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?
Printable View
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.
Sorry to re-post the data but I wanted to draw some conclusions from it:
Looks like you need at least 3400mhz uncore to not hold back the cores at 4.0ghz.Code: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
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.
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).
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
http://cfs15.tistory.com/original/21.../497513c332666
http://cfs15.tistory.com/original/16.../497513c29b30d
i7 920 @ 4.2G
Threads 8
1000MHz 9-10-10-30 2T
32M pi
http://cfs15.tistory.com/original/24.../497513c2c986a
i7 920 @ 4.2G
Threads 8
1000MHz 9-10-10-30 2T
EVEREST C & M
http://cfs15.tistory.com/original/17.../497513c2de088
i7 920 @ 4.5G
Threads 8
32M Pi
http://cfs15.tistory.com/original/27.../497513c2f3c2c
i7 920 @ 4.6G
Threads 8
1M Pi
http://cfs15.tistory.com/original/13.../497513c311af9
i7 920 @ 4.65G
Threads 8
http://cfs15.tistory.com/original/33.../497513c31cea9
http://valid.canardpc.com/show_oc.php?id=488163
^^^amazing, great job :D
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
Excellent work, nameGT! batch# ?
is there a thread showing Batches and results of i7 920 here in XS?
€dit: Just found it here
Sry!
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.
LinX+Prime=FTW. :)