHz on stock volts, and 4 GHz without breaking a sweat. I'm just not finding it so. I got to 3.6 after some conniptions, boosting the volts to just over 1.3, and right now I'm stress-testing at 3.825 GHz (425x9) with vcore set to 1.3625 in BIOS, which is Intel's design limit for the chip (though CPU-Z shows a lower voltage). This is the best I can do. If I go up to 430x9 it BSoD's within 5 minutes. Temps are in the 50-53c range, but I'm leery of going over the design voltage.

I know, I should be happy I'm getting a 25-30% overclock. But everyone says 4 is easy, so I feel like a dweeb if I can't make that Does anyone have any advice? Did I just happen to roll craps on the chip I got? Am I doing it wrong?

System:
EVGA nForce 790i Ultra, P10 BIOS
Intel Core 2 Duo E8400, rev C0
Cooler Master Hyper 212+
2x4 GB Kingston Hyper-X PC3-10666
2x EVGA Geforce 8800 GTS 512 in SLI
Cooler Master GX 750W
Win 7 Pro 64-bit

Yep. Change your RAM divisor and lower the speed of that memory to back within spec. Stop adding voltage!

vig1lant3, I'm running the FSBs unlinked, so the RAM is still at 1333. I ran into that problem early on and figured it out; it was folded into the word "conniptions"

Gonna need some screenshots of the BIOS if that's possible then...been a couple of years since I OC'ed any 775's.

Seriously, stop running your CPU at those voltages. I've seen that chip at 4.2Ghz @ around 1.28V -1.30V. Seen the chip itself run even faster, but can't testify to how much power was or wasn't added. Point is that lack of juice is very likely not causing your instability problem. Not at that speed anyway, so no reason to fry the chip in oil until we know for sure!

OK, for the time being I dialed it back to 3.6 GHz and auto-vcore, which seems to be stable.

I have no idea how I would go about screencapping my BIOS, so I just wrote down pretty much all the settings. I know some of them will be irrelevant, but better to be thorough than go through that tedium again

CPU Internal Cache: Enabled
Quick Power On Self Test: Enabled
APIC Mode: Enabled
MPS Version Control for OS: 1.4

CPU Multiplier: 9X
FSB - Memory Clock Mode: Unlinked
FSB (QDR), MHz: 1600
MEM (DDR), MHz: 1333
P1: Auto
P2: Auto

Memory Timing Setting: Expert
tCL: 7
tRCD: 7
tRP: 7
tRAS: 20
Command Per Clock: 1T
tRRD: Auto(4)
tRC: Auto(33)
tWR: Auto(12)
tWTR: Auto(16)
tFAW: Auto(21)
tREF: Auto(7.7uS)
tRFC: Auto(74)

PCIe x16_1 & x16_2, MHz: Auto(100)
PCIe x16_3, MHz: Auto(100)
SPP<->MCP Ref Clock, MHz: Auto(200)

HT Multiplier:
nForce SPP --> nForce MCP: 5x
nForce SPP <-- nForce MCP: 5x

HT Spread Spectrum: Auto
PCIe Spread Spectrum(SPP): Disabled
PCIe Spread Spectrum(MCP): Auto
SATA Spread Spectrum: Disabled

Limit CPUID MaxVal: Disabled
Intel SpeedStep: Disabled
CPU Thermal Control: Disabled
CE Enhanced Halt State: Disabled
Execute Disable Bit: Enabled
Virtualization Technology: Enabled

Voltages:
CPU Core: Auto(1.31V)
CPU FSB: Auto(1.3V)
Memory: 1.65V
nForce SPP: Auto(1.35V)
nForce MCP: Auto(1.5V)
GTLVREF Lanes 0-3: Auto(+00mv)

NVMEM memory test: Disable
OS Comp: Auto
System BIOS Cacheable: Disabled
HPET Function: Enabled

IEEE1394 Controller: Enabled
JMicron AHCI Controller: Enabled
HD Audio: Auto
IDE HDD Block MOde: Enabled
Onboard FDC Controller: Enabled
Onboard Serial Port 1: 3F8/IRQ4

