Possible mxm rx480 card on Clevo laptops

Hello guys. I'm from China and own a Clevo p15sma laptop.
Currently, a factory in China is selling the mxm rx480 video card, which is built with standard MXM3.0B specifications. I know there is no official AMD Polaris cards for laptop at this moment. According to the manufacturer, a desktop rx480 gpu core was used and built into mxm form. I dont know why they built this and where did the card suppose to go. But it makes a very interesting video card upgrade options for Clevo laptops. Anyway, the price is very competitive, only 2000 rmb (roughly 300$) for a mxm version rx480.



However, according to the users who have tested the card on their p870dma and p770dm, it would shut down after 1min power on with fans spins at 100%. However, gpu-z can recognize the card, even the temperature was sucessfully read.



The poweroff phenomenon after 1min was similar to what would happen if not mxm card was installed. It looks like the EC shutting system down for some reason. And I came across this thread. “Telling the EC not to react when it receives no signal at all...” is a possible workaround for this issue. Wondering is the solution possible for my beautiful p15sma so I can have a taste on the amd card. (Already upgrade from the 870m to 980m for about 370$)

Further information of the mxm rx480 card can be found here.(Sorry for it is in Chinese)

 

You could see if @Prema could work on it with a donation. It would need some fairly low level modding even if it is possible.

 

That means the internal thermal sensor has been build as to Dell/Alienware values and not Clevo's AMD config. That's a hardware and not vBIOS problem...

 

So does that mean no RX 480/E9550 alternative for older Clevo machines, right?

 

Any details about the Clevo's configuration? Is it possible to have a workaround on the issue

 

I'm really curious about this. For Linux purposes.

 

There was a workaround for the EC shutdown issue on the MSI 1070 to get it working in Clevos.

I want it to be available for competition sake. A $300 MXM 3.0b card that performs about the same as a 1060 is good news for everyone.

 

For 299? Where can I find these people if they get this working on barebones laptops like the Clevo P775DM1 that would be great for the DIY market, because jeez look at the prices for the GTX 1060 MSI MXM, it's pure wallet torture.

 

I'll believe it when I see one that someone actually got running in a Clevo machine.

 

Not so easy for Clevo barebones. Besides the EC shutdown issue, the EDID of the internal screen should be also writted in the vBIOS. Someone sucessfully got this card running on a 16f1 laptop with external screen (internal screen would not work due to the EDID problem). Anyhow, I have ordered this card for testing on my p15sma, and it should arrive in the next week. Hoping all things would go well

 

Definitely post when you get it installed, I'll be interested to see how it goes.

 

I want a full blown AMD notebook with Rayzen and rx480 !

 

Someone tested on old alienware 17 before and it did not work.
Should be like what prema said, the card was not configured for laptop. It was originally created for VR ITX anyways.

 

No other news about a possible workaround for the Rx 480 MXM?

 

Due to the low end nature of the chip you are going to struggle to find a lot of interest in getting it to work.

 

Is it not possible to get it working like what people did with the Dell cards in Clevo's?

 

That was at least another notebook they were coming from.

These things are basically SFF desktop cards.

 

What's the main holdup here? Power draw? insufficient cooling? Lack of support for LVDS or eDP?

 

Just a stab in the dark: I imagine it's a financial viability issue.

 

As in there would not be enough buyers interested in the end product to justify development?

 

To justify development, integration, QC, etc. I've thought it to be a lack of demand & financial gain, but it may very well be more complicated than that. Again: Total stab in the dark.

 

The listing has been on taobao for months now. There's been a few threads on baidu of guys in China trying it. One in a p750zm, it boots through to Windows ok but triggers the Clevo auto shutdown thing.

As for a workaround to that I've not explicitly heard anyone say it's impossible... You're right there are plenty of non Clevo cards that don't trigger the auto shutdown (e.g. all the 1060s and 1070s going into Clevos so far)

I just think noone has really tried to get it working as despite the price it's 1070 power draw for 1060 performance... and with Vega due this year I doubt anyone will bother now

 

understood, I appreciate your 2 cents on the matter

this is a good point but considering it's the full fat rx480 which is like 90% performance of the 1070m(which Aetna sells for like $800 a piece according to the quote I got from them) I would think there would have been more interest from Consumers(and from AMD)

 

i guess this is a pass then!

 

That's pretty much what all the high end notebook manufacturers felt by the looks of things.

 

Too bad though

 

Yes, hopefully AMD can come back with a better GPU that can match up properly.

 

How likely do you think it is for a mobile Vega to show up later in the development cycle?

 

there is an amd rx570 in this samsung odyssey 17
http://forum.notebookreview.com/threads/samsung-odyssey-17-with-gtx1060.802813/

 

Seems like BGA to me

 

Now that's an interesting question and it relates to a couple of things.

Firstly the performance per watt has to be up there, secondly the form factor possible can't be wildly different to Nvidia and the platform as a whole will need to undercut nvidia in performance per $ by a fair amount.

It's not fair but due to market dominance AMD will have to pull ahead of Nvidia by a significant margin to be considered properly I think.

 

Considering amd brought polaris 10 via rx 570 to notebooks and vega is supposed to have better performance per watt I would say vega is nearly guaranteed for BGA (and probably mxms for embedded systems) the only question is how cut down it will be and whether or not they can bin them well enough to fit inside laptop thermal envelopes and still be performance per dollar competitive with pascal

 

yep pretty much my thoughts too. If Small Vega does end up between 1070 and 1080 performance, the whybother would be too strong to bother investing in the dev of it, without some other USP (cheaper, great crypto miner hash/W, etc)

A HBM board would have to be completely different layout wouldn't it? Unless the hbm shortage rumours mean small vega is released with GDDR5 (X?)

