Volta: NVIDIA's Next Generation GPU Architecture (2017-2018)

What pops up is that beside the ai chip, is just seems a bigger pascal chip with a slightly smaller node. At least considering the tflops.

Could the design be really similar to pascal?

Enviado do meu MHA-L29 através de Tapatalk

 

Like to see this in a Laptop!

 

I'm betting on 2018 late Q2 early Q3 before it shows up in laptops since it's likely to arrive in desktops earlier in the year.

 

As far as I know, only GV100 have been mentioned in drivers.
It was listed over a year back.

GV104 have not been listed yet

 

They will have project the cooling from scratch I think, chip look like he need a lot of volt(a)s

 

[rimshot]

But yeah, if they're building up cooling from scratch my estimate may be a bit optimistic.

 

Now that Volta has been announced, discussion within this thread is now expected!

Here's a decent blog post: https://devblogs.nvidia.com/parallelforall/inside-volta/

 

It's been pretty "low" Volta here for a while, still waiting for the announcements, or at least a better leak.

Gonna be a while still...

 

Yeah, they've still got more Pascal stuff to release, they won't announce anything until they've got what they can out of that.

 

Pascal "marketing" is on prime time now, waiting for Titan xp. There is no point to feed peoples with Volta right now

 

Plus I imagine they have some information drop ready to take some air out of Vega.

 

Vega 1.0... that's for Pascal.

Vega 2.0... that's to compete with Volta.

 

I'm kinda interested to see if they come through with new FirePro stuff in there as well.

 

Competitinon betwen Green and red ends years ago.

After Vega Frontier Edition they will put RX on market but it will ends with "AMD marketing talk" like always

Is sad but is painfull for ours wallets also

 

The competitiveness of AMD / Intel that ended years ago is now with us again, and if AMD keeps up the good work it will continue forward, even after Intel /Nvidia drop prices and kicks up performance.

Competition doesn't just only mean which is marginally faster in some benchmarks.

It also includes $ cost of implementation and use: CPU + GPU + Motherboard + RAM + PSU + etc.

If AMD's solution is that much cheaper than Intel's, and performs as good (close enough) or better, then AMD choices will make sales over Intel choices.

The AMD Ryzen and coming CPU's are the most "competitive" right now, with Intel being way overpriced, so that will be the first thing to change - Intel must drop prices to be competitive.

AMD Vega 1.0/Pascal and 2.0/Volta can also be competitive based on $, mostly in the Professional and AI markets, perhaps less so in the consumer market if Nvidia drops prices enough.

In the wide range of low / mid range laptops where AMD will be more price competitive. Any marginal performance difference will be ignored by consumers wanting to be able to afford usable gaming performance. Those APU's should provide real GPU performance over the Intel onboard iGPU's.

Enough of us are ready for a solid alternative in AMD for CPU / GPU solutions where we can use them, even if we are giving up a few % points in performance in some things, especially if the things we use them for to make $ money are saving processing time over same priced Intel performance.

If you mean "AMD talk" where some performance isn't up to their claim, that's always going to be with us in some form even with Intel and Nvidia - some things will run as showcased for Intel / Nvidia and some for AMD - their solutions are independent and characteristics of each deem it likely there will be some software running better on AMD and some on Nvidia / Intel.

It's going to be much different now moving forward with AMD back in striking range of performance of Intel and Nvidia with $ savings as an added incentive.

 

Shouldn't even matter if you're a fan of one or the other, being this close in performance will make both companies better.

 

Right, there is no "fanboy"'ism involved for me, I started out on CPU's other than Intel and AMD, and became knowledgeable and critical about features and performance before I "bonded" with anything available today.

It's merely down to helping cultivate the options, and broaden my own experience and knowledge - it's hard to understand just how much it helps to gain many different POV's, one of the benefits is avoiding the state of being a "fanboy" about any one solution.

It's time to spend some time and $ supporting the AMD world again

 

Honestly if I had the money I'd have builds from each.

 

If that's the case I'd think AMD would be the better single choice

 

We'll see when the newer stuff comes out, but for now I'm sticking with what works for me.

 

I've been dying to replace my nvidia cards....I have an immense hatred of nvidia but nothing in the mobile space comes even close.

 

I'd love to see Vega laptops. Though I won't fault mfgs if they are a bit gun-shy about it, some of their last few mobile generations were a bit rough IIRC. (inb4 "but my 6990M worked perfectly")

 

early production reliability aside, the 7970M was the last hit AMD had. stock clock for stock clock after driver maturation, it sunk the 680M. ever since then....its been sad. 2012 was 5 years ago.

 

True. If the card survived that long it ended up the better option, but it left a bad taste in a lot of mouths because of the early failures.

