Pascal: What do we know? Discussion, Latest News & Updates: 1000M Series GPU's

What it means is that the system will still POST even if the sBIOS doesn't recognize it. My P377SM-A even turned on and booted (with a black screen due to eDP not being supported on such an old card) with my 260M GTX and when Sager upgraded my machine to 980Ms, they didn't flash the sBIOS and the cards showed as unknown but worked just fine. Previous Clevo machines would have shut off if they didn't recognize the card. As for vendors, I'm not sure. The 260M was a Dell card though.

I'm not claiming full support for any card but the system refusing to even POST shouldn't be an issue anymore.

 

Is there any chance we will be able to have 1000m series GPUs in a couple of months?

Nothing too high end, just a new standard for notebooks. (I'm buying a cheap 960m notebook and I don't know if in 2-3 months there will be 1000m notebooks for the same price)

 

Same price, not. For sure it will be priced a bit higher.

But it will be a bit better too.

 

Couple of months, possible. Could also be late fall.
In any case, don´t expect the successors to be as cheap as some of the laptops 960m and 965M are today.

 

GTX 1080 vs. 980 vs. 980 Ti: Overclocking Battle - Video

GTX 1080 vs. 980 vs. 980 Ti: Overclocking Battle - Article

8 game averaged test results:

1080 stock 28% faster than 980ti stock.
1080 OC 23% faster than 980ti OC.
1080 stock 5% faster @ 1440p than the 980ti OC.
1080 stock 6% faster @ 4k than the 980ti OC.
1080 stock/OC 60% faster than 980 stock/OC.

 

Someone said a while ago that HBM2 would be the death of MXM - to be frank, I'd be glad to see that, replaced by a new, smaller, more compact MXM v4 standard. Some MXM cards are nearly half the size of the W230SS motherboard, and they could do with some serious trimming.

Anyway, 1060M only matches 970M according to another certain post above? Sigh... Seems like I'll have to carry this little pocket rocket until 2018 at the least.

 

That's at least 60% improvement though using same ol' GDDR5.
Can you remember the performance increase upgrade from 860m-960m was? About or less than 5%. This is good news for people only targeting that section wanting to upgrade mid range. And I know too well what that performance gap is since owning 970m and 960m. I miss my 970m compared to the 960m.
1060m users will be happy if they had a 960m, or less previously.
I'm modest so 1070m instead of 1080m unless the gap is just insane

 

What? The 1060m will have the power of the 970m? You sure? That almost sounds impossible... utterly improssible... but it would be super awesome because I target midrange laptops...

And yes, the improvement from 860m to 960m was a joke.

 

You know what...I don't understand why people say HBM will be the end of MXM. If the Fiji die is anything of a basis, I think standard heatsinks will cover the HBM memory chips just fine. I know the Alienware heatsinks in the M18x and Alienware 18 actually have more copper underneath the black tape, so if the whole core/HBM package isn't too big, it should still be covered by the heatsink provided the HBM is level with the core and the black tape removed off the heatsink covers both the HBM and the core.

 

Yea I can't see it only jumping to a 965m level. After all the raving about is performance. If it's performance per watts is around x1.7 it's 70% increase. It might be a bit less for the 1060m. But even if it was X1.5 performance/watts it would still match an 880m. I think it will be between that and 970m.
I will be CAPITAL EMBARRASSED if my estimations are way off with me hyping up pascal. But from what we've come to learn from the 1080. If the mobile version is anything to go by. I'd completely believe the "impossible"

 

Hole spacing. I think that it would be able to align within it with the new gen stuff.

 

Can MXM 3.0b take that HBM bandwidth ?

 

I don't see why not. M18x R2/Alienware 18 are PCIE 3.0 on the MXM 3.0B slots, M18x R1 is 2.0 though

 

Well, that Chinese chart set the 1060M at +75% of the 960M. I totally believe that jump.

People aren't predicting the end of MXM, just the end of the MXM 3.0 standard.

MXM 4.0 or whatever will take over.

 

so 1060m faster than my 880m? shoot me now lol

 

Could someone tell me... is it a good, bad or terrible time to buy a laptop with a GPU?

 

It really depends on your performance expectations. But if time is on your side, wait a few months and perhaps see what pascal is like.

 

Perf/watt is great and all, but heat is different. If the 1060M is GP107 and it's 70% faster than a 960M, then great... whether or not it'll fit into the systems we're accustomed seeing with 960Ms and remain cool? That is an ENTIRELY different story. We need to see thermals.

