ghoul

yes, that's it.

Don't put liquid metal there But I think any insulation tape should be ok.

I didn't used any on that, just don't use whole 1g for 1 cpu and you will be ok.

Yes, isolation of capacitors is necessary with mobile device, since you move it a lot.  I used quick glue on capacitors, but consider it cannot be revert.

On 2017. 05. 03. at 1:09 PM, TheReciever said:

Here are the results of my fan swap. Hope you guys find my results well...

 

You will need excel or some type of excel viewer to view this data in its current form.

 

@ghoul Were your findings similar in nature when you had the 4700mq?

 

 

Conclusions.zip

HI

Can't say for sure, one thing, which is the most important difference, use metal TIM, like CLU.

However, nice test, lots of data, excellent job. if we could measure fan speed, or pwm percentage of it, it could further explain a lot of things.

Higher average, lower max, but as I suspect, lower fan speeed overall. Can you please share pictures of final fan mod, just for the audience?

20 hours ago, TheReciever said:

Usually people refer to IGP as integrated from my experience so I guess thats were communication didnt line up.

 

I dont think it will be able to draw 170w with just the CPU and motherboard though so I am not concerned about that, whether we have the ability to do so, is what we want to figure out lol

Yes, I'm aware that, but since you cannot enable IGPU with ultrabay gpu inserted, it's pretty straight what happens with IGPU when ultrabay is present. And since we are talking about CPU it wasn't a far shot

19 hours ago, Zwierzak said:

TheReciver as I mentioned earlier I could falsify TDP CPU indication on my current mobo with powercut option from throttlestop and with addition of manually setting VCCIN around 2 volts but basing on the experience with this mobo I think that EDP Current limit is either set by EC or hard-coded into voltage controller and right around the corner - maybe well reach 4ghz stable on around 60 watts. I should be able to exchange new motherboard for proper one (if they have them) in a day or two and maybe ill check the previous one tomorrow so we'll see.

Yes you can do that, but with that you can only fake TDP calculation, current of IA rail, (cpu cores and uncore) which is EDP limit in Limit Reasons stays. Read previous posts from me, IA limit is below default, so you will never unleash MX cpu in this machine even on stock clocks. EC overwrites everything, even if you set current to max, which option is present in unlocked bios. Which is to be honest is not that bad, since if you would be able to, VRM would have been smoked up. It is not designed for that current, and cooling of it is nonexistent.

14 minutes ago, TheReciever said:

Oh no, I was confirming onboard GPU not the ultrabay 755m

 

But thanks!

 

Im curious if it will behave differently when there isnt as much draw from the onboard 755m. I am not sure if it can be disabled to the point of no power draw but it should certainly be reduced. Temps right now for me are about 72c max load but I also noticed that the fan curve has priority on the GPU. Meaning CPU full load = little increase in speed, GPU load increase = fan speed increase to max allowed on curve.

using SLI, meaning with ultrabay inserted and onbard gpu disabled....

CPU and onboard nv gpu is using different power rails, those are not affecting each other, other than common heatpipe.

If you are reaching dc input max, then battery will be drawn, and after that, it will shut itself down, to prevent failure, but, normally you will not reach that.

1 hour ago, TheReciever said:

This was with the on board GPU enabled correct?

 

It stands to reason it may behave differently with the onboard GPU disabled, or you think that is flawed logic?

 

Thanks for your info ghoul!

No, it did not matter, using sli or not. Why would it matter? in case using iGPU, TDP will remain, and some percent of TDP will be eaten up by GPU, so without SLI, you would be in even worse situation. Current limit is separate between rails, but TDP is not.

You are welcome.

18 hours ago, TheReciever said:

@ghoul should be able to settle this then.

 

Was an unlocked BIOS vital to CPU overclocking the 4930mx when you had it?

 

Thanks

Hi

Like I said before, any TDP related settings were ineffective, even tuned down, TDP was at CPU default,  IA current limit was below cpu default, so EC protects VRM circuit, by limiting current.

in XTU any related setting were ineffective as well, ratio could be set, and voltage offsets.

4930MX if you can deal with the heat is way better than 4900MQ because of higher TDP.

You don't need to repair motherboard, if you entered something, which made you unable to POST, just disassemble it, and remove cpu from socket and back, it will reset settings.

One example is setting offset in bios, that will make it semi bricked.

Been able to game with undervolted 4G, but stressing it will reach current and TDP limit, so final ratio will be around 3.7-3.8GHz.

Larger diameter. Maybe it can fit as well, but it will be louder, like the one in ultrabay, fan is too close to shroud.

20 hours ago, TheReciever said:

Sweet, I figured with how locked this system is out of the box that it would likely disable the SATA port if Ultrabay was populated. I dont have an extra drive right now so I cant test it myself.

 

@ghoul

Hmm quite the hurdle.