 

It's almost like AMD was given a 5 year "timeout" for bad behavior... I hope they've learned their lesson and are now on their best behavior

 

Yeah, it would be good to see some competition in the high-end mobile space to bring prices down.

Sent from my SM-G935F using Tapatalk

 

Even starting with the 1050ti / 1060 / 1070 mobile space would be good, that's where most of the sales are right now.

IDK if the top end Vega 1.0 / 2.0 or Volta will fit in a laptop - power and cooling are going to be tough to provide.

 

I like that analogy. Hopefully they can deliver!

 

Depends a lot on how cool/hot they run in general. Pascal runs hot, and many manufacturers are having trouble cooling it so that it does not throttle. And then you have crap like the Razer Blade 14 and Razer Blade Pro, which are crippled on arrival. @Papusan

Sent from my SM-G935F using Tapatalk

 

I'm really hoping some vendors step up and come up with some creative cooling solutions for the coming high thermal output top end GPU's and CPU's, or we are gonna be stuck with desktops for the high end again.

Something "fold-out" with radiators and fans, either integrated or something we click into to provide more cooling.

Maybe ubiquitous liquid cooling with places we can "jack-in" for cooling while mobile will catch on

 

ASUS has one like that, but it's hella expensive.

 

I am still waiting for someone to make a custom water cooling loop inside a laptop.

 

You mean other than the Asus GX700 / GX800, out for about 3 years now, much easier to find for purchase recently too

 

Not enough space inside for pipes, rad and fan that will cool it, you're pretty much stuck with externally attached systems like the ASUS models @hmscott mentioned.

 

The Asus GX700 / GX800 comes with a roll around case to hold the hydro-cooling dock + laptop, and if you wanted to build a custom loop mobile computer, you could take a roll around case about that size and build it inside of that

Then you'd have a complete water cooled desktop "mobile" solution

 

Use this to build: http://www.lian-li.com/en/dt_portfolio/pc-tu300/



Might take some creativity to attach the radiators now that I look at it.

 

Just something that could fight a gtx1070 would be great.
Most consumers shop at this level and lower.
And to game at 1080p you don't really need more for a few years...

But a 4k gpu would be great, even though I guess we'll have to wait for 7nm cards.

Enviado do meu MHA-L29 através de Tapatalk

 

Nvidia want pushing out crippled versions of 1080, to meet the laptop manufacturers!! Nice!! And the notebook manufacturers hope graphics will use less powa with next models. Forget high end cooling!! It will be worse... More Apple Books. With Apple Book cooling!!

THIS!! Tell you everything Damn nice with advertised overclock!! And +70C with single thread load!! With far from advertised OC TRASH!! And people, yeah they buy it

 

Since the new cards seem like they will be between the performance of the chip they're based on and the chip below it, it confuses me as to why they used the naming convention they did. Like why if the new card is between the 1080 and 1070 in performance would they call it a 1080[Q?] instead of 1070 Ti? Even if slightly less accurate for the silicon, it gives the customer a better idea of where they can expect it to sit performance-wise.

 

Maybe call it for what it is, and what we get... GTX1080 FRANKENSTEIN MOBILE

 

It comes from the nomenclature NVIDIA uses to differentiate the power states:

Max- Quality
Max- Performance

For Max- Q Notebook ODMs have to go through a certification process, which requires them to hold x-clocks for y-time while staying below z-volume. Where X,Y & Z are lower than required for the regular version of the GPU.

It's essentially a sticker that says: Hey this system can keep the GPU at certain clocks, cool and silent enough to pass the minimum specs for a 10xxN GPU.

 

Or they could call it 1080Lite

 

Yep, you're stuck with the butt growth! Arrrghhh!

 

Whatever they call it... It is and will be a mess. Gtx2080-2070-2060-2050Ti-2050-2030... Then will they have similar line but with weaker graphics... Gtx2080Lite-2070Lite and so on Or just use what they say... Max-Q. My name for this mess will anyway be FRANKENSTEIN!!

 

Well, when you put it like that it sounds way more disgusting than it is. But yeah, you're stuck with it for the moment.

 

Seen this? More names. 150MX

 

What is the point of that naming convention? My head hurts.

 

Milk more!! Give the OEM's the opportunity to push out more, thinner flimsy Apple similar TrashBooks with even worse cooling!! What else should it be? More trash out... The more money for NGreedia!! And people aka the buyers, yeah they jump on the ****y train-bandwagon.

 

Wow, that's about a 30% boost over the 940mx, I think people were hoping for 2x speed up, not 1.3x.

Those 2 performance charts don't match. One says 30% faster than 940x, and the other says 3x faster. Nvidia's messed up the collateral materials for the release.

And, it should have been named 1030mx, as that's built on the same GPU as the 1030 desktop GPU recently released.

 

That was my thinking. It really doesn't sound like they care much about their budget cards.