Wait for Mobile Pascal on Sager/Clevo?

Hmm. After even more thought, I starting to see why waiting out for Clevos with mobile Pascal seems like a legit move:

1. Current cards cannot do 4K output well.
2. Even if possible, 4K laptop screens eat up too much battery
3. Output to external displays can be hit-or-miss affair due to Optimus/MSHYBRID confusion
4. Current cards are at least an year old
5. Current games seem to be straining current cards already

This train of thought means the following possible courses of action:

1. Wait for mobile Pascal, which should resolve all these at one go. Just a matter of how long can that be...
2. See if current Clevos have MXM, which will allow for future resolution of most points (especially 4 and 5). But will the Clevos support future cards?
3. Change expectations, reconsider the need of how much computing power needed on the go. This means a need for a stronger PC to somewhat "cover" for the loss in computing power, but this also means a need to manage both systems somehow (current plan is to let the Clevo as the main computing device, PC as backup server).

Am I thinking this right or am I missing something else?

 

Pascal does look to be promising. But there isn't set dates for it as of yet, so maybe it just depends on if your current system is failing or barely chugging along. I'd maybe wait a little longer and see what info comes out to actual performance. I'd be skeptical about performance claims until something launches. "It's 40000% better and makes your house smell like warm cinnamon rolls!"

As long as there are compatible MXM cards, I'd assume it would be supported by Clevo, or at least someone might find a way to make it work. But just really won't know until models are out there for sure.

I think you've got at least a realistic look of current gen models. If you are really wanting something for 4k, then if possible, I'd hold off and see if more info comes out for performance.

 

1/ Depends on what games you play, GTX 980 can cope with most games at 4K with med/high settings as can SLI 980M. But if you want true 4K Ultra gaming in an environment that you can get the most out of it, I would suggest a 24" minimum external display

2/ Shouldn't be enough of a differnce in itself to stop you going 4K

3/ This isn't a problem with any system you would be looking at that's capable of running close to 4K. The external outputs in the P870DM, P75*DM, P77*DM, P775DM, P6**R* are all permanently connected to the dGPU. Only the internal display switches between iGPU and dGPU through MSHYBRID.Even in the P640RE one of the mDP ports is permanently connected to the dGPU = 4K output capable.

4/ Can't argue with that

5/ They cope pretty well considering their age http://www.notebookcheck.net/Computer-Games-on-Laptop-Graphics-Cards.13849.0.html

Waiting for Pascal is an option, but it won't solve all these problems in one go, it'll simply give you more performance - if you can wait that long.....;-)

 

My laptop is running nicely but just that the 560M is a little too old for current games. And it's too heavy lol.

Am I right to say that only those heavier models with desktop CPUs like the "Batman 2.0" are the only ones with MXM slots?

1, 2. Ah, I see, looks like I will be fine with a 1080p laptop screen since I am eyeing those 15" and below laptops.

3. Now I see why there is more than one mDP ports... And I really thought that the switching will affect external outputs too :X

5. Hmm. The 960M does not seem to do well with current games already (Playable only on Medium settings for most games); even the 970M seem to be struggling a little. The 980M does well but this comes at a price premium.

 

FYI, in the case of the P6XXRE/G series, the entire system can be switched between Optimus and full dgpu only for the internal screen.

Do not take the 960M as any indication of what current Maxwell 2.0 based chips perform. It's based on the older GM107 chip which is significantly slower (significantly lower power budget to be fair). The GM200 based 970M is a little shy of twice as fast ast the 960M. The 980M is a good chunk over twice as fast. Both do pretty well as far as I can see. The biggest problem I've seen is most triple-A games released in the last couple of years have been pretty shoddy from a performance perspective but haven't moved much graphically. Not to mention almost every franchise in recent memory has simply piled onto the same 5-10 year old engine with not much thought to efficiency (Batman series, Fallout 3-4, COD series, Assassins Creed series, GTA5 etc).

 

Ah, I forgot about the mux that the P6XXRE/G series have. One less concern I guess.

And yeah, good point, I did not consider the fact that current games aren't being well-optimised for now. If this changes, looks like 970M will be pretty safe bet. I am eyeing the 960M though due to it being much lower cost than 970M; I guess there is a price to everything...

 

I think this would mainly come down to when you are looking at actually needing/wanting a new PC. If you don't mind waiting possibly another year, than Pascal might not be a bad option, however there's no release date for it yet. I would assume there will be a x60m from Pascal as well which while it would be more powerful than the 960m, It will be quickly outdated by new AAA games. Other than the GTX 980 the other cards have been out for awhile, so you are probably seeing better price points on these cards than you will at the start of Pascal.

The other thing to consider is 4k is still an adapting technology and games keep pushing the limits assuming 1080p resolution, so there's no guarantee that new cards with new games will be powerful enough to push 4k.

