Mobile Pascal TDP Tweaker Update and Feedback Thread

Thank you, there was some hex bolt that was attaching it.



But now the programmer is saying "unrecognized chip model"...
Oh boy what did i get myself into?

"Manufacturer ID: 0x0
Device ID: 0x0.

 

What chip-set model do you have? And did you connect the pins correctly?
@bloodhawk Have you seen "unrecognized chip model" before?

 

PM me a screenshot, of that window, and a close up of your chip along with how you are connecting the clip. (either here or over at OCN)

Nope, but with my Skypro i always selected the chip manually and then did a detect to be sure.

 

Hi guys ok here is what I have.

http://i.imgur.com/71tTIUy.jpg


I have the sop16/8 connector disconnected right now.
I have two clips, one with "5x2" (10 holes) arrangement and one with "4x2" (8 holes) arrangement.
I can't even get the 4x2 one to stay on the chip.
The 5.2 arrangement one has two extra holes that don't connect to the SOP16 because there are only 8 pins on it. But there are 10 holes.

Pin "1" is where the white dot is on the video card right? (the one closest to the X bracket screw)?

Someone in another picture said "pin 1" is the one FARTHEST AWAY from the white paint dot....that is confusing. Very confusing.

 

1 . Use the 4x2 Clip. I cant tell how the 5x2 is wired. If the need be, hold it down with one hand.

2. Please post a close up of the chip on the card. Positioning / chip type / placement, can vary depending on the brand / manufacturer. Its impossible to tell whats going on and what is what without looking at it in most cases.
Also a picture of how you are wiring everything up. (as in how everything is connected before you try to read/flash)

3. Dont panic.

 

I thought u you guys said that the sliders weren't working.

Sent from my SM-T560NU using Tapatalk

 

This is the 10 hole clip because i cant even get the 8 hole one to stay on even if im pressing down !!!

these are three pictures:

1)
http://i.imgur.com/pyI8BqA.jpg


2)
http://i.imgur.com/o77wrvS.jpg


3)
http://i.imgur.com/xXRobQ4.jpg


"Unrecognized chip model"
Manufacturer ID: 0x0
Device ID: 0x0

I have:
SPI flash
MXIC
MX25U8033E

selected in SKypro USB Programmer

 

What happens when you click the Detect button ?
Also a close up of the Chip and the Skypro Window.

 

@bloodhawk
the clip you told me to buy doesn't work
it's *TOO THICK*.
I can't even get a good lock on the chip because a resistor is blocking the clip from making contact.
It's IMPOSSIBLE to make it work.
The head is too thick.
it's being partially blocked by a resistor.
That "U745" resistor is preventing me from getting a solid connection.
Remember I already said what happens when I click detect. I get unknown chip, ID 0x0, device id 0x0.

I was able to get 0x8e and 0x800e or something similar by almost breaking the resistor, but it still said "Unknown chip"

The clip with the "10 holes" has a narrower ABS teeth end, so that is easier to go on, but even that doesn't work. same error.

Here are the pictures:






Please help. I'm frustrated and want to break something. I'm permanently disabled with a life threatening condition (i was violently ill yesterday). i have extremely bad back and stomach swelling and a metal bar in my spine and marfan syndrome and scoliosis. This is very, very, very hard for me!!!!

 

Here's the "10 hole" clip that is connected perfectly.
It can't possibly get any more perfect than this.
Please look at it.
Also, I looked at the connector. The "bottom 2" holes (pins 9 and 10) have the cable cut so they are not connected at all, so that's not needed.

Please take a look:




How can it get any more perfect than this?

I think I should just put the video card back in my laptop. I'm afraid I fried something.

 

Not sure why your clip is like that, but mine is thinner and was ordered from the exact same link -




This is what my Skypro Window looks like with the proper chip selected and "DETECTED" -



I really have no idea where you are finding those Chip ID's and such.

If the other clip fits then line up the red end with the pin 1 of the chip, for reference this is how Pin one can be identified -



Also are you very sure and can read the exact chip type on the chip it self? And that it says - MX25U8033E , not MX25U8032E??
I cannot make out what it says, since it has the weird crusty/waxy layer on it for some reason.


Finally if you really are not sure what you are doing, i would just stop and stick with stock. As mentioned earlier in this thread , this whole process is not for users who are not comfortable with tinkering with component level hard ware or do not have the exact understanding of exactly what is going on.
As you said yourself, might as well STOP right now, and save yourself future trouble.

@Coolane Do you know if the MSI 1070 has the MX25U8033E or the MX25U8032E??

I have a feeling its the latter one, and @Falkentyne is trying to go off what @Mr. Fox posted , which is valid for the Clevo 1080's. IIRC even the MSI 1080 has the MX25U8032E.

