Core i7/X58 Overclocking Thread

[Hooti3]

I had a similar problem with my first i7 920,it would not fully post above bclk 172 with the same diag code as yours.I have since sold that i7 920 and have a new one which does bclk 200 (previous batch no 3838a820 ,new batch number 3839a557).Stay clear of 3838a820!

I did have the 3838a batch, but I got a new better batch on a new CPU 3837a899.

But I still get that code.

Can anyone help?

[Brahmzy]

great news - congrats!
For the life of me I cant get stable at 19*210 - can you post your complete bios settings please.

This is on an ASUS P6T WS Pro FWIW:

CPU - 19
BCLK - 211
DRAM Freq - 1269 (624.5 (2:6))
UCLK - AUTO (Can go up one step further, but I see no tangible perf. benefit)
QPI Freq - AUTO
DRAM Timing - 8-9-9-23, 1T (any lower and Prime+LinX BSOD's at 4.0GHz)
CPU V - 1.375 BIOS (2 increments lower and Prime+LinX locks box up, no BSOD)
CPU PLL V - AUTO
QPI/DRAM V - AUTO
IOH - AUTO
IOH PCIE V - AUTO
ICH V - AUTO
ICH PCIE V - AUTO
DRAM BUS V - 1.60 (any higher up to 1.66 doesn't seem to help anything except heat up my CPU more. 2 increments lower and Prime+LinX BSOD's)
All other DRAM V - AUTO
LLC - EN
CPU DIFF - AUTO
CPU Clock SKEW - AUTO
CPU SS - DIS
IOH Clock SKEW - AUTO
PCIE SS - DIS

HT - EN

All other Intel CM/TM/VT crap is off.

Funny, everything on AUTO offers me the best results. I tried messing for hours upon hours with PLL and especially QPI voltages, all would break stability. Also, I would disable HT and go for raw clock speed, except I need the 4 other "cores" for rendering purposes.

Full load after hours of Prime+LinX gets my lapped TRUE Black to around 87/88 degrees C. I'm fine with that. We'll see if these i7's degrade, or "burn-in" at all...

And S3 resume (sleep resume) doesn't wake the vid-card up... just black screen. That SUCKS - I rely heavily on using sleep functionality in Vx64.

[JWMc]

Sorry to re-post the data but I wanted to draw some conclusions from it:

Ubuntu Desktop 2.6.27-9-generic SMP x64, compiling kernel version 2.6.28.1
make menuconfig, saved without changing settings.
ReiserFS on a Seagate 7200.11 1TB drive

time make -j8 vmlinux, fresh boot, text mode, 3-6 back to back runs

best real times used:

3.8ghz, 3200mhz uncore: real 1m2.207s/user 5m53.758s/sys 0m42.675s
3.8ghz, 3400mhz uncore: real 1m1.152s/user 5m51.986s/sys 0m42.591s
4.0ghz, 3200mhz uncore: real 0m58.972s/user 5m37.037s/sys 0m42.019s
4.0ghz, 3400mhz uncore: real 0m56.923s/user 5m36.913s/sys 0m41.127s
4.0ghz, 3600mhz uncore: real 0m56.887s/user 5m33.673s/sys 0m41.039s
4.2ghz, 3200mhz uncore: real 0m55.090s/user 5m23.288s/sys 0m41.315s
4.2ghz, 3400mhz uncore: real 0m54.617s/user 5m22.072s/sys 0m39.890s
4.2ghz, 3600mhz uncore: real 0m54.440s/user 5m19.956s/sys 0m39.370s

Looks like you need at least 3400mhz uncore to not hold back the cores at 4.0ghz.

Going from 3200mhz to 3400mhz uncore @ 3.8ghz is a small gain but it seems more stable on my system so I'm using it.

Uncore gains @ 4.2ghz are relatively small so feel free leaving it low to be stable.

[dejanh]

Well, after some time fiddling with the FZE I got some results...nothing that I go "wow" at really, but ok I guess. It shaves off about 7C compared to the TRUE at 100% fan, and about 4C at 60% fan (comparable noise to TRUE).

Testing with my low-voltage 3.8GHz OC using QPI:Uncore in 1:1 (3.6GHz each)...

Nice idles at 25C, but load is not so good. I am expecting load in the mid 50s for this clock, but I am peaking out at 61C using two instances of LinX. The temps are part of the second 20-run (I did one 20-run first to warm everything up).

Anyway, here it is...oh and, there is a slight chance that 2-3/6 of my TECs are not working correctly, so I am RMAing this baby later in the week and will be doing more testing with the replacement unit. I did the "low" idle test and the coolant temp could not drop below ambient (23C) even with TECs and fan at 100% for an extended period of time. Definitely no condensation or anything which indicates a problem with TECs. They should get very cold if working properly according to Coolit, like not ice-cube cold, but very cold-water cold (between 7C and 12C).

[mpower1001]

SSD action? - http://www.xtremesystems.org/forums/...&postcount=531

Wow!! 1323MB/s Burst...Incredible Stevero

[namegt]

i7 920 HT on Threads 8

CPU i7 920
Mainboard EVGA X58 3X SLI
RAM samsung DDR3 1G PC3-10600U CL9 1x3
VGA EVGA GeForce GTX 260
Power Topower Octopus 1500W
Cooler WATER [GTZ]
OS WINDOWS XP PRO SP3 [OEM]
BIOS 12/29





i7 920 @ 4.2G
Threads 8
1000MHz 9-10-10-30 2T
32M pi


i7 920 @ 4.2G
Threads 8
1000MHz 9-10-10-30 2T
EVEREST C & M


i7 920 @ 4.5G
Threads 8
32M Pi


i7 920 @ 4.6G
Threads 8
1M Pi


i7 920 @ 4.65G
Threads 8


http://valid.canardpc.com/show_oc.php?id=488163

[Chicken Patty]

^^^amazing, great job

[GenTarkin]

With these gigabyte boards, I have a question. There are 2 entries in BIOS that say CPU clock drive and PCIE clock drive, both have options of 700mv 800mv 900mv 1000mv ... what are these exactly and should I be paying attention to these for stability increase when oc'n the cpu? Stock I think they are set to 700mv

[roller11]

Excellent work, nameGT! batch# ?