 

You have a point there; maybe I need to reconsider what and when I exactly need in the computers (note the "s"!). I am in a weird limbo state now (and a bit out of touch with tech recently), that's why I am asking these smaller questions first, which will hopefully help me understand the situation a bit better and thus plan better :X

 

Even the modern intel IGP can drive 4k outputs too over display port The biggest difference would be high refresh rate 1440p support and G-Sync.

 

I was replying to the OPs concerns about MSHYBRID, not what the hardware is physically capable of.

 

Everything has pretty much been covered except for this point.

Clevo finally implemented full MXM support... Any Clevo MXM card should be a plug and play affair (with a modded driver INF).

The real question is whether or not Pascal will be MXM.

 

I would hope so. it would be crushing to see that not be an option, especially now that there is a desktop GPU on MXM that opens some new doors for possibilities.

 

I hope so as well, it just seems like Nvidia is trying their best to put an end to MXM so we will just have to wait.

Nvidia and Intel don't care about enthusiasts and they stand everything to gain by soldering everything.

 

And also whether more Clevo models have the MXM slot or not (I understand that as of now, only those bigger laptops with desktop CPUs have that).

Same, not having that option is kinda counter-intuitive.

And looks like I may need to rethink things a little. Initially I thought that I can just get a lighter laptop that can handle almost everything (gaming and photo-editing especially) and just a home server to take care of backups e.t.c... I have a gut feeling that this plan will need to be tweaked a little. Time for me to get just a bit more data, sort out my school exams on Thursday and then open a separate thread to deal with this?

This thread is still free for discussion of mobile Pascal on Sager/Clevo though!

 

I hope so, there is no reason certainly why the smaller cores should not fit within the mounting holes and you have a LOT of room for power circuitry at that point. The Fury-X core is HUGE but the die shrinks should help with that a lot.

 

The decision has been / will be made from a business POV, in the case of next gen it won't be restricted by the form factor itself.

 

Yes it will, though change always costs money.

 

But will Clevos users still be beholden to Prema for future BIOS support?

 

What is being said is that due to the standard nature of the MXM implimentation that is not needed.

 

Me personally, I will absolutely wait for Pascal to see what it has in store. The question for you is, can you possibly hold out until summer next year for a 50% better overall purchase, or do you absolutely have to buy it now?

I am not saying the new platform will offer smooth 4k gameplay, but who needs that on a laptop anyways....I still have nightmares playing 1080p on a 15,6" screen, especially RTS games....stuff is tiny enough as it is, lol.
But its definitely worth waiting for Pascal. Who knows, if it comes out with 7-th gen intel procs, you get double the benefit of waiting!

 

I doubt 7th gen of Intel will be out by then though lol. But yes, the question is how long more can someone wait it out. Or alternatively, if someone has to buy now, has she/he have the chance to upgrade once Pascal is out.

And yes, I need more time to think through what my problems exactly are; sudden influx of work and school stuff took up more time than expected :S

 

seeing as the ODMs like Clevo are the ones responsible for designing and building MXM gpus, one can only hope that they will keep to their "rebel" stand of continuing to cater to enthusiast´s wishes of fully serviceable laptops so yeah, even if nvidia doesnt intend for pascal to be released in an mxm factor, clevo can just show them the finger, buy their gpu chips and smack them onto standard mxm form factor boards: WIN!

 

O_O Whoa, I didn't consider that possibility!

 

I feel like that would result in Clevo getting sued...

 

unless nvidia makes them sign some kinda agreement restricting the usage of their chips to specific form factors, there wouldnt be any legal grounds to sue them. also, clevo is one of a few select brands to buy off nvidia´s top-tier mobile gpu chips, so suing clevo would be shooting into their own leg

 

Actually most MXM designs are based off the Nvidia reference design... I think the 980 desktop ones are a bit more original but likely done with a lot of help from nvidia.

 

I can see the rumour mill running on this for a while yet, all I will say is this - an ODM won't need to show Nvidia the middle finger in order to offer what has been suggested above, it may just be that the ODM has a little more work to do compared to launching previous gen MXM cards......as I work inside the industry I really can't say more than that for now.

 

always open for pm's and job offers @XMG but whatever you may or may not be implying sounds encouraging for us folks sporting mxm machines

btw, juuust out of curiosity: what kinda profile would one have to sport in order to be able to land a job with u guys?

Sent from my Nexus 5 using Tapatalk

 

If Maxwell is the last generation to be MXM compatible/upgradable, I will never buy a gaming laptop again after my current rig becomes obsolete - I'll switch to the desktop world. I've owned 4 different gaming laptops over the past 9 years, but there's not point in owning one if it's not upgradable.