 

Thank you for the pics.
Yeah they sent me a clip that doesn't even fit.
Take a look.



the BOTTOM clip is the one they sent me (from the link you told me to buy from).
It's nothing like yours. It's WAY too thick.

The top clip is the one from the RT809F adapter I had in storage:
it's smaller and actually clips on the chip....

 

Yeah that clip is way too fat, i guess one can never trust these Chinese clips. Thats why i use these ones since @Johnksss@iBUYPOWER pointed me towards 'em - https://www.amazon.com/gp/product/B00HHH65T4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1

For the Clevo cards however either of those clips will work just fine.

I took a look at @Coolane 's image he posted quite a few pages back and seems like he is using a similar clip to the 5x2 you are using.
Good chance you are selecting the wrong chip type /model.

 

It should be a SOIC8-150 1.8v chip from the shape. Pin 1 is the top right corner if the card's MXM contacts are at the bottom. Make sure the connections correspond to what's going into the programmer.

@Coolane @Johnksss@iBUYPOWER should I try 151-200 but set to adjustable?

Sent from my SM-N910G using Tapatalk

 

Pin 1 is here according to your picture(this was also mentioned on few pages back), NOT the WHITE dot. The white dot is only to indicate that is the vBIOS chip-set.

My chip-set is MX25U8033E, the 1070 that @thegh0sts has I think it's a Winbond.

This is so true! @Falkentyne, if you don't feel comfortable of flashing it right now, maybe take a break for now.
No rush and take your time to order a better/compatible clip and do more research, so that you can feel more confident before you decide to do it again.
Put the disclaimer here again, :

 

Can you guys read this?


Looks like winbond 25Q80EW or something?
That for sure doesn't say "MXIC"...

 

We were talking about the temperature slider is not working. The TDP slider is working on both Inspector and Afterburner.

If you don't mind chance of burning your card, then go for it, .
If you care lifespan and safety more, then I would stayed at/below 170W.

 

I read an old post that said "pin 1 is in the top right corner if the card's MXM contacts are at the BOTTOM.

So I put the clip the other way and INSTANTLY when I tried to select MXIC, it instantly went to WINBOND right away.

So I hit AUTO and this happened.....



DOES THIS MEAN I CAN NOW REPASTE AND PUT THE MXM BACK IN THE LAPTOP ???

 

That's a Winbond chip. If you don't mind, you can scrape off the white dot paint gently to show the complete model name.

Edited: No need, I saw your model name showed by the software: W25Q80EW

 

@Coolane
I reversed the orientation of the clips after finding an old post about "pin 1 in the top right corner if the MXM contacts are at the bottom" (completely randomly) then I hit detect and it INSTANTLY detected the winbond. Even when I tried selecting the MXIC (on purpose) it went right back to the winbond instantly when I hit detect (no errors). I dont think I ever hit "Auto" so fast in my life. It flashed in 10 seconds and verified...am I good to go?

My 'original' attempts at pin 1 were actually correct.....just the clip (the one I ordered) was too large so it always errored on ID and then i started doing it backwards.

I had to use the RT809F clip. And then had to reverse the position again.


*EDIT*
THERE HAS TO BE CLARIFICATION for future users on PIN 1.
There is conflicting information in the first few pages.

Some posts say "pin 1 is marked by white paint".
I don't know if this applies to some cards but not others, but pin 1 was NOT where the white paint was on mine. Pin 1 was completely opposite (with a small indention in the chip, the MXM contacts on the bottom and pin 1 in the top right corner).

Are there people with pin 1 where the white paint is? Because that wasn't the case on mine...

Clarification on this can save a lot of frustration.

Now time to remove this CLAY PUTTY............(I guess that's better than Intel Pigeon Poop right?).

 

Did you load the vBIOS that you modded? I am not sure what the "Auto" does, does it automatically do "erasing, blank checking, programming, and verifying"? If it does, then I think you can put it back and see if it lighted up and showed any difference under load.

 

Dude seriously what are you talking about? All previous images clearly show that pin 1 seems to be on the bottom left. (Depending on how the card is pictured)
Please dont go about posting about what pin is what, without clear pictures or proper descriptors like the indentations that every SOP / SOIC chip has marking PIN1.
Vague pointers like will end up having someone burn/destroy their chip.

That is a Winbond chip. This is exactly why i asked multiple times for a close up of the chip itself. There are different chips used in different batches/versions of the GPU's and can vary from manufacturer to manufacturer/brand.

Please tell me you made a backup of the data on the chip before flashing the modded vBIOS?

FOR THE RECORD - PIN1 is the pin next to the tiny indentation on the chip. Doesnt matter where the weird paint/markings are. Always looks for the tiny dot / groove.

 

AND IT WORKS!!!

VALLEY BENCHMARK SITTING AT 1835 MHZ (+200 MHZ) with perfcap reason blank!
Thank you!!!!!!!!!!!