I could not begin to tell you. As far as I understand it, Pascal will be stronger yet hotter and consume more power per relevant part (or they'll simply gimp mobile GPUs further from desktop counterparts in speed etc). But Maxwell is using bottom barrel parts and users are complaining of bad OCing potential and relatively high failure rates. AMD might as well not exist right now as far as mobile users are concerned.

 

I would say bad moment to buy one right now. Well, not entirely.

There is a chance that Pascal is really much better and will OC, but will probably be more expensive, and right now all 9xx based laptops seem to be on sale. You could get a 970m based laptop for the same price you will be getting a 1060m with a high probability, that it would end up with the same performance....

So right now, it depends on what laptop with what card you want to buy. If you want enthusiast level, it is a bad moment, if you want a cheap mid entry, or mid high, it is a good moment .

 

High failure rates? Really?

 

Not necessarily: it could be that Pascal's performance per watt is better at lower TDPs. However, even if mobile GPUs are further from desktop ones than they were before, mobile Pascal should still be significantly better than mobile Maxwell.

 

Keep in mind that the latest mobile chip has not been officially announced and even after the announcement it will likely be months before it is widely available. If you've already got something functional and can wait 6 months then I would wait.

 

Agree with this.

I am waiting with a 460M based M17xR3 and looking to buy a pascal based enthusiast system. I don't mind waiting till December to be honest.


Sent from my iPhone using Tapatalk

 

Great advice. Nailed it.

 

Benchmarks of the GTX1070 have been leaked.
http://videocardz.com/60265/nvidia-geforce-gtx-1070-3dmark-firestrike-benchmarks

It seems that it is slightly faster than both the 980Ti and Titan X. So, if the GTX1080M is what we have been guessing (i.e. the exact GTX1070 chip with lower core and memory clocks), I guess we will get a mobile graphics card that performs like a 980Ti? (give or take a bit depending on actual clocks we eventually get)

 

Newer cards, as far as I've heard, have rather high failure rates. At least from users on this forum, which I would trust.

I don't know about that. Pascal has better perf/watt, which means, for each watt consumed, more power is produced. But then every new architecture has performance per watt improvements. It ain't anything special to write home about.

The issue is that Pascal cards are designed for a higher TDP target. 165W Maxwell vs 180W Pascal, and Pascal can't OC as high (Percentage based) as Maxwell without hitting power draw limits.

This means that if you put Pascal's GP107 with an overall 70% improvement over GM107 (and that 70% improvement is higher than the 61% improvement GP104 gets over GM204, mind you) then two things are going to be a given: The new card MUST draw more power, and the new card *MUST* be much hotter (as we've seen with GP104 being as hot as GM200 or more).

So assuming the card is coming out as-is, then one of two things has to be the end point. Either:

- The card will be too hot and power hungry to be put into the laptops we're currently witnessing, and the 1050M will have to become the new "960M" in terms of its place in smallish midranged notebooks around the world due to heat and power draw issues, or
- The card is in fact GP106, and is meant to replace the 965M instead, and is neutered to the temperature and power target limits that the 965M has, and thus the improvement becomes less if you compare it to GM206 to where it makes sense.

 

good to know the rough performance level of the 1070, thatll be the max. 1080M level and thus +40% of my current setup. not sufficient for me to upgrade phsew, my wallet just gave out a sigh of relief

Sent from my Huawei Mate 8 NXT-AL10

 

The GTX 1080 literally drew 7 more watts in max power consumption tests, versus the 980, and it ran at the same exact temperatures. You conclusions are frankly quite off base from reality.

Also you seriously are going to need to provide some evidence of Maxwell chips having high failure rates. You can't just say that and leave it there.

Faster than a Titan-X is way more than +40% over the 980M, if the system in your sig is what you refer to as "my current setup".

 

I'm talking about the rated TDP. 980 is 165W and 1080 is 180W. And 1080 has hit TDP limits in overclocking as well. And where did you see it run at the same temperatures? I need YOU to provide something here for me now. I'll use THIS VIDEO as my reference with heat.

I have been told by users like @Mr. Fox and @TomJGX and a few others that getting good 900M cards has been difficult and that prematurely dead cards are a thing. I've seen no big threads making a stink about it, which is also what they were somewhat annoyed about. You can ask them to provide more proof, because I really don't have it. I simply said I was told by users of this forum that they've seen it and I trusted that information.