I don't think I'll be the only one. NVIDIA and OEM's will loose customers and profit, if this occurs.

 

I don't think you will be either, but I suspect the enthusiast market that empathizes with the sentiment is a niche that NVidia and/or OEMs are prepared to frustrate, and even lose. MXM cards will certainly be missed if they go by the wayside.

 

Are we expecting the GTX 980 laptops to be Pascal compatible?

 

We can't know at this stage.

 

With compatibility, it is always hard to tell until new cards are launched, or if there is some credible info before launch, which might not be much to do NDAs.

 

If Pascal is going to move to stacked components, then I hope the MXM add-o boards are smaller and can thus fit in smaller, thinner notebooks. The current MXM cards are gigantic; I think my W230SS motherboard is not much bigger than the video cards themselves.

 

nooooo, no thin n light crap please! sure, make the mxm boards smaller but keep the size of the chassis and fill it up with moar coppah! also get rid of 2.5 inch slots

Sent from my Nexus 5 using Tapatalk

 

Hmm. So for now it's either MXM with a bulkier chassis or BGA with a thinner chassis?

 

pretty much! well, looking at your signature, ud be amazed at how much thinner the P5 and P7 DM series is compared to the HM series. my previous machine was a P150HM with a 2960XM and a 7970M, both overclocked to within a mm of their respective lives (and in the case of the gpu and mobo, a mm too far in the end *lol*). when i switched to my current dark knight the first thing i thought while unpacking was: OMG! ITS SO THIN! and guess what? its about twice as fast....

 

Yeah, I am amazed when I saw those thinner series being sold over here; technology has changed much such that my laptop is being heavier and slower than those newer models. One of my main decisions lie in whether to go ahead and get these BGA laptops that are thinner but un-upgradable or reconsider what I want/need.

 

Quantum computing + photonics = next to zero heat energy dissipation = MacBook Air that's more powerful than your P751ZM. Laptops are going to get thinner while getting more powerful. Already happened with tablets and smartphones.

For example: my laptop is thinner and lighter than my sister's (i7-4710MQ + GTX 860M vs i7-2630QM & GT 525M), and it's orders of magnitude faster. Her laptop? I could barely run COD BO2 and Kerbal Space Program at medium settings with 30-40 FPS. Now, on those two games, I just max out the settings with no drop in performance. Just look at the P640RE. Same weight, even thinner than the W230SS - but - more powerful ( GTX 970M vs 860M). That is some improvement in just one year.

Efficiency is the word. If someone made a laptop thin, and light, and yet powerful as hell, I'd snap it up in the blink of an eye. To be honest: thin laptops are actually quite cool (looks and technology-wise, not temperature-wise).

Copper is heavy. Heavy = not good. When laptops run cool, they don't need so much cooling apparatus (screw Asus and laptop liquid cooling - pile of useless, heavy crap).

Why? Not everyone has enough $$$ to get 1TB SSDs when they need the space.

 

I think he meant he doesn't want "thin n light" non-upgradable crap. If I could have the laptop in my sig in a less than 1" thick form factor, you better believe I'd take it. But if that came at the cost of no longer having modular components, then there'd be no way I'd want it.

 

When it's non-upgradeable? There I agree with you, and hence all MacBooks are worthless in that regard. Stupid piles of locked-down crap, indeed. You can't even change the screen easily if something goes wrong.

Well, you technically can (Clevo P640RE), but, yes - CPU and GPU are BGA (and no SLI). Which is why I started this discussion in the first place, hence coming back to square one. Modular laptops are bigger because the modules and the add-on connections are big, as well as the add-on boards. I've seen MXM cards, and there seems to be a lot of empty space (I'm not a hardware engineer, so forgive me if it sounds ignorant), and I believe both the silicon die and the entire board could be made smaller. Why put the VRAM so far away from the chip? Do it like an SoC and voilà, you have an add-on board that's 50-60% smaller. That's where Pascal comes in.

Unfortunately, the only way to get a socketed CPU now is to go to desktop chips and I feel Intel has let all of us light-laptop gamers down by getting rid of rPGA and socketed mobile processors.

 

I don't think is worth the wait. Even if the performance and efficiency increase is massive you have to wait +6 months more, it is not like Pascal is around the corner. Upgradeability is nice but I would not count on it. Nvidia has no competition in the laptop segment atm so they could easily push a propietary conector instead of MXM. Maybe some company will invest to turn the pascal cards to be compatible but only if it is profitable.