ACPI function: Enabled
ACPI Suspend Type: S3(STR)
Soft-Off by PBTN: Instant-Off
WOL(PME#) From Soft-Off: Disabled
WOL(RI#) From Soft-Off: Disabled
PWRON After PWR-Fail: Off
Power-On by Alarm: Disabled
POWER ON Function: BUTTON ONLY

Init Display First: PCI Slot
Reset Configuration Data: Disabled
REsources Controlled By: Auto(ESCD)
PCI/VGA Palette Snoop: Disabled
PCI Latency Timer(CLK): 32
Maximum Payload Size: 4096

Try loosening up your RAM timing a little. Without knowing your exact model of RAM, I can't say specifically, but DDR3 1333 Hyper-X comes in two timings, CL9 and CL7. I assume the difference is just binned chips that should stand up to the tighter SPD table. Yours may very well be CL7, but even though you're running unlinked, as you push the FSB higher you might consider running more traditional 9-9-9-27 CAS settings. That would be the first likely hiccup to rule out...

I loosened it up and that helped very little. I appear to be stable at 3.825 and I was able to turn the volts down a notch or two, but it's still over 1.35. Any less and it crashes.

My RAM is KHX1333C7D3K4/8GX, it advertises 7-7-7-20 timings.

Should I try 2T timings? I've heard that Kingston RAM prefers 1T...

You can set it to 2T just for troubleshooting purposes. On Intel platforms it wasn't until after socket 775 that the CPU's began to gobble up northbridge components, so I'd bump the voltage on the chipset and see how much mileage you get out of that. While you're at it, observe the chipset temps as well. That could be an issue also...one that is easily resolved with aftermarket parts if need be.

Er... noobish question: SPP or MCP voltage? How much should I feel good pumping it up?

I'm not sure I'll be able to tell what the temp is. According to Speedfan, my "system" temp is 124c. I somehow doubt that

More concerned with SPP, but you can bump both. Think of the MCP as a more traditional southbridge. Just touch the heatsink, you'll know if you have a problem.

OK, I got Everest... sorry... AIDA64 (bleh) to read my MCP/SPP temps, they're around 50c and don't seem to change much with voltage. I have the SPP up to 1.45v without any change in stability (still bsod's at 1700 FSB after 5-10 mins). I can push it to 1.5 or 1.55, but the BIOS shows those as "red" so I'm scared to do it. OTOH, the NB temps only rise a couple of degrees.

Trolling some other forums, I've also done the following:

Set the PCIe clocks from auto to 100 MHz
Set the SPP<->MCP clock from auto to 200 MHz
Disable all spread-spectrum, execute disable bit, and virtualization technology

I can push the FSB voltage as well. Should I try that?

Yes. The FSB voltage is a higher priority than the other clocks. The PCIe clock should always be left at 100Mhz (especially when running an SLI video solution). If adding voltage to the northbridge didn't yield any benefit, there's no reason to push any further. In addition, if adding voltage to the FSB doesn't help, then you've likely reached your ceiling for overclocking.

I havent posted here in a long long long time, however i read this thread and it got some gears turning. from what i recall back in the LGA775 days, the 790i wasnt really the greatest when it came to FSB overclocking, that distinction came with the P45 chipset. let me ask you this, have you dropped your cpu multiplier to say 6 and tested to see how high of an FSB you can get? if you get a 6x400 and its stable through testing, that will put you well below your processors stock speed however it will tell you that the FSB isnt the issue. at that point i would say it could very well be the chip itself. I myself had an E8400 that didnt like to run much beyond 3.8ghz



also looking at that and going back even further to my AMD Athlon 64+ nforce 4 days, does the HT link with an nvidia chipset play as big a role in an Intel system as it did back during the Athlon 64 days? can someone fill us in on that? if so, that could be your problem as well. could try dropping that multiplier down to 4 to see if it helps any

FSB does not seem to be the issue. I was able to run 475x6 at stock voltages and 500x6 by ticking it up 1 notch to 1.35 vtt. Rock solid with Prime95 and IBT

Reducing the HT multipliers seems to have helped marginally, i.e. it takes a bit longer to bsod at 3.825 (10 minutes vs. 5). Still not stable though :-7