[anzial]

Excellent work, nameGT! batch# ?

It's 3838B079
Me thinks these are suicide runs, none are 24x7 stable Note, how he never posts temps under load, only idle

[Amuro]

i have a 965xe. Am i right in saying _ALL_ i need to do to overclock this is just increase the multiplier (along with Vdim) and not worry about the QPI? I am currently running at 4ghz stable with:
BLK: 133Mhz
Multiplier: 30x
Core voltage: 1.275

My QPI is currently 3.2GHz.

I'm wondering about the same thing. How much boost in performance will overclocking the BCLK result in?

[SAE]

Note, how he never posts temps under load, only idle

You can see max temps in realtemp, so what...?!

[D3mon_Hunt3r]

That's what I did, but are you cooling a GPU with that? If you are, you'll see an increase in temps depending on the card. I have 2 - 4870X2 in the same loop with a double and triple rad. I removed the vid cards and my temps are 5-7C cooler idle. I still have to run some tests to see if the cooling difference is greater under load.

I have individual loops. CPU, VGA('s) & MB/HD's.
I'm going for a triple with push and pull for the CPU, the two dual's i'll keep for the VGA's.

[peebee]

is there a thread showing Batches and results of i7 920 here in XS?

€dit: Just found it here

Sry!

[Lukee]

With these gigabyte boards, I have a question. There are 2 entries in BIOS that say CPU clock drive and PCIE clock drive, both have options of 700mv 800mv 900mv 1000mv ... what are these exactly and should I be paying attention to these for stability increase when oc'n the cpu? Stock I think they are set to 700mv

800mv for first one and 900mv for second one.

When you load optimum settings it will give you those.

[Lukee]

Well, after some time fiddling with the FZE I got some results...nothing that I go "wow" at really, but ok I guess. It shaves off about 7C compared to the TRUE at 100% fan, and about 4C at 60% fan (comparable noise to TRUE).

Testing with my low-voltage 3.8GHz OC using QPI:Uncore in 1:1 (3.6GHz each)...

Nice idles at 25C, but load is not so good. I am expecting load in the mid 50s for this clock, but I am peaking out at 61C using two instances of LinX. The temps are part of the second 20-run (I did one 20-run first to warm everything up).

Anyway, here it is...oh and, there is a slight chance that 2-3/6 of my TECs are not working correctly, so I am RMAing this baby later in the week and will be doing more testing with the replacement unit. I did the "low" idle test and the coolant temp could not drop below ambient (23C) even with TECs and fan at 100% for an extended period of time. Definitely no condensation or anything which indicates a problem with TECs. They should get very cold if working properly according to Coolit, like not ice-cube cold, but very cold-water cold (between 7C and 12C).

Why are you running 2 instances of LinX?

[Lukee]

I have individual loops. CPU, VGA('s) & MB/HD's.
I'm going for a triple with push and pull for the CPU, the two dual's i'll keep for the VGA's.

That'll work. Are you interested in a couple 4870X2 w/ waterblock?

[ReverendMaynard]

i7 920 HT on Threads 8

CPU i7 920
Mainboard EVGA X58 3X SLI
RAM samsung DDR3 1G PC3-10600U CL9 1x3
VGA EVGA GeForce GTX 260
Power Topower Octopus 1500W
Cooler WATER [GTZ]
OS WINDOWS XP PRO SP3 [OEM]
BIOS 12/29

Can we see that running more than single thread apps GT? Your clocks are awesome, but you ran superpi.....great. Get us some wprime numbers and lets see more than one core operate

[dejanh]

Why are you running 2 instances of LinX?

One instance of LinX only loads 50% of the i7 if running all 8 threads.

[JWMc]

One instance of LinX only loads 50% of the i7 if running all 8 threads.

Only if you forget to re-enable 'Two Instances (For HT cpus ..)' in the settings when you come back from testing with HT off. :P

[Lukee]

One instance of LinX only loads 50% of the i7 if running all 8 threads.

Which version are you running? To my knowledge, version 0.5.5 stresses 100% of the threads.

[JWMc]

Which version are you running? To my knowledge, version 0.5.5 stresses 100% of the threads.

Oh it does, you will hit 100C if you push it hard enough on air.

[GenTarkin]

yeah linx can be quite the beast in getting much higher temps than prime95 does. Also, try putting it on max problem size your amount of RAM will allow, I found that my processor can be stable all day long with problem size of 10000 but, if I put it up higher to where it uses 100% of my RAM thats free or so, it will find instability much sooner.
Think maybe this happens because its putting more stress on the IMC.

[Brahmzy]

LinX+Prime=FTW.

[JWMc]

LinX+Prime blend=FTW.

Fixed, and nice car. I autocross mine.

I'm about to post up my screenshot of LinX x 20 @ 4.2ghz, 6gb, HT -on-, AIR COOLING. It has taken me a while to learn the right settings to use!

I will do Prime blend next and then go cry in a corner when it BSODs.