Clevo Overclocker's Lounge

Fortunetely for me the only instability so far with XMP1 or custom with the same settings has been with overclocking over 4.4ghz. I have not tried much since I just got the ram but anything over 4.4ghz is unstable. I was able to reach 4.6ghz with 2400mhz ram with -20 adaptive. Some fine tuning If I plan to overclock more at a later date. Sounds like there is a difference though from computer to computer.

 

Try it with HWiNFO64 running with SMBus Support enabled if you have not already. Would be interesting to see if you get memory test errors that way. If you do, disable SMBus support, shut down HWiNFO64. Wait a few minutes and test again. If no errors launch HWiNFO64 with SMBus Support already disabled and see if you still have no errors. Use TM5 in Windows to save time. If you have no errors with TM5 then chances are good Memtest86 will also pass without errors.

https://drive.google.com/file/d/0Bwdqi25LDwZyN1g3X01LZHZiOHM/view?usp=sharing

 

Boot XMP1, then reboot to bios and change to custom and then set 1.25V. Then reboot and and wait.....When it boots to windows you should be stable from that point on. Depends on what is stored in nvram....

 

This might not count as an overclock, but I did it and it was pretty funny.

So I was playing Shadow Warrior 2 at absolute full settings. Then my brother asked me to play dota2, which also runs at full settings.

So I double timed the games using ALT + TAB and I played both Shadow Warrior 2 and Dota2 at the same time. Changing between them was seamless and the whole experience was as fun as you would expect. Won both dota2 ranked match and the boss fight in shadow warrior. With those two games, temps stayed below 90C on CPU while the GPU fans barely needed to spin up, 1080 drove both games at the same time with ease.

This is how it makes one feel

 

I think this is what you were asking? Enabled



Disabled

 

Thanks I will give that a shot. 1.25V will be safe?

 

How is it your test is done in 3 minutes?

1.25V should be fine for you.

 

Thanks again, ill have to do some additional testing was able to get to 4.6 without locking up. Just a real quick test.

 

Voltage with adaptive looks rather high though so stopped with the test. Will have to dial it in manually.

 

I am not sure have never used HWInfo to check ram stability and that was what it was set to when I downloaded it. I did run memtest86+ last night with no errors.

 

Yes, thank you so much for that. I see you have errors both ways.
Try rebooting after disabling SMBus Support. If that does not work, try clearing NVRAM and test again to see if the errors are gone with SMBus Support disabled. There could be something saving in NVRAM that is naughty. But, also be aware that running XMP 3000 default values I still have a few errors (like you posted) no matter what. Only switching to custom profile from XMP and tweaking the voltage and tREFI stops them completely for me. The SMBus Support being enabled in HWiNFO64 was causing my SPD checksum errors.

That would be normal. SMBus Support is typically enabled. It was only by trial and error searching for a reason to help me understand what was happening that I discovered this. I was just trying everything I could think of and that turned out to be relevant. It's not supposed to be that way, but until Clevo (or @Prema) can figure out what got messed up this workaround may be the only solution for people that insist upon using HWiNFO64. Bottom line is, whatever is causing it needs to be fixed. It seems like a mistake on the Clevo side of things rather than the HWiNFO64 side of things. I value HWiNFO64 way too much to stop using it. I would not know how to act without it because I have been relying on it for real-time critical system information so many years. I could use other things, but HWiNFO64 has almost zero overhead and the RTSS/OSD support is something I find extremely valuable. I don't even enjoy playing games without that OSD thing any more because I am so accustomed to it and OCD about knowing what is going on with my system at all times.

 

The second test in the info I posted with SMBus support disabled showed 0 errors.

 

Ooops, sorry I missed that. My bad.

 

So with tweaking and voltage were you able to get 0 errors with SMBus support enabled or do you still get them?

 

No, I cannot have SMBus Support enabled at all, or it causes SPD checksum errors and errors from memory during testing. Fire Strike and 3DMark 11 also frequently crash due to the memory errors. The tweaking in the custom memory profile gets rid of a few straggling errors after disabling SMBus Support. With SMBus Support enabled I get literally thousands of errors sometimes, and almost always hundreds of errors.

Something is definitely not right with that. Brother @afloyd was having the test finish too quickly at one point as well. Maybe he can tell us what fixed that for him. With 32GB of RAM is should take between 10 and 11 minutes and just about half that much time with 16GB of RAM.