....

Ultrabay fan is weak even at max speed, but, you can fit original main fan to ultrabay, lower pitch than y500 fan, which is stronger by the way, but it is still enough for 1 gpu, and noise is less.

I think it held very well, regarding the size of heatsink, and vrm overload, but that is out if its league.

47W but larger cpu could work, since under gaming load, it can boost up to 4930 speeds, but under heavy load, tdp will keep it cool ( compared to 57W)

I bought is from local marketplace, it was around 200EUR, and sold it around 170... ebay is waaaay overpriced. 4200m was arround 40EUR...  a 4340m slipped out of my hand, and that was even a bit cheaper. that would have been the perfect cpu for me.

Max temps will be only lower if you increase disspipation space on  other side of pipe, like larger fins, increased airflow, lowered ambient temps.

And of ocurse, @skavi use CLU, and with that less is not always the better, apply it to chip and heatsink as well, and increase pressure on chip via washers.

I don't have to mention, when using clu, insulate smds beside the chip, do I?

@TheReciever

First of all forget 4.5 even large AW machines are unable to run them at that speed, heat and power is too much.

4.2 with AW, but never gonna happen in this machine for multiple reasons.

First of all 4910MQ is 47W cpu. 4930MX's deafult IA max current is 95A and Lenovo bios sets it 86 if i remember correctly. Evven with unlocked bios, tried to increase it, but nothing happened.

So did the cTDP failed to work, no matter, how i set it in bios, MSR as set according, and XTU was able to read those settings, but throttling occured way before that, porbably,because of current limit, which is programmed by EC after bios init, so those settings are overwritten.

With undervoltin, was able to run Throttlestop bench @57W TDP throttling, cpu speed was hovering between 37 and 38 multi.

Don't expect more from Lenovo, and with 47W cpu, it will be even lower.

With more current usage,like AVX load, speeds where down as 3.0 with 4700MQ, didn't tried that with 4930mx, probably you will understand why

Lenovo repair manual for motherboard states clear, 47W cpu max, use cooling on vrm, and expect it to fail over time And remove battery, when you tempering around VRM, those are under power ,and risking shortage( been there, done that... tried to move thermal pad over with screwdriver, motherboard almost died, had to repair it)

What spacer are you trying to remove? If you want more pressure on chip, use washer mod, copper shim will introduce another layer of thermal interface between chip and heatsink.

Regarding extra heatsink... For heatplate( let's call it that), as you can see, you can add copper sink, but, this is a two sided weapon. You will gain, for being more thermally capacitive( English not my native, excuse me if it is unenglish ), and heat from chip will be absorbed by copper heatsink as well, delaying temp increase nad flattening down heatspikes.  But, downside is, the are dissipating to chassi, so eventually, when final temps are equalized, temps will be as high as without them. Cooling down will be slowed as well, so only reason to use them is for this.

But, putting those on heatpipe in the chassis, will make it even worse, because it will downgrade the efficiency of heatpipe, since not all heat will absorb outside the chassis, but part of it inside. Strongly unadvised.

Cutting chassis,so fan will be able to breath more fresh cold air is a better approach. But don't forget, VRM will be hotter,since airflow over it will be lower.

Those outside sucker are as good to practice throwing to dumpster. Use brown wire cut switch, and it will force fans to run at max speed.

Sata can be connected via flexy cable, so 1 m.2 and 2 sata devices can be inserted, if you strip sata ssds, 2 can fit on top of each other at original location.

Almost forgot, fan is XN0G5 part number , AW m14 R2 fan. 4 pins inner two pins have to be switched, and that's it.

Edit, did forgot something:

Even with original 4700mq downvolted, inner power cable at motherboard side was almost burned. Switched out, but when used with 4930mx motherboard pins became so hot it occured after two weeks or so. If you want to use it for folding, replace power cable and solder it onto motherboard. I used 2.5 mm2 speaker cables, which was a pain in the ass to route under left speaker, and getting motherboard out became harder as well.

@Tesla That's why I reconsidered my opinion about, if is it a mod,or not. However, for most of us, who use it as a notebook, it is unusable.

There are plenty of mods, which can be done to maintain lower temps, keeping the device usable as a portable one like chassis cut, dust filter remove, CLU repaste, heatsink retention washer mod, fan replace mod, fan speed switch mod. Personally i think downclocking anything below rated speed should not be a solution for engineering mistakes.

Yes, it was a very funny experience, hence larger notebooks struggle to run that little inferno too. Since cTDP is not working, even to lower below rated TDP, so I say, don't do it, use a 47W cpu like 4800MQ, since VRM is running on ( over) the edge with 57W, so do the cooling. I'm not sad, I use it very rarely for gaming nowadays, so silent and cool browsing, general usage is more important for me than raw speed.

