*** The Official MSI GT75 Owners and Discussions Lounge ***

So, a month ago I put IC graphite pad on my CPU, and now, running prime95, I'm getting exactly same temps as back then (every core has same temperature).
I know maybe I'll have better temps with thermal paste, and less core difference with 0.5 arctic pads, but I'm ok with current temps and I know they will never get worse.
So, considering that repasting CPU requires dismantling of all 3 heatsinks - I will stick with IC graphite pad.=)

(with loadlines=0 it's 6°C cooler, with hottest core averaging at 72°C.)

 

There's *nothing* wrong with IC graphite pads if you are not overclocking.

For stock performance the pads will function better than the stock MSI thermal "Lairdtech" crap and will handle the convex heatsinks quite well and will not degrade. For the GPU, you will ABSOLUTELY NEED to use insulation (3 coats of nail polish around the SMD's will work; you don't absolutely need super 33+ tape instead of nail polish, as you aren't trying to help make an airtight seal, but it won't hurt to use Super 33+ tape).

The only problem with the IC stutff is if you want to overclock. The pads get saturated after a certain amount of heat output and falls off hard. They are significantly worse than Kryonaut. Just better than stock paste. Just be careful when applying them, make sure you have some sort of insulation and make sure the cutout is slightly larger than the BGA surface itself. You don't want hot spots because the pads slightly shifted when applying the heatsink.

 

MSI Sponsored FREE tickets to ESL ONE NEW YORK 2018 for both Saturday and Sunday...September 29 & 30.

Anyone interested...Please feel free to email me at [email protected] or email me your phone number and I can call you between 4:00 PM and 1:00 AM GMT (9 AM and 6 PM PDT) - Monday through Friday.

I have 5 sets...first come, first serve.

 

@Donald@HIDevolution are you paying my tickets from germany too?

Nice action from your side!

 

No, but if you pay for my tickets to Germany...I will bring the tickets to you!

 

hmmm, no sorry i need the money for the new rtx 2080 notebook version or maybe an new taptop

 

help desk information and device manager information about my bios

 

Looking at these RTX benches, MSI absolutely has to give the next generation of this laptop a 1440p 120Hz or 144Hz option.

I'm not interested in 4K yet, and the 2080 would be massive overkill and wasted at 1080p/120Hz.

 

Finally picked up a GT75 Titan 8RG-055. Does @Phoenix still do his magic for new laptops??

 

Phoenix is no longer existent.. the Person yes but Not the nickname.

Did you bought it from hidevolution?
If yes use @Ultra Male for further help.

 

He's not an HIDevolution Gaming Team member so the answer is No.

 

I was praying to find a 1440p/120Hz panel this generation. Nobody seems to want to make them. Probably cheaper for them to just put in a 4K panel since they are made for everything else.

 

Well there was a 1440/120Hz AUO1496 screen out there for a while, utilized by Clevo, Alienware, and others, but it had a color banding issue that caused many complaints and its use to be discontinued. Hopefully a high quality one pops up here soon. The 1080p/144Hz displays seems to be all the rage right now, judging by most laptops on sale.

High refresh rates are something I won't compromise for massive resolution. I'll go 4K in 3-5 years when 4K/120Hz is hopefully trivial.

 

Work in progress.

 

3200 mhz memory running in that BGAbook, eh?

 

Has anyone else done this before?

 

On the 16L13, I think so. Didn't @Mr. Fox get 3200 mhz RAM running on it with some mods?
Not on BGA garbage however. I know 2666 mhz memory works in the GT73VR (I've made Kingston 2400 mhz valueram run at 2666 mhz with extremely vomit luster timings). And I recall someone had 2800 mhz memory working in one of them.

To be honest, memory multipliers up to 15 exist in the Bios for the GT73VR. No idea if it works or not (doubt anyone would care anyway. Who's going to try to put extremely expensive 4000 mhz memory into a GT73VR?).

 

That is possible, i would have to double check.
4000Mhz?

 

Well the memory multipliers for it exist in the Bios.
Question is if they are actually functional up to there, if the IMC can handle such a speed on discarded bin BGA chips, and if it can train properly with "supported" RAM.