Yes I had all backups of the chip.
And yes I loaded the modded Bios right away. (remember I used programmers before, but only to flash monitors, not IC's).

(remember my blurbusters posts?)

Note: I modded the "8A" Bios, NOT the original Bios. Auto overclocking sucks...boost clocks locked at 1835 mhz instead of jumping over 2000 mhz in low loads is much better.

 

I was talking about the location of pin 1 not the actual vbios chip.

 

I

I'm gonna check a 1070 FE vBIOS to see what else is different to mine.

 

Yes that's what im referring to. Be more specific and use pictures point describing pin layout. And SPECIALLY dont not use vague terminology like "white paint".

 

NO POWER LIMIT!!!!!!!!!!

 

Either way they got it working now so

Looks like a 1070 FE vBIOS has the TDP as adjustable so I'll go ahead and make it adjustable but still keep the same 151-170W TDP limits.

That'll be the last flash I do for a while....unless @Coolane is able to crack the throttling temp issue.

 

GO FASTER! you can probably hit just under 2GHz.

 

Did you guys say the temperature slider isn't working?
I set my throttle temp to 70C in the Pascal Editor (instead of the 91C in the profile preset) and right now I turned off cooler boost fan so it's on auto, and the GPU temp is 64C right now and still at 1835 mhz, no throttling.
It was at 55C for an hour running Witcher 2 at the tutorial knight battle point (that was with the fans at 100%).

Decided to quit and turn off the laptop at 65C; Grizzly Kryonaut was doing too good of a job with automatic fans from letting it reach 70C. . Checked GPU-Z right before turning it off and no temp throttling whatsoever. The clocks were hard locked at 1835 mhz without any variation.

 

setting the temp doesn't work IIRC but @Coolane might know better.

You should now have equivalent or better performance to the desktop 1070

Enjoy!

 

I just made a clarification on the first post, hopefully it will help out other users of not making the same mistake.
Btw, glad that it works out for you! Cheer!

 

Thank you!
At +250 core, clocks are at 1886 mhz and not moving at all whatsoever. It's a complete solid 1886 mhz without moving down even 1 mhz.

Are others getting the same results (zero temp throttling of boost clocks?)
Remember I modded the 8A Bios, which disables "boost 3" (automatic overclocking past boost clocks), not the stock Bios.

 

I'm gonna flash my vBIOS one last time tonight....so much to do and so long to wait till i get home from work

Where's that vbios?

 

Maybe next time try to show a picture for clearer explanation. Connecting the clip in the wrong way with the pins will have chance killing the chip-set due to short circuit.

 

Adding 250 on core should give you over 2GHz, what is your default base clock and boost clock?
The 8A BIOS, is it a down-clocked vBIOS? Do you think you can show me a link of it?

 

I too want to see this vbios.

 

here's the 8A Bios.
Default clocks are the same as all others: 1443 mhz and I forgot the base memory clock, but I was +300 on it. (1443 mhz base, 1643 mhz boost).

+250 mhz overclock gives you 1693 mhz base core, with boost being 200 mhz higher than core, so 1886 mhz (however that adds up incorrectly). I was running +250 mhz core +300 mhz RAM.

going past 2 ghz with +250mhz core will only happen if you have "automatic overclocking" (boost 3) enabled. And that only happens in slow load situations. 8A disables automatic overclocking. Better to have stable locked clocks old school rather than the card change speeds by itself.

The 8A bios was released by MSI China to fix someone's problem (with the 1070's that were crashing past stock). The 8A Bios disables the automatic overclocking, where the card will overclock by itself past boost clocks. Example if you had +200 mhz overclock, in *low load* situations, the clocks would exceed 2050 mhz (!). This would cause people with "2016 year" 1070s' to crash.

2017 year 1070's would not crash, but instead would throttle due to "Voltage reliability" (VREL) instead.

It's unknown whether the 'bad' 1070's were using an older revision GPU or were using the "Micron" memory that desktop 1070's also suffered from, rather than the stable samsung memory (but this is another topic).

There were two Bioses MSI released to fix the "2016" crashing" 1070's: the "8a" Bios, which disables only "automatic overclocking" past boost clocks, and another Bios (44? 4b? I forgot), which everyone hates, because that one disables boost clocks also. Many people with fully working 1070's tried the 8A Bios anyway and preferred it, because the card seemed to throttle less and had less variation in clocks compared to the stock Bios.

With 151-170W TDP, and 8A Bios, I never once exceeded 100% TDP and the core clocks never, ever dropped past 1886 mhz (I did keep the temps below 70C, which was the temp throttle point I set in the mod).

I have no idea if you guys are getting temp throttling in the stock Bios. You guys aren't keeping your core clocks 100% of the time, after raising the TDP?

 

Thanks for the share!