@Tesla You don't need to be sorry, no harm was done,just wanted to point it out, that's all.

As a matter of fact, removing keyboard is more of a mod, than a repaste job or downclocking.

TP was meant to be used, so if we want to be correct, not using it should count as one, even if it wouldn't be effective (like a removed keyboard )

Thanks, and it is just a 6.4CFM fan, and there is one in this size with around 11CFM, now that should do the job properly (with a plane taking off like sound)

And if you want a small desktop, build an ITX machine, maybe with a huge passive heatsink

Before someone asks for 4930MX benches, a had to sell that cpu, currently running an i5 4200M, stressing it is around 60C with fan barely ramping up to audible speed.

This is not a mod, this is just underclock. Title couldn't be more misleading....

Let me show you a real mod:

Cut out original fan

Cut out AW fan, and glue it into original fan shroud

Other side:

Valley stress test:

Valley + TS bench

No laptop cooler,or anything, just slightly elevated chassis

With this, been able to use 4930MX@4Ghz for gaming...

Disassemble laptop, remove cpu from socket, reassemble and bios will reset itself.

16 hours ago, octiceps said:

Hey @ghoul, have you noticed the overall system power limit? It seems the board is limited to a max of ~140W, pulling 155W-165W from the wall depending on PSU efficiency. Any more than that, and it will drain battery. Try running TS Bench 1024M and Heaven/Valley at the same time with your CPU and GPU's overclocked, you'll probably see it.

Hey,

Not really,because i rarely push HW that far,and by default i use undervolt on cpu/gpu even under oc.  But it can happen easily, if you overvolt gpu-s and cpu is overclocked, not to mention, with 58W cpu it can be easliy reproduced, and i read about this somewhere, but that was with different modell.

When tried  once 1.3G gpu clock with 1.2V on gpu-s with original 170W psu, overcurrent was tripped sometimes, but probably it drained battery at that time as well.

You can try larger dell psu, or parallel 2x 120W or another 170W lenovo one.

However, be really careful, and check internal DC cable connection on motherboard side, because now for me the second one got burned. ( 58W cpu is not really supported, despite it works fine, first was burned with 4700mq ) Pin is ok on motherboard side, solder seems ok, but the female pin in connector gets really hot, and burns plastic around. Easy fix is ( wich i did already) is to direct solder a thicker cable, but  notice, there is very little place left to route cable under speaker.

On 2016. 09. 04. at 9:02 AM, daicharade1 said:

Ok so I could run Valley at your clocks and voltages but not games. Especially Rise of the Tomb Raider. What voltages you run yer setup at while gaming and what games?

These are my only voltages, i use them for games as well. Tried ROTR, it was stable. Could run valley with even lower voltages on ultrabay, but to be game stable it had to be raised to 1.05V.

Every chip is different, mine does lower volts than yours, but for exchange, memory oc is crap. Beyond +100 it is doing texture errors. Nothing guarantees that,even if they are the same ASIC quailty, if mine does certain speed at a voltage, yours can too,

dupe

10 hours ago, daicharade1 said:

 

Sorry for my noobness.. What do you mean fused? Why testing without SLI? 

Means default voltage for chip set by nvidia. Because there you can test cards one by one, and one doesn't affect others stability.

22 hours ago, daicharade1 said:

How come your default voltage is 1.05 and 1.08? I used svl7's bios and vbios unlock and my voltages are 1.1V (750m SLI, 1058,1250). My Asic quality is 75.4 and 89.1 yet it crashes out on yer clocks and voltages

Becasue those are the fused ones, before modded bios. Of course, after modded bios, everyone will have 1.1V, unless you extract bios files ( 1 for sli, 1 for optimus) and mod voltages further.

Regarding stability at certain voltage, every gpu is unique, respond different at different voltage levels, guess i was lucky with mine.

Test them out one by one, don't use sli for undervolt testing.

On 2016. 08. 26. at 5:05 PM, daicharade1 said:

Did you replace both fans? Your temperatures are good but your GPU memory isn't loaded to its full? What is your ambient temperature, and temperatures while gaming? And what is your ASIC quality for both the cards? What drivers you using?

Scroll up a bit, you can find almost all the answers there. For the lazy ones:

Yes, both fan was upgraded. No,because GPU load matters more,than saturating  memory bus and GPU is waiting for that. Ambient was around 24C. Temps under game depends on the game, this is why i choosed this test, because it can be easily reproduced, hence comparable.  Asics: 84.6 Internal, 76,6 ultrabay. Driver was upgraded since then to latest, so i dont know, but don't think it matters at all, unless you use ancient drivers.

Temps nowadays skyrocketing, so does cpu speed, switched to 4930MX, GTA5 can be played @4GHz, with thermals 90C +/-2C top ( depends on core), gpus are chilling, since fan speed is higher because of cpu temp Synthetic test is forgettable, if I want keep machine alive, but gaming is fine, so it is ok.