 

We are talking about MSI. Pretty much none of their notebooks recommend anything above 2666Mhz, let along it actually working.

 

I only saw one person who had 2800 mhz RAM working in the GT73VR in the main thread.
And since @hackness got his 2400 (!) mhz samsung RAM running at .......3000 mhz in his GS65 (that's not a typo), then it's definitely capable of it working in at least some brands. Doesn't matter if it's "supported" or not. MSI is run by a bunch of buffoons. Wonder if it's the actual engineers directing the shots or just....suits.. MSI doesn't "support" the laptops running at full AC unthrottled power if the battery is disconnected either, yet it's possible...

 

Only problem is the time it takes to find a working model.....Mine runs at 3460 Mhz stable, but 3600 Mhz is another story.

 

HID tech support tested out different faster ram sticks (as well tried different settings) for *Phoenix* when he purchased his first MSI the Gt73 and with no success. Maybe newer ram sticks would work better on the old models. I don’t know.

 

Back then, "Clear Cmos" trick for RAM timing fails were not known either.
And wasn't the Bios extremely horrible back then?

I had some RAM timing fails also when trying to tweak my timings. Sometimes it forces a manual clear cmos (60 second power button press on Titan laptops). sometimes it automatically reverts back to stable timings after a fail.

MSI default sticks (Kingston Valueram or MSI branded Vaueram) at 2400 mhz 17/17/39 2T will work at 2666 mhz at atrocious timings of 18/19/43 1T, even with all four dimms populated at 4x8 GB.

If this is possible then normal higher end sticks are definitely capable of being run now.

 

It would be interesting to go back and test it again with these newer sticks. That's TBD and yet to be tested.

 

I do remember that.

Well, looking at the picture, she is running 17-17-17-39-420-2T@3200 Mhz 16GB(4) and an I9. So maybe....someone is listening....

 

Not that advanced yet, I stopped at 2933MHz When I have more time I'll keep pushing!

But I might start again soon, since the new ThrottleStop (8.70) came out and it actually separates iGPU and SA, that is good news!

I always have the SA overvolted for 125mV which I believe can be some of help on the RAM Controller, but the problem in the old ThrottleStop (8.60) is that it was linked with the iGPU so I had to OV the iGPU back then.

Another thing is the Core iccMax is limited to 928 (232A) in the GS65 BIOS, but you can push further with the new ThrottleStop! Nice!

I have a strong feeling that there is this RAM lottery just like the silicon lottery on the CPU and the GPU. This is the lowest Timing my Samsung RAM would stay stable:



Note that with my current setup, I'll actually need tRFC set to 470 to not run into BSOD and tREFI to nearly stock level to not run into freezing problem (tREFI depends on how large the RAM is, the larger the more frequent the recharge is needed, mine is 16GB per stick).

By the way, I think Samsung RAMs are very capable of OC'ing.

 

Looking good!
So it passes TM5 & Memtest86 also? If so, that is awesome once again.

 

Yes! Wow you are actually running at 1.35V?

 

No. I'm actually running at 1.25V.

The factory setting for XMP is 1.35V

 

That is impressive! Currently, I'm also running at 1.25V to solve minor freezing issue at 2933MHz, I think with the same voltage, the MHz be pushed further on mine, can't wait to try again!

 

Should be interesting to see. I have ran 2400 Mhz at 3000 Mhz before, so it should be interesting. 1.2V

http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/page-573#post-10367270

 

Very good demonstration, but the tRFC timing seems to be pretty loose, was it only for testing if the 3000MHz runs?

 

Actually. To be perfectly honest. It was an accident. I had forgotten I had left the stick of 2400 Mhz Hyundai memory in. This was when trying to run 3k ram was super unstable. So you started out lower and worked your way up. Then a few bios revision later, 3k ram worked as intended.

 

That is a huge leap you did there, I now have more confidence of doing further testings with 1.25V after seeing the successful example of 1.2V @ 3000MHz, thanks!

 

Indeed.

