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

Yeah... I take no JayZIdiotCents video as factual. Almost every single time I see a video from him, he does something either flat out wrong, or absolutely stupid.

 

lmao

Well, you kind of made me curious about a ti...4-5 months worth of wait is well worth 60% out of the box performance gain over a 1080, and your predictions were pretty accurate in the past

 

Think I'm kidding? This one time he was testing the Titan X for 4K gaming saying it was worth it... using Battlefield 4... with MSAA off... saying it's the ultra preset... because he has "4x MSAA forced in control panel for all games". Except that BF4 denies driver-level AA overrides without enabling a hidden flag via nVidia Inspector, which I can 100,000,000,000,000,000,000,000,000,000,000.01% GUARANTEE YOU he did not know about, far less do.

This other time he tried benching DDR4 vs DDR3 using AIDA64. What does he do? Runs DDR3 in dual channel using Z97 board on a 4790K. Then benches it against DDR4 using a X99 board... with the memory running at 4K in quad channel... gets about double the speed and much higher latency, then claims DDR4 is 2x as fast as DDR3 even running at the same speeds. Like... THE VIDEO HAS OVER 100K VIEWS AND IS MOSTLY ALL LIKES. NO. YOU DO NOT GET TO BE SO FAMOUS AND HAVE SUCH GARBAGE DRIBBLING OUT OF YOUR MOUTH GETTING IT ALL OVER THE FLOOR SO THAT A STREET SWEEPER HAS TO RUN THROUGH YOUR HOUSE EVERY TIME YOU MAKE A VIDEO TO CLEAN ALL THE GARBAGE YOU EXPUNGED FROM YOUR THROAT CAVITY.

*is finished with his rant*

 

lmfao dude Ikr

 

Except GP104 is highly sensitive to temperature compared with cards like fermi. Who just flat out dont care...

GP104 throttle point is like 83c. GF100 is around 97C??

 

It would probably be less confusing if they brought back the M paired with MX designation and used that instead of the 1080M and "1080 notebook" it could be 1080M and 1080MX

Sent from a 128th Legion Stormtrooper 6P

 

Not as good marketing material.

 

I really can't see Nvidia using GDDR5X in a higher spectrum of 10xx when AMD is coming through with hbm2 in what's rumored to be around October

 

Sounds great to me! >2000 cores is what I want to see!

980M has fewer cores than desktop 970. If 1080 for mobile has more cores than desktop 1070, this is exactly the right direction to go in.

NVIDIA will continue to close the performance gap between desktop and mobile. Awesome!

 

I have an unpopular opinion.

I wish they would just give up on battery life. Just give me 10 minutes of battery life and the best laptop ever made.

The reason is simple. Whoever plays multiplayer, when current goes out, so does wifi and internet. Any single player can be closed and saved in 10 minutes.

And as much as portable usage goes, I see little chance of keeping something like P870 in lap and doing anything. Even if it doesn't heat, it weights about 5Kg for heatpipies and other components.

So yeah, I would like a laptop that did not take any battery into account when it was built. Just highest grade components, and making it more portable than a computer is.

 

Yeah, every reviewer has a flaw... But he did make a good point about NVIDIA showing off the 67C temperature with a 2.1Ghz overclock.

I take all reviews with a grain of salt as well, but it's clear how the card performs. Yes, it does "throttle" once it hits the temperature threshold of 82C or whatever, but this can be raised quite easily, as with any reference GPU in the past. I'd wait for other models to hit the market this week to see how they perform before declaring once and for all that the 1080 has a throttling issue. New cooler designs should eliminate the "issue." Fairly certain this was done on purpose, too, for GPU Boost 3.0.

 

Yeah, but then the temp rises and if you want that one in check, then the noise rises as well. In the end you have a hottish and noisy $700 GPU, which you have to tweak (and most can't be bothered). Not cool. And people laughed at AMD for bundling AIO, at least it was in the price and came cheaper. BTW has anyone thought about power comparison with R9 Nano? I mean, it's a 175W GPU, have you seen the cooling? It's half what the 1080 has.

 

That's just how every reference card has been, though. Everyone is treating it as if it has never occurred before.

The only difference is the Founders Edition being $100 more, which is a scam, no doubt.

