[Team.AU]Gigabyte X58-UD5, 3GB Corsair 1866MHz CAS8, Core i7 920 Testing thread

[dinos22]

i've only tweaked two timings which i mentioned earlier

this extra performance was from higher NB speeds and i think wazza may have worked better this time

still not sure what the difference is

[dinos22]

Hey fellas

small update from weekend LN2 bench session with the core i7 920

I've also interchanged between a GTX280 and 4870X2 as per screens

i think i've gotten as much as i could out of it and its time i do some serious benching with a 965 next





SuperPi 32M and 1M subtimings i used in bios
(its still a stab in the dark sort of thing so more tweaking is still possible )

[eva2000]

Congrats there dino excellent work indeed

still have your older GPUs for a few more points ?

[dinos22]

Update

gigabyte can go higher than 222MHz BCLK it seems

i've completed 32M SuperPi runs at 227.50MHz BCLK (gotta get the screens off the drive)

will report findings soon

http://valid.canardpc.com/show_oc.php?id=568106

[eva2000]

nice what gpu used and bios version ?

[Booj]

hmm interesting, i've seen hardly any Gigabyte boards over 222. My C0 can bench anything at 222mhz, but won't post above that. It seems strange not to have a decreasing slide in stability as bclk increases around that point. 222 all ok, 223+ nothing.

Curious to see if this is with newer F8 beta Bios, or if it's a D0 cpu thing. I don't think i'm too fussed anymore though. I see a 975 in the crystal ball hehe

[dinos22]

Here is the full update

I did some QPI/BCLK/bus frequency testing to see if i could break the 222Mhz "wall" I had in my previous tests. It was a bit of a tussle but in the end i managed to get some extra bus speed through few different settings outlined below. It seems as thought PCI Express frequency, uncore "sweetspot" volts and settings you boot at are the main driving factor from my initial testing in achieving higher/stable bus speeds or QPI or BCLK or whatever people call them. I was maxing out at 102.5MHz PCIe frequency in windows and while i also managed to find a sweet spot to let me OC a touch more (103 or so MHz) i was not able to make a run stick even though i did have a CAS6 run running at 228.3MHz bus only to crash on 16th loop...it looked like a 7m30s run ...

Here are my findings so far. Disregard the SuperPi efficiency.......i was just testing frequencies.

I'd love to hear from Gigabyte X58 board owners about their experience and whether these settings help anyone reach higher clocks.

I'm in touch with Gigabyte guys about this so hopefully they can figure out what it is exactly, whether it's the PLL chip or signaling/traces issue or whatever. All i care about is 250+Mhz bus speed stable on any GFX card

Test Configuration:

• Gigabyte X58-UD5 F4n bios (old bios for initial testing)
• Intel Core i7 920 retail chip 3849A910 D0 stepping
• 6GB 2000MHz CAS8 DDR3 Corsair Dominators
• Nvidia 8400 PCIe Graphics Card + Xfire ATI 4890
• 1000W HX1000 Corsair PSU
• 1hp C-BuZz Single Stage (kayl evap) > on CPU
• 160GB IDE Hard Drive
• Windows XP Pro 32bit SP3
• Windows Vista Ultimate SP1
  

BIOS SETTINGS:

• CPU PLL Voltage: Auto or 1.80v - same effect
• CPU VCORE: 1.35-1.36 - same effect
• IOH/ICH 1.1V Voltage: Auto or 1.20v - same effect
• IOH Voltage: 1.42v (i havent tried auto...dont think it matters)
• ICH PCI-E Voltage: Auto or 1.5v - same effect
• QPI/VTT: 1.433v or 1.455v = this chip's sweet spot (any higher or lower and i was unable to push past 101MHz PCIe frequency in windows)
• VDIMM: tried between 1.66v - 1.94v same effect
• BUS SPEED: booting from 200 or 210 same effect, lower or higher i had problems going past 101MHz PCIe frequency in windows
  for the record i could boot at 220 straight from bios to windows...sorry didnt try higher..will do next time
• PCIe FREQ.: Auto or 101MHz - same effect....102 would occasionally boot but every now and then throw an F6 error iirc...anything higher and no boot at all. In windows i did not see 101Mhz PCIe freq. in SetFSB even though i was selecting it in bios so not sure it was actually sticking at all.

222MHz Bus Frequency
101MHz PCI Express Frequency
32M SuperPi stable

recorded in the middle of the run (before i moved on to higher frequency)




223MHz Bus Frequency
101MHz PCI Express Frequency
32M SuperPi stable





225MHz Bus Frequency
101.8MHz PCI Express Frequency
32M SuperPi stable




226MHz Bus Frequency
101.9MHz PCI Express Frequency
32M SuperPi stable






227.5MHz Bus Frequency
102.50MHz PCI Express Frequency
32M SuperPi stable

[eva2000]

Interesting so this was reproducible on normal air/water for >222bclk ? Just wondering how much was due to the right settings versus how much was due to sub zero cooling ? Tried loosening those sub timings for higher bclk ?

Was the >101mhz PCI-E frequency being set using setFSB or what you ended up with after setting from bios boot ? i.e. 101.8mhz = 102mhz in bios for PCI-E ? I ask as on my DFI UT X58 setting PCI-E frequency in bios shows ~+1mhz higher in setfsb i.e. when benching with HD4870x2 CF with 110mhz PCI-E setFSB shows 111mhz for me.

Nvidia 8400 PCIe Graphics Card + Xfire ATI 4890

interesting combination ??? why the 8400 card ? for higher PCI-E ? Or to force PCI-E link bandwdith from x16 to x8 ?