Taming the Beast: 4930mx

This will mark my adventures with taming the 4930mx, a notoriously hot chip of its era, and the last rPGA CPU available to us.

For my particular platform, the methods available seem to be able to push the chip to 4.3Ghz. This is usually done with a 10.4-55cfm fan, the aftermarket 4 pipe copper heatsink and liquid metal.

I aim for 4.5Ghz

Yes, this will be a frankentop, no it will not be mobile, and yes its a DTR. No, I do not want a Desktop, and I have reasons for that. With that out of the way, I will share the ideas I have come up with for cooling the CPU.

Pictures to follow.

 

Pictures taken on my phone are beyond the 2MB limit, Here is a link to imgur for now.

https://imgur.com/a/KcKt4fh

 

No luck with getting the 4980HQ or some similar Crystal Well CPU working on your laptop?
Given that it seems to only need the micro-code to be present, it should give you better clocks, and better perf due to the extra L4 cache.

Also, is your motherboard VRM's up to the task?

 

No, crystal well never worked. We have injected crystal well microcode as well and still no post. I'm not a bios modder just a humble hobbyist sadly but if anyone wants to try their hand I can be a Guinea pig no problem lol.

I believe it's likely due to the lack of digital output but I can't be sure, as the repackaged 4980hq doesn't output digital signals. That alone shouldn't be an issue but perhaps there are other aspects of the cpu that aren't functional that are preventing it from posting. I'm happy with where it's at though, the iris pro helps out the t440p quite nicely.

I'm hoping if the cpu is cooler then the vrms should be OK as well, time will tell though

 

The t440p seems to have the QM87 chipset, while the AW has the HM87 chipset, but there isn't any massive difference and both should support the same CPU's..

What is the intel ME firmware version running on the t440p and what version are you running on the AW? Maybe its linked to either ME version, or ME firmware need a nudge using FITC(or maybe going nuclear and disable_me).

Also, nice work on the heatsink, if the VRM's are near, maybe solder a piece of cooper plate to provide extra cooling to the VRM inductors/MOSFETs.

 

Im not sure, I can check that when I get home tonight.

The problem here is I still need to make more adjustments, I want 4 heatpipes in a single row.

The bigger problem is that I have never soldered anything in my life lol and dont know anyone willing to take the task even if compensated. So either figure it out with caution (heat pipe is essentially a pipe bomb) or contact someone that is willing to take on the task.

Everything else I can prep myself. I guess I can ask my brother in law if he has used 63/37 solder paste before and how to work with it.

 

DONT use regular solder!!!

You need low temperature solder with a high bismuth content, something like chipquik that melts at 138°C, regular solder that melts at or over 220°C will result in broken/exploded heatpipes.

If the heatpipes are already covering the lenght of the die, unless you have a thick cooper "heatspreader" more heatpipes side by side wont help much.

 

@Khenglish had used that paste with his 1080 heatsink unless I read it incorrectly?

There is some of the heat spreader that doesn't have heatpipe contact, I can't really stack as that will also increase the height and becomes larger than the fan I plan to use. The idea is to mitigate burst power fluctuation.

I'll look into that paste you mention, does it have the same thermal conductivity as 63/37?

 

Its more or less the same paste used by the heatsink oems, they all use solder with high bismuth content, or they would burst the heatpipes.

If you want burst handling you need a thick cooper heatspreader to act as a thermal damper.

If it wasnt chipquik it was something similar and I have also used it with good results.

Lots of flux on all parts, plenty of solder paste, arrange everything and lock all in place with a jig, toss in an oven, heatsoak for a couple minutes ate 80°C, raise to 140°C, watch till solder melts, give it 10 seconds to flow/wet well, turn off heating and open the oven. Its easy, but try with a couple pieces of scrap to learn, and then off you go, Just make sure you dont use regular solder.

 

Did you use something like This?

 

Thats regular solder.
Its this range of products
https://www.chipquik.com/store/product_info.php?products_id=440003

It must have Bi in the composition and spell out low temp 138°C.

It seems like their current chip removal alloy melts at less than 90°C, you want the one that melts at 138

 

Those are all for removal? Those can be found as amazon as well.

I guess I am a bit lost on that front...

 

Check my edit, sorry.

Im linking the site because its easier for me.

 

No worries, soldering in general is uncharted territory for me.

I check this against the against the site and it appears to be the same?

https://www.amazon.com/Solder-Bi57-...aste+Lead-Free+chipquik&qid=1574710442&sr=8-5

Oh hey even @Mr. Fox left a review on that listing, small world.

 

I've always used 63/37 solder paste for heatpipes. Heatpipes do expand a little at 183C where they are not clamped, but they don't burst. Make sure they are clamped good at contact surfaces to avoid expansion so you don't lose contact surface area. An unleaded 217C paste probably will make the pipes burst.

Heatsinks definitely do come with a lower temp solder paste, probably 138C like Senso mentioned. I've learned how to work with 63/37 so I've just kept using it. I would recommend using a lower temp paste if you can get it.

 

Thats the correct one.

 

Thanks guys, your info has been invaluable. Now I feel like I have a path, may not get it right the first time but I can easily source more parts if needed. The main benefit of using a Dellienware system.

 

ha funny, i was actually thinking of Fox when u guys linked to that solder

 

I ordered that paste and it should be here today, just trying to get everything ready for when I try to get this all implemented.

Side note, Just got my AX50 router yesterday, with steam streaming I have no need for my AW 13 R3 anymore, will be getting that up for sale here soon. Going back to only rPGA systems for a while oddly enough.