 

Ambiguity laden tactics do have their marketing potential, but that potential tends to function on a kind of u-curve when it comes to their effectiveness -- in the right amounts, they can dazzle someone into spending their money in line with the marketing agenda, but when such tactics are taken too far, one can end up losing a frustrated and potentially confused customer. It can just as well be argued that getting too liberal with confusing naming schemes -- where it becomes difficult to tell one product from the other -- could actually divert individuals away from the target/relevant product. Case and point: a relatively layperson will have a much harder time searching for a desktop version of the GTX 980 in notebook form because they'll invariably hit a torrent of GTX 980M offerings, or systems where both options are available with no real clarification, because of the utterly mangled naming scheme. I would argue that the whole 'desktop' edition of mobile cards as a naming scheme is not going to yield the ideal results they desire over the longer term; even as marketing material, it's quite counter intuitive.

 

I never said nvidia marketing is good at what they do. Better than AMD? Sure, but it really doesnt say much.

I originally remembered 980 was suppose to be named 990m. It was probably called 980 desktop edition to sell the fact that a laptop is on par with a desktop(not true at all).

 

The 980 desktop edition was often faster than the 980 desktop reference cards though.

 

It is a small die card. The titan x and 980 ti are cheaper and performs far better.

I consider the big die cards to be proper desktop performance

 

nVidia learned their big die lesson after Fermi. For all intents and purposes, the 980 notebook edition was just as good as the desktop version of the card, many cases better, which was a first.

Now with the 1080, nVidia wants to drop the M, still cut the die down, and confuse everyone.

I'd call the 980 notebook edition a win and the 1080 a fail.

 

Apparently the next generation graphic card

Boring boring Chelsea

 

Honestly, I think he hit the nail on the head and did not realize he did. That was a surprise for me.

Everybody claims how these cards aren't hot and they're running fine etc. Has every other reference card been vapor chamber? If you slapped that 1080 reference cooler on a 980, would it overheat without 80%+ fanspeed? Or would it run cold? That's the comparison here. Aftermarket coolers WILL keep temps under control. Sure. But there'll be a clear difference in things at the end of the day. People down here who went through 580s, 770s, 970s, etc... my country is extremely hot. They all have basically the same story: 580s nearly sat at overheat temps. 770s were almost the same and sometimes throttled. 970s got warm but never overheat ready, and were overclockable without hitting thermals. When the 1070 etc comes... we'll really see. I still think that the reference cooler, while performing give or take around the same as previous reference coolers did let's say, is of a much higher quality, and if they were on other cards they'd perform much better, and that would prove Pascal is hotter than Maxwell which everybody seems hell bent on fighting to prove isn't the case.

 

High-end, high-end everywhere, and not a single mention of GP106...

 

What if they use the GP108?

 

Scary thought, but not likely. The GP108 is strictly low-end - even nVidia classifies it as only worth a fight against Intel HD Graphics. They may not even go GP107 this time, since the entirety of the 1000M lineup probably will be all-new chips. My guess is: GT 1020M to GT 1040M: GP108; GTX 1050M: GP107, GTX 1060M: gimped GP106, GTX 1065M: proper GP106, GTX 1070M-1080M and their 'desktop edition' variants (bloody stupid naming, as I and several others have already mentioned) would obviously be GP104.

 

I wouldn't be surprised if they opt to use Maxwell cores for some of them. Why waste a perfectly good desktop chip when you have Maxwell trash bleeding out of your ears still?

 

2017:
GDDR5X for mobile and we'll finally see HBM2 in the 1180Ti and Titan with the second iteration of Pascal, which will be an exciting re-brand.

 

Well they're testing GP106.

http://videocardz.com/60950/nvidia-pascal-gp102-400-spotted

 

Hm. A friend of mine put it very succinctly: he's not very interested in what is simply a die-shrink of Maxwell, which was what Maxwell was supposed to be in the first place (20 nm fiasco). I'll probably get a 970M laptop for cheap, and wait it out for Volta.

 

Wouldn't surprise me if the just rebrand the 970M as the 1060M.

 

apple dropped Nvidia from their macbooks a while back

 

I know 580 uses vapor chamber but thats fermi. Does GK110 reference uses vapor chamber?

 

Nah, we gotta compare GF114/GK104/GM204. This is GP104 here, not GP100 or GP102. I mentioned the 580 before, because it's what most people had at that price point back then, but I *WAS* considering aftermarket coolers in my statements as well, not reference. That's why I was pointing out that vapor chamber cooling for midranged Pascal then saying it performs like non-vapor-chamber cooling on midranged maxwell/Kepler is a rather unfair comparison to heat production of the dies.

