Prime no 100 % capacity utilization on all cores

I have a gt73VR and repastet it yesterday.. now i want to try what are my limits with oc.

i never had such a problem on a desktop pc ... i used 1344k prime 26.6 with this setting


after some time the load dropt on 2 cores .. :X no idea why

 

I've seen this on my laptop. Something causes Prime95 to lose CPU thread focus. Not throttling or current limits either.
Can you please run prime95 with:

1) XTU completely closed down? Should not have this program open. XTU has been known to conflict with both Throttlestop and HWinfo64 because of them hammering the same CPU registers
2) HWInfo64 running, showing CPU TEMPS and VID in the background.
3) Change your prime window to have 8 windows open, 1 for each thread. This is hard to get correct, but it should look like this, without being super tiny:

 

ok thx 4 advice not to use hw64 and xtu

hope that screen is that what you want ?

 

Since Prime is showing a failure, your overclock isn't completely stable. Some folks don't think Prime is a good indicator since it puts extreme strain on the CPU, especially if their machine is fine otherwise. I don't fall into that camp personally, but how you want to handle this is up to you.

 

Yes, excellent. This is much better and shows each thread.
Ok you have a 6820HK, not a 7820hk.

And your prime dropped 2 threads with errors. That means your CPU is not stable. And since you are NOT using AVX or FMA3, then you are not stable at all. Prime should not drop threads with errors if AVX/FMA3 are disabled or you are using an older prime without AVX.

So your voltage settings are too low. If you want 4.1 ghz to be stable, you need to increase your voltage. I would try a +20mv higher voltage offset here. So if your current voltage offset is +10mv, try 30mv. If it's 30mv, try 50mv.

Make 100% sure you are using cooler boost 100% fan speed please. Do not use auto fans for this.

However i would not test prime at this point, if you are using 100% fan speed already. You are already getting high temps, so I don't think you would keep the temps under control very well at another +20mv. You can try it, but if it gets too hot, I would give up and try something like 10 consecutive runs of CINEBENCH r15 (10 runs in a row without waiting).

If you get no errors on your next prime and cinebench test runs, you are NOT finished. Please check windows EVENT VIEWER -->SYSTEM logs, and look for WHEA Correctable errors during the time matching your last test runs. If you get any of those, it means again that you are not stable, and the voltage is too low.

Keep in mind that the 6820HK CPU's tend to top out at 4.1-4.2 ghz on overclocks, and some reach 4.3 ghz, but usually on liquid metal repaste cooling. Looks like yours will be limited to 4.1 with a little more voltage and 4.2 ghz at absolute max (will require a repaste and much better cooling and a lot more voltage).

 

you safed my soul with this anwser with your answer below. I thought its like on my desktop 6700k ... where 4.5 to 4.7 was not unreachable with good cooler

thx also 4 that .. didnot know hat ... even that my core crashed the log would be clean or ?




prime startet on the log below at 12:10 prime ...

error 1 was on 13:30
error 2 was on 14:14

next time i use cooler booster. Rpm were not very high .. arround 3000
vcore offset was +30

i am no hardcore ocler and a lot of guys say 1 hour 1344k is enough. Can i consider my game now as " gaming " stable because a lot of guys would not go beyond 1 hour of testing ?

 

I wouldn't consider it stable unless you can pass at least 12 hours with AVX/FMA3 instructions enabled and 24 hours with them disabled.

 

@saturnotaku You can't pass 12 hours with AVX/FMA3 on this laptop with any sort of overclock
The EC will not allow it !!

I've posted this many times in the past but either people keep forgetting or they just don't care.

CPU_CrtT = 1.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0

Change CrT 1 to 0 for CPU and SYSTEM without permanently bricking the system then you can pass AVX/FMA3 up to 98C.

 

At least you can do this within the OS right?


Also running P95, not recommended.

I'd run realbench CPU instead, in all practical purposes if it passes, it's stable.

 

@Mobius 1 No, you need to *EDIT* the EC to do it.
Look for yourself in HxD. you'll see multiple strings. For CPU, System and even for Battery!
And these are not exposed in "EC RAM". Rather EC RAM can cause certain conditions to be "met", e.g. the Throttle flags turning off instead of on (like changing the power ID from 230W to 330W). But the Current Trip flags are forced ON with no corresponding "off".

And you can't just hex edit the EC. There are checksumS. That's plural.
Not to mention that 32 bit checksum at the end of the file.

Senso tried to disassemble it and he said its complicated and there's tons of empty space and strange code.

 

Why am I not surprised?

 

@Mobius 1 Um I don't know if that means I asked something bad or good?

 

I think they weren't surprised that MSI made the firmware so cancerous.

 

@saturnotaku

I just wish someone had a "spare" like half broken GT73VR.
So they could actually TRY hex editing that "1" into a 0 for CPU_CrtT

Then see if
1) ECflashwin_041.exe bombs with an error "Incorrect checksum!"
2) if #1 doesn't happen, then see if the laptop never boots again after the flash "completes"

3) if neither #1 or #2 happens, SOMEONE is going to become very rich...and popular..........

@Papusan

 

Misunderstanding

@Falkentyne you're ok

@MSI Myco @MSI_Darutohne @MSI_Josh you guys should be ashamed

 

@Mobius 1
If I had my Clevo P870 TM1 right now (im waiting for 8 core CPU, so is brother @Papusan ), I would hex edit that byte in the EC, change the 1 to a 0, try to flash it.
if EC flashwin says bad checksum, I would take apart the laptop and put the Skypro programmer to the EC chip.
Then throw crappy thermal paste on the CPU and GPU heatsink (removing my PERFECT THERMAL PASTE JOB with 0C-1C core temp difference:




Then I'd try to boot the laptop

If I bricked it, I'd take it apart, reflash the stock EC, then sell it or give it to someone who needs a taptop.

if I didn't brick it and the caveman hex edit worked and I could run 4.7 ghz with AVX and 5 ghz Cinebench with NO POWER OFF, I would be the most famous laptop user IN THE WORLD !!!! INNNNN THEEEE WORRRLDDDDDD!!!

 

If I can source a 17A2 bare motherboard and the accompanying cpu heatsink >_>

 

CPU heatsinks are easy but cost time and money I got one from Tom Ho (Shehary posted his email a few times in the main MSI thread), but took about 2 weeks because it had to get shipped from China (they don't keep them in USA, just fans). But cost me $70.

From what I read about a Lenovo EC, their EC had a main file checksum (like the FFAA or 8DB9 checksums or similar short checksums), which you can also see in HxD. Those are easy enough to circumvent the lazy way, if you change one byte to a 0 and another 0 to a 1, but that's probably not the first checksum the flasher checks.

@Mobius 1 : take a look:

And this was for a Lenovo
http://zmatt.net/unlocking-my-lenovo-laptop-part-3/

God only knows what technological terror MSI created with theirs....

So how does one change:
CPU_CrtT = 1.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0
to
CPU_CrtT = 0.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 0.SYS_ThtlT = 1.SYS_ThtlT = 0

without disassembling the entire EC, doing that there then redoing all the checksums and all that wizardry?

 

ok i am going down to my 3.8 ghz spot. I am no Falkentyne




But one thing was very interesting.

110mv prime crashes after a few minutes
100mv screenshot... stable all night.

this is my sweet spot. I stay on him and look forward to good temperatures.

 

and thx 4 advice with the WHEA Erros .. i was happy to found it also in hwinfo64 .. all on one screenshot and no searching in the windows log files

*only 4 info*