 

So is there a way to disable this in the bios or just stuck with it being enabled? I will run some 3DMark as well when I get some time and see if I crash.

 

No, SMBus Support is a software feature of HWiNFO64. Simply take out the check mark and that will disable that feature in HWiNFO64.

The actual System Management Bus cannot be disabled in the BIOS. ACPI can be disabled, but not the entire SMBus. That's an integral part of the laptop. >>> http://smbus.org/specs/

 

I figured out why it was only 3 minutes. Instead of using right click and run as administrator I went into properties and assigned the program administration. I did this with SMBus Support enabled and got 0 errors at 1.25 voltage. Going to reboot with it enabled and try again.

 

After rebooting.

 

Just one error is too many to say your machine is stable. Does the error go away if you remove the check mark from SMBus Support, reboot and run TM5 again?

 

Yes it does go away when I removed SMBus Support; however, I also ran at XMP2 custom with 1.25 Voltage and here were the results with SMBus support enabled. One run is probably not enough for testing purposes but I will have to make more runs tomorrow after work. Did not show any errors.

 

Well, maybe the people at Futuremark will pay more attention when they see this and people start complaining and asking questions. What is happening brings the integrity of their product under question and scrutiny. If their database is becoming skewed and unreliable over time because of erratic changes between GUI versions, then nobody should give a rat's ass about Fire Strike scores any more.

Gradual Degradation of 3DMark Fire Strike Produces Unreliable Results

The lowest score is the "latest and greatest" version of Fire Strike. The benchmarks were run only minutes apart on the same machine with identical CPU and GPU settings with no changes whatsoever except for the version of Fire Strike.

 

I think you will find it is consistently error free with SMBus Support disabled. I hope so, because I do and I tested it about a dozen times randomly over the past few days before I posted the information today. You may need to bump the tREFI as I showed in the example photo of my BIOS settings before the errors are gone and stay gone. Only increasing the voltage wasn't enough for me. The tREFI was too low.

 

Remember....This is only if you run hwinfo in the background. If you don't use it it should still be error free.

 

Yea the first few times I did TM5 I was getting low test times. Honestly not sure what I changed that fixed it though. Tried switching memory profiles a couple times maybe thay jogged it somehow....

Also, never really thought about it until you've mentioned it here. But when I run HWInfo64 I have always had the SMBus disabled because of the warning it shows that it could cause system instability... I'll have to try TM5 with & without it again tomorrow and post back as well

 

I have sent a new feedback mail to Kokko on Futuremark as we have agreed(The 10th or 15th mail in the row now). This time with a little more powerful patron... Futuremark have said long time ago to me in mail that "They have reproduced in-house and investigation is ongoing" Nothing happens!!!! And when they finally push out the new 3DM version after 3 months FIX, so is the benchmark software in an even worse condition... Like the last time... New 3DM suite UI 2.2.3488 64 version out 9th Dec. = Fiasko!! Then they need to push out an even newer one The UI 2.2.3491 64 - 1 day later aka 10th Dec. Same ****y happened last two time as well(in July and Aug). Futuremark have BIG problems with their 3DM Suite!!!

 

Right clicking on hwinfo64 and going to properties to set the program to run as administrator is how I got it to run properly.

 

Just as a update even though it passed as XMP2 with SMBus support enabled it failed the second run. I am going to go back to XMP1 at 1.25 Volts and will run several tests over the next few days with support disabled and if it fails I will bump up the tREFI as well and give it a try. Thanks.

 

I don't think that's what caused my short TM5 runtime if that's what fixed it for you... I've always just run HWInfo64 without administrator privileges. However, TM5 I do have to run with administrator privileges, otherwise it shows an error about that in it's console window.

 

I hate to be a wet blanket, but are you /sure/ those days ever existed?

Sent from my SM-G900V using Tapatalk

 

@Mr. Fox tRREFI was not a option for me to change the values in the bios I have. After running 2 cycles with support disabled I still got a error. I guess that is only with the Prema mod that you have the option to change tREFI.

 

Yes, but maybe I am old enough to remember them. It has been a while. But, we can get back to that if enough people wise up, do the right thing, and stop letting others do all of their thinking for them. Sometimes it takes hitting rock bottom before anyone looks up.

Yes, that is true. I forgot it was not available in the stock Clevo BIOS.