If a GTX 1070 is give or take around Titan X levels, then mobile cards are likely to max out at 1070 levels of performance, is what he's saying, and he's considering max OC to max OC for upgrade potential. If we get a 1080M to 1070 ratio like 980M to 970 is, then it'll not be 50%+ better in raw power (though in some games, due to better tessellation engine, it might prove to be).

 

I was a bit going to ask to what it is compared to be +40%, because the 980 desktop 200W edition is +40% over a 980m. It would be so and so for a 1080m to barely reach 980desktop levels. I mean, I don't dislike the idea, but I would certainly hope that 1080m and 1070 is a bit faster than 980. A bit around Titan X levels means around 980ti levels or so, and this is about 10-20% over a 980, if the scores I read were correct

Either way, it is good that these are coming with more speed, I cannot complain. I just wish they would roll faster, and cheaper as well.

 

i was comparing the firestrike graphics score of the 1070 to my oced graphics score sorry if i didnt elaborate before

 

Don't worry one bit

You must be OCing like a pro there, to have this rate of difference

 

I am excited to see what the 1040m offers. Provided it isn't a rebrand, a ~600 dollar laptop with decent gaming ability would be great for the Fiance. She's still on an APU with no dPGU.

A very cut down Pascal within Zenbooks could be great too. I know we usually don't care about the lower end but there's potential for cheap gaming machines here.

 

The 1040M is rumoured to perform as well as the 860/960M, which is a great deal there. Skyrim on max settings at 60 FPS, for example, or Witcher 2 on very high settings at 45 FPS.

 

Nvidia cheaper? The 2 words do not belong together!

Sent from my SM-A500FU using Tapatalk

 

I seriously think that if the performance jump looks like that, they might jump the prices too. It would make sense, especially as kaby lake makes 920 redundant.

 

At 388% upgrade over my 960m. I don't care what my wallet thinks. Although I think I'll settle for the 1070m. My wallet is happier with that.

 

I'm going to do the same - order the 1070 (Desktop Version) when its available. It will be a massive upgrade from my Radeon HD 7770 and all I've got to do is upgrade the PSU on my gaming machine to a 500w.

Its a bit frustrating having to wait for its release though, especially as I've watched some YouTube videos of TW3 and really want to give that a try next!

 

about 4 860m cards in performance. And I am able to play most games in 60 fps, FHD and full details.

I am not entirely happy with the price and still not being for laptops, but still, that is a huge performance leap.

 

I've been assuming the 1050M will perform like a 960M, and a 1060M will perform about 40% better than the 960M.

But if the 1040M will perform like a 960M, then that's very, very impressive. The 940M is weak.

 

I'm guessing the 1040M will perform like an 850M/950M. It shouldn't perform more than 50% better than the current 940M.

 

I wonder if 1070m will match stock 980(desktop) without OC'ing

 

I kinda doubt that.
GTX 980 is around 40% faster than the GTX 980M.
GTX 980M is around 30% faster than the GTX 970M.
So, GTX 980 is around 80% faster than the GTX 970M.
GTX 1070M could perform around the GTX 970.

These are just my assumption.

 

New info about upcoming Pascal chips (from AIDA64 system info tool developers):

http://videocardz.com/60289/breakin...rs-confirm-pascal-gp102-gp106-gp107-and-gp108

 

^ makes sense. I knew GP100 was launched already, which means they're simply milking GP104 for a while. They could just toss GP100 into gaming cards and be done with it; some of the Tesla cards are already like gaming cards with simply extra vRAM and different solder etc.

Oh well.

 

It wouldn't be nVidia if it retained compute capabilities.

 

As far as I heard, OpenCL is even worse on Pascal since they favour CUDA as usual. But if they're removing DP encoding blocks from GP100 and calling it GP102 and still overpricing it, then... I hope Vega 10 is actually not the top and Vega 9 exists and AMD hasn't even told anybody but their engineering department that it exists yet and that it owns GP100 (far less GP102) for like $650-$700.

*goes and makes wishes to the gods of IT*

 

I suggest you get a world religions book and make offerings to every deity from each one lol.

 

I will get a GP100 based titan instead of a GP102 titan if there is two options...

So one compute enabled titan vs disabled titan?

 

I only have one soul to offer

I am simply speculating. There is a very large chance that they just invented "G*xx2" as a class of GPU silicon to do exactly what I said and they won't release GP100 to mainstream, and there is also a chance that they will release both GP100 (as Titans) and GP102 (as DP-less titan-class cards), and then charge $2000 for a Titan or some crap. Who knows.

Either way, it's not good for consumers.