 

Still waiting to hear if you got your pipes bent?...or straightened or exploded or whatever


I was giving serious consideration to the 4930mx....But KingofSwag pointed out the heat issues and i decided not to torture myself and just accepted a 4910mq.

 

Those are kind of pointless as yes you can achieve a 4.0ghz clock you can't sustain it at all.

 

4910MQ can be overclocked to 4.3GHz on all cores.

 

Im aware, however you are still limited to 47w long term power. It becomes a moot point.

 

Not if you have an unlocked BIOS that allows you to raise the power limit, or use ThrottleStop’s PowerCut feature.

 

If thats all that was needed, you wouldnt have microcode hacks for that generation.

Powercut isnt guaranteed to work on every system either.

I have had the unlocked BIOS for some time, it does not manipulate the CPU's microcode limits, it opens it up on the system side.

 

The 4910MQ has a partially unlocked multiplier (+6 Turbo bins). It doesn’t need ucode hack to overclock.

 

Again, I am perfectly aware of that. However You are still limited to 47w TDP for long term power, that is built into the CPU, not the motherboard.

Once you hit long term power limit it goes back to 47w unless I am missing something?

 

You’re talking about stock power limits. Obviously you don’t leave power limits at stock if you’re overclocking. Heck, I had a 4720HQ that could draw almost 100W in Prime95 Small FFTS w/AVX before the VRMs overheated:



If I limited PL1/PL2 to 65W, I could hold that indefinitely on the edge of overheating:

 

None of which was held at 4.3Ghz on a Ranger?

 

What does that have to do with it? You claimed Haswell couldn’t exceed the stock long term power limit, I proved otherwise.

 

You're welcome to discuss further in PM if you would like. I have reservations of your evidence.

 

Averaging 55W in a 4-min wPrime 1024M run:



Holding 55W while running Prime95 Small FFTs w/o AVX:

 

Cool.

 

Woah, what fan is that?

 

IIRC its off of the G703, people liked using them to replace the fans for their Tornado F5

 

That thing is beastly, might have to look into sourcing a couple of them. What's the height on them?

 

Thats Wicked sick!

 

If memory serves me right, 25mm

 

This was explained 5 years ago here, unfortunately the images are no longer served.
http://forum.notebookreview.com/threads/tdp-and-power-limiting-haswell.766743/

TLDR, in a lot of cases it's not only the BIOS firmware that may need addressing but also the Embedded Controller firmware which runs under it's own microcontroller, different to Intel. For my own laptop this firmware reads the SKU power of the CPU (TDP) then uses a lookup table to set PL1 and PL2 power and time via PECI (basically a backdoor to the processor). These settings are stored differently to the other PL1 and PL2 settings and arbitration takes place where lowest wins.

If the maximum limits are defined as unlimited (can be read using HWiNFO) then that's what they are. Microcodes while very powerful are not responsible for this.

Here's an example of an i7-4980HQ running an app called Firestarter to consume a constant 100W for 1 hour.
https://linustechtips.com/main/topi...mation-guide/?do=findComment&comment=12168073

"Taming the beast" is a great title but certainly not limited to only MX CPU's although technically the MX should be a top tier chip. Look forward to seeing what you can achieve.

 

There's realistically no reason to get a MX chip other than binning.

 

Unless proven in the ranger, it is required for overclocking.

Now, if anyone has further points they can share in pm otherwise I'd like for this to get back on topic if you all don't mind.

 

yeah so....how is the heatsink?
Got the pipes all bent? Pictures please

 

Ill be grinding that out tomorrow mid morning after the gym.

I have to grind the contact plate to accommodate the extra pipes.

This is all of course as long as I get the fan working first.

 

MX isn't required for overclocking, not in the slightest lmao

 

I found the thread you referred to,
http://forum.notebookreview.com/threads/msi-16l13-eurocom-tornado-f5-fan-modifications.810712/

The fan is AB07512HX26DB00, from what I can see it's 80x80x25mm? Very cool stuff.

 

Yes I believe that is the one

Ill start this whole process tomorrow, using the short holiday to get this all done. If plan A fails I still have a plan B which isnt as pretty but should still suffice.

 

???...pretty?......I think that went out the window long ago

 

The g703 fan now works with my cooling pad. Getting it to work was much simpler due to the wires following the same color scheme as desktop as another member had surmised could be the case.

Fan control also works as well.

Do note that the fan is powered externally.

Also the fan fits almost natively in reverse. However I want to flip that so I will be cutting off the mounts to allow that scenario today.

Now that I have been able to confirm this much the rest of the project can proceed.

Time to get something to eat!

 

One thing I forgot to add yesterday was that the CFM the replacement fan outputs is nuts. There is no 5v fan that even compares and now I can see why they are typically used in the larger machines.

I have been working on the chassis again yesterday for the majority of the day to accommodate the scenario. I did make a mistake and broke part of the chassis but I dont consider it a huge loss, just some additional compromise to the overall integrity that I was ideally hoping to avoid.

Today my brother in law is bringing his powered dremel as I have been doing what I have up to this point with a file and a power drill, which can hit higher speeds which should reduce potential damage to one of the heat exchangers I plan to trim down.

One question to you guys though @Papusan @Khenglish @Mr. Fox and any others experienced in these sort of things.

I have one contact plate that is completely copper, however it is very slightly bent. I did put it into a vice and tightened as much as possible, and I had also planned to attempt lapping it.

The question is this, I have two other standard contact plates (copper contact, surrounding material aluminum) that maintained integrity, would be a better idea to simply use those instead?

Thanks

I'll upload pictures later today hopefully, not sure how much I can complete today when family is coming over, we always do christmas on the eve.