AMD also has nearly half a decade of complete absence from competitive notebook GPUs to undo you'd have to assume their relationships with the OEMs have completely withered and died

 

USP?
AMD has beaten NV in opencl every generation. I doubt vega will upset the trend so I assume crypto miners will flock to vega in desktops and possibly mxms (like that zcash poster who used taobao rx480 mxms in a custom mining rig)

not really, the gpu will be a bit bigger than otherwise but that isn't deal breaking and there won't be any BGA ram chips around the gpu since hbm is on die

true but amd has still pumped out commodity gpus to the many of the same oems so hopefully there is some good will there

 

Makes sense, and I do also recall some problems the last time Clevo added AMD cards to their lineup.

 

After Maxwell a lot of Crypto miners started using Nvidia again. Even though the double-precision performance was lower than the competing AMD cards at the time, the power consumption was also a lot lower. The total perf/watt/card was superior once scaled up. The increased efficiency has knock-on effects on cooling and PSU requirements as well.

The rx480 naturally changed this, but the shift to ASIC miners at the time also didn't hurt them too badly.

I'm curious as to just how low Vega will scale if they target mobile with it. Assuming they're targeting 250W to compete with the 1080ti, they'd nearly have to cut it in half to get it close to standard MXM power envelope.
Or they could make a 180/200W suped up version with external power but without any current governance on how external power should be implemented on MXM we have the current clown fiesta of non-standard MXM cards.

 

"A unique selling proposition (USP) refers to the unique benefit exhibited by a company, service, product or brand that enables it to stand out from competitors"
https://en.wikipedia.org/wiki/Unique_selling_proposition

 

Compared to:
- the already odd shapes of everything that still uses dedicated modules on the market (i.e. msi and clevo)?
- the already soldered variations that occupy the remaining HUGE chuck of the market (everyone else, the above two including)?

So where exactly redesign is needed? Let me put it more correctly - redesign happens for each and every machine nowadays. Once upon a time, the MXM form factor was an excuse (although I said quite a few times that an HBM chip could fit, given different hole spacing). I think that we can all see that this is no longer the case. So again - they are redesigning everything, so why not for AMD?

Then comes the DELL's Precision line (the only remaining machines that still use standard MXM-B format and recent AMD chips (WX 7100 = RX 480)). I'm eager to see if and how Vega would fit in there.

 

Is there anyway we can grab the vBios off that wx7100? I'd love to see what happens if it get's flashed to a Taobao RX480

 

I thought that was just forced incompatibility/lazy over specific design, the MXM standard is designed to allow/require up to 10 amps to the vid card from system main power and supposed to allow for 9 to 20v IIRC for example 19.5v is pretty common in laptop and at 10A the card would be able to get 195W from the MXM slot

 

This type of users is not among the most cooperative (there are always exceptions though) and given that there are less enthusiasts than ever before - try your luck asking in the Precision section, but don't expect response.

 

The problem is the power delivery pins can only deliver so much power. While the spec will be up to a certain voltage and up to a certain current that does not mean both at the same time.

 

interesting, do the power pins fail when run at Max spec?

 

Probably not the pins themselves, but you'd probably melt something else. In some cases the metal can become brittle and break under high current, but I think that's less likely.

 

19.5V is only a common input voltage, not a common running voltage. 19.5 is common it's a multiple of ~3.3v which is what Lithium battery cells run at. The other trick with the battery, is you'll cause voltage drop if you actually tried to draw 19.5V

As far as MXM spec, it is fairly lax on power delivery, but for the most part it's reliant on 12V input since that's fairly standard. Going to a higher voltage would mean having to run an entirely new rail, then redesigning the VRM to deal with it. Since 12V is a standard it allows the GPUs to be fairly similar across mobile/desktop at least in power delivery. The MXM connector is also quite dense, so I'm not sure how good of a time you'd actually have running more than say 150W down it without causing noise issues.

 

19.5V has nothing to do with battery voltage. Batteries commonly output 12.6V and 16.8V. Multiples of 4.2V, not 3.3V. The 16.8V Clevo battery typically outputs 16.8V down to around 13V on a 100% to 0% discharge cycle. The P150EM motherboard specifies 12V as the minimum voltage but I've never seen it get below 13V and still run.

MXM boards directly receive the 19.5V (plugged in) or 16.8V (battery) input voltage. There is no step down to 12V. 19.5V allows for higher power efficiency as it means the card draws less current for the same power draw. Even higher would be more efficient, as most power FETs are rated for 30V.

Exceeding the MXM slot rating just results in reduced power efficiency. My gpu has pulled over 270W and nothing bad happened. I did double up my MXM power FETs though as the default FETs weren't rated that high.

 

Colour me educated then. I was under the impression source power was 12V@10A.

 

that is more inline with how I interpreted the MXM3 specification I read a week or 2 ago

what voltage was it pulling that 270W at? what gpu were you powering?

You removed stock FETs and upgraded them? HxC

 

Interestingly enough the wx7100 mxm looks nearly identical to the gecube rx 480 mxm, if only we could get someone to post their wx7100 or E9550 vBios


Props to @song_1118 for tearing down his dell p7720 and posting high quality pictures

 

yep, thats @Khenglish for you my friend, basically the equivalent to @Prema on the gpu hardware side swapping out mosfets, gpu cores, overclocking gpus 2 tiers / gens forward, etc....