@Bobbey: Doesn't get that hot? What's the difference between 750m and 755m? First the memory is running higher, 750M uses 1250Mhz binned Hynix with 1.5V voltage. 755M uses 1500Mhz binned same Hynix with same 1.5V running @1350Mhz. Thermal differences between the two is close to none. Mine is running only 1300Mhz, gpu is ok,but memory is shitty like hell.

750M runs @1058Mhz boost clock, 755M runs at 1097Mhz.

The gpu voltage is based on leakage( ASIC quality), and you can get a 750m which runs at higher voltage than 755m. Thermals between few watts only.

And limited by TDP as well, mine has like over 100W/gpu in bios( actual is way lower of course )

Please note, my gpus have 1.05 and 1.08 voltage as default, running at 1.0 and 1.05 , so gpus are undervolted, and besides that, overclocked beyond 755m speed.

Cpu is overclocked with the 2 extra bins, which is present in 4700mq, and undervolted via -85mV and -50mV cache.

Memory is faster than stock 1600Mhz, they are running at 1866.

So if you had instability, that has nothing to do about being a 755m modell and running things at default speeds could  be done with even better temps.

Simply you have undervolted too much,because of thermal reasons, and/or you are not able to undervolt more, but cooling can not cope with heat.

This is what this and previous hw mods are done for, so you can use the performance present in the hw, and don't have to do underclock, which is unacceptable for me.

If you are using underclock for like longer battery life, that is ok, I can accept that, but to do it, because cooling is messed by lenovo, well..., not so much.

Lucky for me, i could do a workaround for their mistakes and now happy with my laptop, more than ever.

4 hours ago, Tesla said:

@ghoul Oh wow. I'm impressed. now that is what you call a mod. literally lol.

I aint a fan of under clocking either. I have the single gpu model, so every Mhz counts.

Thanks, pictures has been uploaded to local forum, here: http://prohardver.hu/tema/lenovo_ideapad_y400_y500/hsz_14270-14270.html

On 2016. 07. 16. at 2:54 PM, Bobbey said:

Hello Guys,

 

undervolt and underclock....

Really? Undervolt is ok, but underclock... Come on.

DO heatsink, mounting mechanism mod, do fan change, do metal thermal paste mount.

If you are willing to keep warranty,and performance as well, sell it and buy one with proper cooling.

Recently installed an AW m14 R2 fan ( only the 6.4CFM version, lame ebay seller sent wrong one) but still  stronger than original.

Had to cut out CPU fan motor and mounting legs of motor from heatsink ( round hole left after)

Cut the AW fan out from its chassis, keeping a bit longer legs cut out from plastic.

Now easily insert fan from keyboard side into heatsink shroud, place it in position and glue tha fan to shroud.

Now, switch middle to pins in the fan connector and thats it.

This is with valley running paused at scene where gpus loaded properly, cpu is locked to 3.4G so lenovo throttling is skipped.

This is with stock fan control, no outer cooling aid, just built in fans with slightly elevated chassis.

Now lets see what happens if i run TSBench beside Valley, to load cpu as well to its max.

Now, that is what I call a near final solution.

11 hours ago, octiceps said:

If anyone has purchased this, can you post a review? Thanks.

 

Edit: Also does this work with the laptop internal screen, or only external monitor connected to the GPU? And what's that 20-pin connector, is it for powering the adapter?

I would, but 60 EUR + shipping for a pcb with 3 connectors is a bit pricy for my taste. Actually I'm waiting for a dual pcb version, where it is connected via flexi cable. One for getting an external pcie 8x connector on the side of the notebook, and one for card to be able to build external gpu case with psu. Then it is more than welcome, building one with a ref 980 or something like that.

I think selling ultrabay gpu was a mistake, but if you can build egpu, than probably ok.

Egpu over the wifi cards slot... well,if m2 slot unused, i would use that, or install wireless card there, without it, you will loose mobility.

Perfect solution would be using ultrabay pcie connector which is regular pcie3 X8(custom connector of course), there are schematics online, if you can build the adapter, then it should work, ultrabay gpu is using standalone bios.

About cooling, if you can reach something like this, then probably you maxed out cooling efficiency( without fan speed , see below)

This is with default fan curve,  no external cooling aid, 25+ ambient, no stands. Valley freezed at a frame where both gpu loaded near full, and 20 minutes heatup. Cpu is locked with Throttlestop, and overclocked, undervolted.

I was able to reach over 4K firestrike with this( check HOF for 4700MQ+750M ), but of course that is just bench speed, i use it like in the screenshot, mostly because i hate fan noise.

PCIE -ultrabay adapter would be fun, both ways, i could use my ultrabay gpu as physx card in my main computer