 

Well, our friend Tamas has already responded with a fix for the CL38 AIDA64 glitch. That guy is really awesome. No wonder AIDA64 is so excellent.

https://www.hwinfo.com/forum/Thread-SPD-Checksum-Errors-w-HWiNFO64-Running?pid=12891#pid12891

 

Thanks, is there a way to clear the NVRam without taking the Lithium batter out?

 

Please buy a license to support this awesome guy and tool

Same to you @Papusan

 

I run all tools as admin, in fact, I right click on them, go to compatibility, then change settings for all users, then set them to run as admin by default so I don't have to right click and choose Run As Admin everytime

It's very essential for tools that interact with the hardware/OS on a deeper level.

I do this for AIDA64, HWiNFO, CCleaner, AS SSD Benchmark, CrystalDiskMark, GPU-Z, etc.

 

Yes, I agree. I have purchased many and will continue to do so. AIDA64, HWiNFO64 and ThrottleStop are all amazing products developed by amazing people.

 

I forgot to change this after W10, explains a whole lot!

 

You mean my payd Aida64 version? Are you kidding with me? I know what is awesome hardware and software!!

 

I know you know what's awesome. You don't know how to joke?

 

And you took me serious?

Spoiler: Don't click here :)

 

Indeed.

 

I have not bought Aida64 yet, but I think I will finally pick it up this week. I probably will buy the boxed version, because I like my software to be displayed on my shelves
The boxed version is cheaper than the digital version (at least for 1 year license) and the box itself also looks awesome!


It's simple, but I like it.

I only regret not buying it when I was a student, the 50% student discount is really nice.

 

@Mr. Fox I have a few more questions about TestMem5 If you don't mind. I see while running the program with HWMonitor enabled my CPU is stressed at 100 percent. Is this program running stability based on the CPU overclock and the Ram. Will it show if both are unstable and throw a random error based on either being unstable. I did not see any issues previously when running a blend with Aida64 at 4.4ghz and XMP1 but am seeing a error with this program. I reset my CPU clock to default and 0 adaptive voltage in which I was running -140 with a 4.2ghz overclock for the initial tests. What I saw at stock was TestMem5 showed 1 error with XMP1 at 1.2 Volts. I then ran XMP1 in custom at 1.25 volts with default stock clocks and 0 change with adaptive voltage. I was then able to run TestMem5 3 different times and had 0 errors. I just wanted to make sure I was going about this the correct way and what other programs I should use to test instability. My next step I believe would be to either overclock to a higher level and test stability or if I want to keep stock clocks to begin undervolting the CPU till I start hitting errors again and then back off. Is this correct I was unsure what this program was testing exactly.

 

I think you are doing everything fine. TM5 is not a CPU stress test, although it will load the CPU. If the CPU overclock is unstable TM5 can capture memory errors from that. TM5 is not as thorough as Memtest86 in terms of isolating the memory only because it is run in Windows and other things can affect it. Although, I note that if I see errors in TM5 I almost always see errors in Memtest86 and vice versa. At this point it seems your system is working pretty well. If all games and benchmark run stable and temps are good I would not worry too much with memory testing. The reason I started all of this testing was because I was having lots of stability problems with memory running faster than 2400. I still do not know why I was able to overclock the Kingston 2400 to 2800 and 3000 without test errors and no SPD checksum errors, but the G.SKILL and Corsair had problems. HWiNFO64 did not have the same effect on the Kingston memory, but maybe it had something to do with it being 4GB sticks.

 

Ok, good deal just wanted to make sure I was doing everything I could to be stable and was only using Aida64 and memtest86+ to base it off of.

 

I don't use it much, but it is a good product. I got the 3 year license because it is a better deal and it is great software...

 

Wow, just noticed F8 actually works now without making changes to the bcd file. Did anyone else know about this?

 

That usually works for me as well if I am using either UEFI w/ CSM or Legacy Mode (UEFI disabled). I have not been able to do it with pure UEFI mode, so I still use the command prompt BCD edit to force the legacy boot menu. Older versions of W10 and W8 did not work without the bcdedit. I don't know what version of W10 it started working. I discovered it totally by accident. What is strange is if you don't do the bcdedit sometimes it will be black like that screen and other times it will have a blue background. That part I do not understand.

Edit: It may have something to do with dual-booting W7. Is yours set up to dual boot at this time?

 

I have the dual boot in windows 10 boot manager.

Edit:
I think you are right.
On My asus drive it boots straight without stopping. No matter how many times i hit F8.