Nvidia Volta Megathread

Page 7 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
if titan v is any performance metric, the ampere/volta gen will offer tremendous advantages to miners. if anything, nvidia will optimize for mining, not the opposite.

mining and crypto are going nowhere anytime soon so dont hold your breath. this is the new norm for gpus scarcity and will continue for at least the next year or two.
 


it's not about find it better for mining or not when there is "on purpose" restrictions on Ampere to make it so bad at mining by nvidia. at least this is some of the speculation that i heard. it is a wishful thinking but some people really want it to be true so miner get their hands off from gaming GPU.

about the current issues they have with supply i think it has to do with upcoming launch as well. nvidia for years have a habit of stopping the production of their chip much earlier to prevent over supplying the market and have to give official price cut later (there already rumor saying that nvidia already stop GP102 prodcution in november last year). in the past few years TSMC is the only one manufacturing nvidia GPU. but starting with pascal nvidia also use samsung albeit it is strictly limited to GP107/108 only. maybe this time nvidia will fab more GPU at samsung to keep up with demand.
 


isn't that expensive?
 


for their compute line up then yes. but their gaming line up most likely going to focus more on gaming workloads. the interesting part is if we are going to see complete redesign of pascal (like from kepler to maxwell) or just another improvement like how it goes from maxwell to "gaming" pascal.
 
Date:
Thursday , March 08, 2018
Author:
Kyle Bennett
GeForce Partner Program Impacts Consumer Choice

While we usually like to focus on all the wonderful and immersive worlds that video cards and their GPUs can open up to us, today we are tackling something a bit different. The GeForce Partner Program, known as GPP in the industry, is a "marketing" program that looks to HardOCP as being an anticompetitive tactic against AMD and Intel.
https://www.hardocp.com/article/2018/03/08/geforce_partner_program_impacts_consumer_choice

Also, talked about here:
https://www.techradar.com/news/nvidia-faces-allegations-of-anti-consumer-practices-in-the-graphics-card-world?utm_content=buffer9e16f&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer_trtw

Looks like Nvidia is might be using the old Intel "rebate" plan here.
 
nvidia need to prepare themselves for the upcoming onslaught. some people said nvidia should not resort to this kind of tactic when they have the dominance in GPU market but to nvidia this is the right time to pull this kind of move. they will not going to wait until they were in the world of hurt first before they start defending themselves . intel or AMD alone can't stop nvidia advance. but the two can joined force and put more effective pressure on nvidia. remember nvidia can't make x86 CPU. what happen if intel have much stronger GPU? worst thing can happen to nvidia is both intel and AMD lock their CPU to only work with their discrete GPU. no matter how fast nvidia GPU is if they were "gimp" on purpose by AMD and intel CPU then consumer will definitely going to move away from them. so before that happen nvidia need to make sure that their GPU will always needed by consumer. always. OEM will move where the consumer is going. remember when Qualcomm once mock mediatek "octa" core CPU? in the end they make their own Octa core CPU (snapdragon 810). they outright admit that they do it because of consumer want an octa core CPU in their phones and not their more effective quad core.

to me if consumer does not want nvidia to pull this kind of "rebate" plan then we should force both intel and AMD to open x86-64 license to nvidia so they can make their own CPU. that way we will have three company effectively competing in the same space (x86 and GPU). but intel and AMD will definitely don't want that to happen. but AMD probably will let nvidia get their shot at x86 if JHH no longer helm the company 😛
 
It would definitely be interesting for nVidia to have access to X86 and X86_64, but I don't really think they'll care that much. Their core strengths are not in CPU making, but GPUs and with ARM they have plenty CPU horsepower to feed their own GPUs in several designs.

But hey, you never really know. If nVidia was able to offer a closed platform (like Apple), they might even dominate the PC world and take it by storm.

Also, like you say renz, I don't think AMD wouldn't license X86_64, but Intel won't lose their tight grip on X86 and they've been very explicit about that fact.

Cheers!
 
The thing is the gaming market is nvidia primary cash cow. There is no pc gaming that is non x86. Take this example: we know when it comes to compute accelerator market nvidia is the king. But if nvidia only use the revenue from their professional division to fund their hardware/software development they will not get to what they have today on professional market. Currently the large part of nvidia R&D is sustained by the revenue they get from gaming market. This is one of advantage that many nvidia competitor did not have. Intel understand this so they decided to attack this source of revenue directly. those marriage between intel CPU and AMD GPU in EMIB package is touted as powerful gaming solution. If this solution become more popular then both intel and AMD can slowly erode nvidia source of revenue . There is no way nvidia will let that happen easily. Some AIB might not like it but if nvidia is going under they still able to sell AMD GPU. Heck they probably even start selling intel GPU. So for nvidia their own survival still come out on top.
 