 

Thanks for sharing the vBIOS.
I am using the "86.04.56.00.3A" vBIOS, which is the normal version, it throttles the dynamic boost clocks (auto-overclocking) like others once the temperature has reached about 46C. Even though the card can boost to 2076MHz(+189MHz) at the beginning, it throttles to 2025MHz when it reached 62C.
Maybe you don't see temperature throttling because your vBIOS has auto-overclocking disabled.
Do you think you can run a test for me? Try to add 400MHz on your core and see if your card can still sustain 2036~2038MHz when it reached 65C.
If it does sustain, I think I will switch to using the "8A" vBIOS.

 

Good idea. I plan to flash it tonight.

 

@Coolane I can't do that.
As soon as I hit "Apply" in afterburner at 400 mhz, the display driver crashed on the desktop, before I could even load a game. It crashed instantly.

When I did +350 mhz, Valley benchmark crashed instantly.
+250 mhz is fully stable.

 

Probably means you've hit the limit of your GPU then.

I'll give it a try but I don't intend to push an OC more than +150.

But that is ok since the TDP boost effectively means you have a desktop 1070. Any OC would make it a 1070TI of sorts.

 

Tried the 8A vBIOS and I got a black screen. @Coolane, since you have the same laptop you might get the same result.

But then you have a v1.0 board and I have a .0b board so there might be a chance it could work for you.

 

The 8A Vbios is a MSI Bios and supports Gsync. Is your card compatible with MSI Bioses?
This was the original Bios from my card (3A): does this also black screen?

 

My card is an MSI but the motherboard doesn't support gsync from the GPU...only from the 980DT that came with it.

 

Sounds to me that by using a vBIOS that disables Boost 3.0 that you're missing out on performance. Nothing wrong with seeing VREL as a throttling reason in GPUz - my desktop runs permanently like that - I think it just means that it's using the max voltage it's allowed to for any given temperature and frequency when you see that message - i.e. it's saying it's limited by voltage for that frequency & temperature. (On desktop cards increasing the voltage slider allows GPU boost to sustain higher clocks at higher temperatures through the Boost feature, even at the same core offset overclock). I think you would have more performance with Boost feature enabled.

 

I compared already, the 8A Bios (boost 3.0 disabled) and the stock Bios (boost automatic overclocking enabled). In Valley benchmark. There was a difference of 30 points between the two Bioses.
30 points is not performance. It's margin of error. When I saw I was getting no performance difference between disabling boost 3.0 (NOT boost clocks, but automatic overclocking) and enabling it, I decided to stick with the 8A Bios. I don't remember the score I got but it was in the low 3000's. From faint memory, it was like 3360 for 8A Bios and 3374 for 3A (this was like a month ago. Both were set to +250 mhz core, +300 mhz memory). No difference in Valley overall.

So I decided to mod the 8A instead of the stock, and I'm happy I did that. there's zero thermal throttling on the boost clocks (70C limit). It looks like the thermal "Pascal" throttling still affects the automatic overclocking when not at TDP, so I'm glad I don't have to deal with that).

I did a 4.5 ghz run with the 8A Mod, at 1886 mhz fixed clocks and scored 3800. I'm happy with that.

When I was using the stock Bios, the clocks would go up to 2000 mhz+ if I was overclocking by 200mhz+. However this happened when *GPU UTILIZATION* was extremely low, like 40%+ etc (on intermission blank screens and CPU limited screens, where the GPU was still needed). This doesn't affect your fps if the GPU is waiting for the CPU because there's no work to do. And what is the point of getting 2500 FPS when you close valley (close screen image), at 2050 mhz and 33% GPU utilization? How does that help anything?

Whenever GPU was fully loaded (100%), both Bioses were throttling to the same point (Pwr Limit) and would never go past boost clocks.

I'd rather have fixed stable clocks than clocks going all over the place.
Now in 60 FPS locked games like Dark Souls, none of that matters. The card won't even go into boost clocks because the load is so low (stays at 1443 mhz, and any FPS dropping under 60 (e.g. Ice cavern area is from hard CPU limitations).

*edit* wrong benchmark. he tested heaven. I got 35.4 min 209.5 max and a score of 2546 in heaven vs his 2616 (no idea what his 5820k was clocked at). And that's a 6 core 12 thread cpu (no idea if heaven can use that).

Believe me, I did my homework before making my decision. I always do.
(Learn to swim @Papusan @Phoenix ).

 

Ah, if you were hitting your Pwr Limit then if you could remove that & as long as you're not having temperature issues, THEN gpu boost might help, but I agree if you're hitting power limits at 100% GPU utilisation in most games with either BIOS, then it's all a wash really.

 

I see. Thanks for testing it.
What is your default voltage under load, is it still 1.05~1.062V?
Normally a 1070 can be stable at 2GHz+ with 1.062V, some can do with only 1.000V.