 

i want higher efficiency in terms of HIGHER PERFORMANCE! what weve seen the last few years is identical performance in smaller form factors or at less power consumption = "higher efficiency"

i dont want the same performance of my dark knight in a 1" thin turdbook, i want the increased efficiency to give me MOAR POWAH with full upgradeability, naturally.

also, just to clarify: definition of "order of magnitude" - a class in a system of classification determined by size, typically in powers of ten.

which means ur machine is AT LEAST ten times faster than your sister´s laptop? doubtful.... definitely more powerful, yes, but not by a factor of ten or even higher

quantum computing and photonics - less heat generation, sure, i love it! but still, as mentioned above, i dont want a ten times smaller package with the SAME performance! i want MORE performance! so please use the available space and fill it up with MORE computing power and dont artificially restrict it just so that u can brag about having a thin laptop, thats all smoke n mirrors in the end if u cant get your job done

best example is intel: sandy bridge to skylake is a 28% increase in computing power at identical clocks, so thats just 7% per generation. compare to that the iGPU performance power, THATS where ull see "several orders of magnitude" in performance difference. but seriously....who the heck cares? i dont want 1000% boost in iGPU, i want ACTUAL cpu performance boosts! because thats what the CPU is for! iGPU should be just enough for desktop, office, browsing, video playback, thats it.

 

There's no difference. The ratio is what has improved: computational output (in FLOPS) ÷ power input = efficiency. It's what manufacturers choose to do with this that matters.

Clearly, one size does not fit all. You'd rather big and bulky and powerful; I would rather thin, lithe, light and equally as powerful. What's the point of having your laptop weigh 4 kg and look and feel like a brick when you can have another one that feels like a feather but works just as well?

Yes, I know what an order of magnitude means. I was exaggerating. Throwing your spear back at you: definition of exaggeration: 'to represent (something) as being larger, better, or worse than it really is.'

Ah, you miss my point, or maybe (more likely) I wasn't correct in expressing it. Laptops will get thinner and they get more powerful. There's just no point in keeping a notebook ridiculously thick (and as you put it, more copper) when there's no need to. QC and photonics promise nearly 100% efficiency - meaning zero cooling apparatus. You still want to keep your bulky P751ZM when you see me playing Cyberpunk 2077 or a hypothetical Crysis 4 at maximum settings on a 1 cm notebook at 4K 60 FPS?

As I said - my own W230SS is thinner and lighter than my sister's 2011 Sandy Bridge notebook. It is also more powerful.

This shows silicon's failings. It's not possible to eke out a 50%++ performance increase every generation nowadays. Moore's Law has stopped working for CPUs.

As for iGPUs - with all due respect, I think they have come very very far. Compare the Intel HD 3000 and the latest Skylake Iris Pro. If you have a choice between GT 930/940M and Iris Pro, it's much smarter to choose the latter. Also, the latest Iris Pro 6200 gets about 1800 on Fire Strike. For an iGPU, that's bloody good.

All in all, we have derailed from the topic. I simply asked if a smaller MXM chip would be possible, because the current form factor appears unnecessarily big. What's unnecessary has to go. If they can make it smaller, then manufacturers could fit it in 13-14" notebooks, all while keeping the weight down.

 

MXM Cards are (especially NVIDIA reference cards) are actually very packed - you can't fit much more on them. That's one of the reasons the 200W and 180W GTX 980 mxm cards are larger. The main issues you have are fanning out all of the large bga packages on the card (16 memory chips and the gpu core itself), keeping trace lengths from the memory to the gpu of equal length, and providing thick enough power and ground traces to handle the current throughout the card.

The only main reason why desktop cards are so much larger is because they require much larger power traces and planes to allow for a 300W load. I'm actually surprised all this can be made to fit on an MXM card.

You're right though, HBM will free up a lot of space. You might be able to make dual gpu MXM cards with HBM. Imagine Quad SLI in a notebook.

QC isn't here and won't be mainstream for a long time. Arguing that point now is kind of moot, quite frankly.

Making a smaller MXM module with Maxwell, I'd argue for my above statements, is not possible - if it was, then the 180W and 200W cards would be standard size. With HBM maybe, but power draw does a lot to dictate size. Also, why make it smaller, when instead you can pack more onto the existing card, and as I said make dual gpu MXM cards?

 

The power density gets too high with dual chip boards really.

The 980M PCB struggles around the 1500mhz memory clock area due to the PCB, any more packed and speeds would tank even more.

Currently we have a maximum of 4 phases for the core and 1-2 memory phases with a 256bit memory bus running at around 1.5-1.6ghz (on the extreme end with rubbish yields) . None of those with the current chips could be increased without decreasing the other.

 

However with pascal, the memory will no longer be limited by the pcb, and isn't pascal supposed to draw even less power than maxwell?

Edit: Ohh, never mind - you were talking about the memory being power starved. I was thinking about crosstalk because the pcb is too dense. However, not having to fan out 16 memory chips should allow more room for more phases . . . hmmmm

 

Somehow it got posted twice =P

 

Having the power phases is one thing but trying dissipate 250-300W from a few inches square gets very difficult.