Presenting the All-New Intel 10th Gen MSI "GS66"!

You can be the Usain Bolt of the GS66 2070S, don't give it your all but just enough to be/beat WR. Then you can just continue to beat records.

Sent from my BLA-L09 using Tapatalk

 

@werdmonkey4321 did you do anything in regards to the TPL PL1 and PL2 limits for your GS66? Any changes in ThrottleStop or IMON Slope offset in BIOS?

 

I have not done any changes to those settings. Those are all stock.

 

Do any of you know how to cap the power limits for the GPU in Ubuntu 20.04 LTS? I tried nvidia-smi and greenwithenvy to lower the power limits, but it says its disabled when I used nvidia-smi -pl 70 and the power slider is greyed out in greenwithenvy. Another option is to lower the temperature at which the GPU thermal throttles at, but I don't see any options for doing that in Ubuntu or the MSI Advanced Overlocking menu in the bios.

 

You make a plateu in the voltage curve, at some voltage/clock point, lets say, you limit it at 1300Mhz at 0.725V and test if its good enough, if not, tweak it around a bit. Not sure if there is any way to change the voltage/freq curve on linux.

 

I know how to undervolt using MSI Afterburner in Windows to cap the power draw, but I just don't know if its possible in Ubuntu. I only need to limit the power draw in Ubuntu.

 

Werdmonkey4321, I think I successfully did the shunt but scores dont' get too high. Looking at the score results the GPU speed and memory speed seems a bit off: https://www.3dmark.com/3dm/50986882?
Any ideas? This was on 140 Core and 900 memory in afterburner.

 

That's a regular overclocked score. My guess is that you only made a cold joint.

This is mine for reference
https://www.3dmark.com/3dm/46827874

 

Crap. Well this is my solder: http://imgur.com/a/rktsNoI

Can't tell if it's a cold joint or something else.

 

Why is it so bloody tall

 

The solder wouldn't stick as easily otherwise. The tip on the weller was a bit thick and couldn't use the original set of tips I bought for another solder. Will try again to see if I can shorten it.

 

Ok made a bit shorter http://imgur.com/a/cRHqnht

 

Does it work? That's what's important

 

Your ~9000 graphics score is about right with only a 0.01ohm resistor. I did some mods after the fact and increased the power a slight bit more when I benchmarked to arrive at my results. I was also running +147 core, +1100 memory when I benched if I remember correctly.

Your overall score is quite a bit lower though and that has to do with your CPU score. My CPU has a pretty heavy undervolt on it(see previous posts). PL1, and PL2 are both at stock values of 65W and 125W, with short turbo time limit also at the stock setting of 56 seconds. Make sure all background applications are closed. And run the benchmarks in Discrete graphics mode for Time Spy. Also make sure your turbo ratios for the CPU are back to stock values.

Also how is your GPU getting an average temp of 49C? XD That is extremely cool.

 

No its not right.

Take note he has a 2080 super Max-q btw

 

Woops I didn't notice that. In that case his benchmarks are way too low.

 

Still working on getting the shunts working I guess but the 9400 score I think was from working shunts as now I can't get anything higher than 8400 with the same overclocks. I had the test done using coolerboost so maybe that's why the low temps? Also is it possible that I could have burnt the resistors somehow? I'm soldering at 610F

 

If it's working you should be getting 10000 graphics score minimum

 

Ok S I ran the tests again and this is what I got:

With +140 Core +900 Mem and throttlestop (-201mV Core/-110mV Cache) this is the results: https://www.3dmark.com/3dm/51009467?

With +140 Core +900 Mem and no throttlestop I get: https://www.3dmark.com/3dm/51008960?

 

Yeah thats a regular score for an overclocked 2080 super Max-q.

Shunt mod ain't working.

Btw, just focus on the gpu score.

 

Damn! My soldering skills need help. First time soldering so could be why. Guess it is just cold joints that I'm doing. Any tips or tutorials?

 

Let the iron heat up more time, use a lot more flux, remove all that solder castle and solder the new resistor as close to the original ones as well, we are talking about mOhms here, all that solder is most likely adding so much resistance that the parallel resistor isn't going anything due to the total resistance of the solder and resistor.

 

Forgot one very important thing. Set up hwinfo64 to show an OSD reading of the gpu power. Then run a gpu workload.

Tell me what reading you're getting.

 

Ok I will try to get it as close as possible. Should I try using a wire instead to connect them?

Also last time I used flux it was not giving me good contact only leaving a sticky film on the PCB but that could be because of the bad soldering station I got.

 

When using HWinfo I get readout close to 90W as max Wattage but depending on the test it is reading 86-89W this is under the GPU power line. Last test was giving me 88.967W

 

Would

Would it be possible for you to flash a 90w 2080SMQ bios with dynamic boost onto your 80w?I tried getting someone else my cards bios before but something wasn't working right so I'll need you to walk me through it but It will definitely be worth it for you.This thing's getting 10200 graphics in timespy now and I wanna see what it will score with that shunt.