This is just to show that my ram is stable at 3467

 

That is great!

I'm also working on mine now.

2400MHz -> 3200MHz = OK!!

I'm still tightening the Timings but... These Samsung RAMs are just crazy!

However I've noticed the timing becomes significant looser especially the tRFC going to 500+ on mine.

 

Nice! Samsung ram can do amazing things in the newer machines.

Only one thing... You need to run TM5 in administrator mode.
1: right click properties on TM5
2: click compatibility
3: put a check in "run the program as an administrator" box
4: click apply
Then reboot.



Then run it again.

 

Thanks! Good to know that!



It is on only doing 2Gb x 12 for some reasons though.

I've tried 133 x 13 however even with super loose timing and 1.3V it still won't boot, I've tried Odd number, becoming 3333MHz, it boots, but then I have problem loading Windows, I guess the maximum I can go with my GS65 + Stock Samsung RAM is only 3200MHz.

I've noticed that the tREFI can directly affect the reading and copying speed, however the stability is traded for that, I knew that when I was testing with 65535 on the 2666MHz a few months ago.
tREFI = 35188 has been on mine for 3 hours now, no unstable sign yet, I'll see in the next couple weeks if I encounter anything unstable.

 

Try upping the uncore voltage

 

I went to BIOS > Overclocking Performance Menu > Uncore and added +125 to Uncore offset with "+" for Uncore prefix, but in HWINFO64 the uncore still stay the same as before which is 6.922V.

And then I monitored the Uncore Wattage using HWMonitor, but I found the wattage to be the same.

Overclocking Feature is enabled too.

Hmmm... Am I doing it wrong?

 

Rule of thumb. Never assume all settings are accurate.

The way to fine out what is the uncore voltage is to Set your TS settings back to factory, but unlock the system agent, intel GPU and igpu. do not change anything else.
Exit out and then reboot to bios. You are going to need a bit more than +125. Closer to like 200 to 300. This should also help stabilize your memory, but going this route will probably have you relearning your memory timings for stability.
You kind of need to make these setting changes right after a cmos reset. So they train under the different voltage.
When you go back in to windows, open TS and see what changed. This will tell you which one of those is the uncore voltage. This seems to change over different revisions of TS.

Edit:
In TS you need to tick the 250Mv setting for SA

 

Ahh! I didn't know the Uncore in BIOS is the SA in ThrottleStop

Your hint strikes once again! Thanks!

2400 -> 3467 = OK!!

Super loose timing but at least it's running!

The tweaking queue is getting longer now!

 

That just depends on what version of TS you are running. Some times that is analog or digital io voltage setting.
Looking real nice. Now that is a worthy accomplishment to be proud of!! Not like the other ones before it weren't already though. LOL

 

Can't even rep @hackness anymore because I repped him too much already.

 

Yes, I have noticed in 8.70 the I/O voltage has gone missing. I didn't think this thoroughly, I should have apply the SA voltage since power on and not waited till Windows is loaded, good to know that SA is the uncore in the BIOS, thanks again

Thanks @Falkentyne ! Much appreciated! The RAM discussion we had few months ago in the GS65 thread has been really helpful!

 

Why is it that it's the "Uncore" voltage (System agent) that makes your RAM work, and not just increasing the DDR memory voltage?
I thought memory voltage was what made RAM run at higher speeds than normal?

 

That's a big 10/4 negative.

It's a combination of voltages and uncore is a pretty big part of that.

Edit:
Ummm, waiting on you to get back to the memory testing as a matter of fact.

 

I was thinking it might help the memory controller to deal with the higher clock rate

Seems stable now,

I've lowered the Uncore offset also,

No impact yet.

The tRFC is now going to 664 but the test is passed without any error or BSOD,

I've tried 640 and that was a NO GO with errors coming up in Testmem5.

Once I've tweaked to the tightest timings I'm going to lower the voltage and start raising tREFI for higher read speed.

 

@hackness
I'm not really sure why the test is not using all of your memory for testing. You are only using 24GB out of 32GB.
But still, that is a very nice accomplishment!

Can you get the latency any lower?