 

Not really. They've been alternating between nVidia and AMD for a while, and I get a feeling this time we'll have Pascal rMBPs. Which are equally overpriced and run a crappy-looking OS (Why the hell did Apple have to kill off its iOS 6/OS X Snow Leopard design?)

 

I believe GM204 can get very hot if you start seriously pushing load on it with an unlocked. Its something...
I think Cayman(HD69xx) was of similar die size and uses vaporchamber for stock cooling. So did HD 7970 GE. It is very possible for highly clocked small die cards requiring vapor chamber.

The better answer would be, GP104 is highly clocked at stock.

 

that sounds very depressing

 

Well that's at overclocked states. This is stock. Stock is always rather low on the clock bumping spectrum for nVidia. I need to compare stock v stock. Pascal can OC rather far (not as far percentage-wise as Maxwell, but still far; easily 30%), so that means it's rather fine to compare them.

AMD is another story... their cards have generated a lot of heat for quite some time. GCN on the whole was rather hot. They're JUST fixing that, really. I need to compare nVidia to nVidia. I don't even think the 560Ti needed a vapor chamber, though. I could be wrong.

 

Could be wrong but some leaks suggest insane CUDA core counts and GDDR5X for the 1080Ti and Pascal Titan. No HBM2 in sight.

Nearly 4000 cores on the Titan and 3500 cores on the 1080Ti. Still a nice performance improvement, though.

 

Hope they going to release it or at least announced it around October if Vega rolls through
would have hbm2 REALLY mattered, though ?

 

I dont know where is everyone getting the idea Vega is launching in October...

AMD said something to the effect of "we only have polaris slated for 2016" in a shareholder meeting or something.

 

http://videocardz.com/59808/amd-vega-gpu-allegedly-pushed-forward-to-october from articles like this parroting everything.

It makes sense, though. They NEED to compete with the 1070 somehow. RX 480 is nice and all being around $230 or whatever, but it isn't a 1070. Now, if the RX 480 single card was around AMD's Fury level of performance, which is smack dab inbetween 980 and 980Ti cards in performance, then for $230 it wins. But if that's only in DX12 AOTS, and it's only around a 390X level of perf, then it's not good enough to pull sales. And they need sales more than anything else. It's why all of us are so annoyed with them.

 

Well, according to the 3d mark graphics score, rx 480 IS at about Fury level. The reason, AMD did not present these benchs in Computex was that they couldn't showoff CF variant as beating 1080, since CF is obviously (as of now) poorly optimized for 3d mark.

 

Btw, nothing was mentioned about that C4 card, while C7 was already confirmed to be RX 480, so I'm guessing C4 is a mobile variant of C7 with 12.5% decrease in performance.

 

If you are waiting for mobile lower-end this year then better don't hold your breath or you may never see 2017...

EDIT: As a fellow W230SS owner I know you eyeing the P640 series, but they have just refreshed it with the N16E-GR (GTX965M v2 / GM206)...it's a very power efficient chip and good OCer.

 

I don't know why they are peddling these CF benchmarks. It lame. Just let it rest on it's merits as a single card.

 

Do you have your P870DM as well? =D

I must agree. Crossfire is nowhere NEAR being ready to tell any consumer to get it. I can't even tell people to get SLI anything under top end GPUs... they want to show off crossfire with no nVidia Profile Inspector equivalent, and still unable to function in windowed mode? Especially since they don't release crossfire profiles nearly as fast or often as nVidia does either... they should not be pushing this at all.

 

Never even got as far as touching one...

 

My guess is that the 2048 cores is the 980M direct successor part.

The 1080 for notebook will be the full 2560 core part.

I've just checked prices of the current generation of GTX 980 for notebook models and holy crap I never realised they were so expensive! £3,500-£4,000 is the typical price! That's crazy

I've seriously got to figure out how I can fit a desktop into my current 2x2.2m room that already has a bed, desk and cupboard. I need to work miracles.

 

> Makes ultra mod unlocking and stabilizing and overclocking every aspect of machine
> Never touched one

#PhoenixMagic

What happened though? I know you had one on pre-order, and you wanted the 4K screen. #Life?

 

I did my part:

https://www.techinferno.com/index.p...fficial-thread/&do=findComment&comment=143282

Others didn't...

 

Ah, makes sense to upgrade to that. Unfortunately, Aftershock PC hasn't made the GTX 965M version available here in Singapore - only the 970M, and it still costs nearly S$2000.

 

I certainly hope that my post didn't inadvertently insinuate that you were saying such, because that was certainly not my intent. I was agreeing that Nvidia are getting lost in their own marketing myopia.