 

So I'm not sure if the shunt actually worked for me since Seanwee pointed out I should be getting over 10,000 on timespy but i'm only getting 9400 at best. Wattage does show higher than 80W so I'm not sure if it is reporting error or it's because that's how it normally shows up as without the shunt. I didn't find nay suitable 90W vbios so I couldn't do the vbios flash. I think I had asked you in the GE66 owners thread for your vbios but gpu-z didn't allow you to get it.

 

It most likely the solder touching the mainboard isn't bonding correctly. Your iron is probably set to too low a temp. I had similar issues when I was doing these thing initially.

Just as an example with the tips I used and the iron I used the Hakko 888D, I had to heat up the iron to 645F. Then I touched the iron to board first to heat up the metal pads and touched solder to tip of the iron. If the pad is heated up properly, the solder should melt like butter onto the metal pad without any resistance. It will look like the solder wants to go onto the pad instead of you fighting to get it on there. That's how you know its heated up enough.

 

Ok great! I just didn't know how hot the iron should be and didn't want to burn the PCB. I will try to heat the pads on the PCB first then place the solder on the heated pad right? The way I did it was heat the pad then solder on the iron to let it flow onto the pad. It stuck but probably not good enough. Also on the resistors should I heat the pad then put the solder on the pad or how should I do that part? Thanks in advance!

 

For the temperature of the iron I would slowly increase the temperature though to figure out the best temp for your iron. I also don't want you to burn your PCB by setting too high a temp initially. XD

Yes for the resistors you want to do the same thing. Heat the pad, add on solder first. Once you have done this with both the resistors and pads on the PCB then you can start trying to join them together by putting the two next to each other and heating up their solder to connect them.

 

The way I had done it was heat the pads (probably not hot enough) then solder on the tip of the iron and put it on the pads. Same process with the resistors and for both I made them cone upwards. I then soldered the solder on the pcb and the resistor together. Problem was that sometimes the solder on the resistor would melt off the resistor due to gravity lol. I will try raising it to a higher temp as right now it is at 610F on my weller. Also using the tip that came with the weller which is not as fine as the tip on the yihua I had gotten but am returning.

 

So you should touch the solder to the iron while the iron is still touching the pads, like in this picture.


You shouldn't heat the pad, remove the iron, put solder on the iron, and then put it on the pad. I might be interpreting what you've written incorrectly, but I just wanted to make sure you aren't doing the latter.

 

Yes I'm heating the pads then putting the solder on the tip of the iron while it is still on the pad.

 

Yeah if the tdp shows up as 80/90w (depending on your vbios) then you're just at stock.

Also, what you want to track is the stable wattage under load, not the max wattage as 80w vbioses can spike up to 90w and 90w vbioses can spike up to 100+w

 

Got it! I will wait until I get a new tip and then try again with the soldering.

 

I think I got it to work.I originally tried uploading it to tech power up but i'm not sure if it went through..I did end up being able to save it as a TU104.rom file..Is that all I needed to do?Let me know if I should upload it to a site or email it.

 

Oh yea that's all I needed to test it out! If you can send a download link or email it to me. Just send me a dm. Thanks!

 

Alright just to clarify you can only flash Max-Q to-Max-Q right?If I wanted to try flashing a150w 2080 Super mobile.That wouldn't work on a Max-Q right?I get so frustrated seeing these low GPU temps but it's still power limited.

 

You can certainly try, but remember to look at the clockspeed and performance.

You'll know it's in safe mode if performance or clockspeeds go down

 

That's what everything I been reading points to..It will flash but it'll be stuck in safe mode with no GPU boost..I've seen a couple people posts about success with 2070SMP to the 2070SMQ but they never ended up following back up if it's working right.
I wish nividia could just release a warranty voiding flash at your own risk bios haha.

 

Ok so using the 90W vbios from klGE66 I was able to get a timespy score of 8766 up from 8466. Also I see my GPU power at 100W instead of topping out at 89W. New score is here: https://www.3dmark.com/3dm/51078904?

Going to see how much more it increases with some shunts properly done. Hopefully soon!

 

That sounds about right without a OC applied..I'm up to 9700+ consistently in timespy and 10200 graphics with my OC so once you dial it in it should be right up there...Really fine tuning that curve makes a huge difference.

Are you able to enable dynamic boost on it now?

 

Happy to see it worked and if anybody else wants the bios just let me know or I'll eventually upload it somewhere and post a link.

 

@klGE66, I'm interested. I also have the 2080SMQ and would like to try

 

I got you later today when I get home.DM me your email to send it to.

 

Looks like I don't have enough posts yet to make a PM lol

 

Here is the file from klGE66. 2080 Super MaxQ from his GE66 which is the 90W variant. https://drive.google.com/file/d/1nB79wR8ZzwG0xoeKAQJNELQcayViHZAZ/view?usp=drivesdk

 

Hey so anyone knows what happens if one of your R0005 becomes chipped? Yes I think I used too much heat soldering or was it that I left teh battery plugged in not sure but low and behold there was a small pop sound and I notice the R005 got a crack in it. System still boots and running a timespy test right now so see what I get as far as power draw is for wattage.

 

You might have killed it or just cracked its plastic housing. I've removed both shunt resistors on my zephyrus G and directly soldered across the pads without issue.

Btw if you guys need help quickly and easily you can join the msi discord. No stingy 2mb file size limits there.
https://discord.gg/TxnnB3V