Well, first, I do not think nVidia is in any imminent danger. This venture from Intel and AMD is just that, a venture. It's like a one night stand. Intel was mainly testing their EIMB tech, so it's not something to be worried about for nVidia. Even more, I would not be surprised if Intel will try with nVidia next.

That is to say, nVidia's current share in the gaming market is not going to shrink due to this venture. Low entry gaming is always shifting and I'm willing to say, Intel is still the dominant GPU maker in there. AMD is trying hard and maybe even take market, but not from nVidia. Until AMD comes up with a decent GPU that can fight Maxwell and Volta now, there's no danger for nVidia on any front of their dominated markets; at least, that is how I see it.

That is why I also believe they wouldn't be interested in X86, since the investment they would need doesn't translate into profit without risk. Plus, most important IPs for X86 CPU making are hoarded by AMD and Intel.

Cheers!
 
there will be no danger to nvidia in near term but they probably still think the need some kind of "insurance". just look at imagination technologies. when it comes to mobile GPU design they were probably the best. and specifically for mobile they most likely even better than nvidia. but they pretty much put their egg on one basket: Apple. while there are other soc maker did try using their GPU design it seems they did not push it hard enough. then suddenly apple decided to ditch IMG in favor of their own custom build GPU. now the future of Power VR is even more uncertain than Qualcomm Adreno and ARM Mali.

nvidia for their part will not want x86 for themselves. they already burn themselves once because of that desire. what they really want is something else other than x86 to compete or even take over x86 in desktop space. so they bank their bet on ARM. that's why nvidia really put an effort with Windows RT before. but all that is in the past now. right now they need to create a way to ensure they are still the preferred choice by gamer even if AMD and intel have competitive product on the market. and this is also the reason i think nvidia will not going to take advantage of the lead they have over AMD right now to continue milking pascal and delaying their next gen GPU to maximize profit like many like to believe. doing so will give the crack AMD (and intel) need to pin them down in future. heck this time nvidia might even seriously attacking the console market that right now dominated by AMD. at least with console market x86 is not a must have. having same architecture as PC might make things easy to game developer but even if they have to support different architecture it becoming much less of an issue for them as long as there is excellent tool for them to do the port.
 
Well, nVidia is definitely not making the same mistake as ImTech: they have a good portfolio of products that cover very profitable markets. Plus, they did have consoles before and they can go back into that market if they wish. In fact, the current Nintendo Switch uses nVidia, so they're already in the game (pun not intended).

Like I said, making CPUs is secondary to nVidia; they're only interested in displaying their GPUs in a good light, so they won't invest much into CPU-side other than what is needed for having a good GPU showing. Again, ARM fits that bill perfectly fine.

As for discrete GPUs, AMD would be the only other company that might be able to do something, but then again, AMD is still on a stubborn stance with GCN, so I don't foresee much movement there. Although, Vulkan 1.1 might help them a bit.

Cheers!
 
Yuka you hit the nail on the head. Vulkan is the great equalizer when it comes to gaming. Drivers are a big factor. They can mess with the numbers like FPS with drivers. Making certain video cards render all the objects in a screen like some very far away to draw the average frame rate down. While, other cards only render the immediate scene blurring out far away objects to increase average frame rates. There is a lot that can be manipulated this way. Witcher 3 and Nvidia cards did a good job of things like that.
 
problem is vulkan probably not going to gain much traction on windows. also while it has better performance than OpenGL it still doesn't mean the API is much easier to use. some developer will probably avoid using vulkan on windows for the same reason why they avoid using OpenGL in the past.
 
We could spends days and days debating merits of DX over OGL and now DX12 and Vulkan, and vice versa, but the crux of the matter is Developers want ease of development over costs. MS has always been, and will most likely continue to be, at the best possible position to bundle DX. Even with Android out there, OGL-ES and Vulkan are not inter-operable (AFAIK) and need a full fledged port just like you'd need from DX to Vulkan.

So, all of this to say, Vulkan just has the same uphill battle it's had until now and 1.1 won't really make that change. HLSL is great and all, but I hardly see it as a "killer feature" over DX12. Hence, GCN will still suck, most likely.

Cheers!
 
This GPP issue probably is more about nvidia and their board partner. Not saying the concern how it affect consumer is not valid (just look how nvidia restrictions on overclocking affect how overclocker can tinker nvidia cards) but for several years nvidia put more and more restriction and some of their board partner probably start feeling uneasy about this.
 
I just saw this about GPP: https://www.youtube.com/watch?v=Saqt_TXH14k

I did not know to what extent it went, but jeez... I mean, that is just bad optics all around plus being the written definition of "anti-competitive". I mean, it's a copy-paste of what Intel did to AMD! For shame, really!

This GPP stuff involves nVidia not giving out money and inventory to companies if they're not partners. That is pretty damn hurtful to consumers, no matter how you slice it.

We should create a thread to discuss GPP on it's own, since it's really ominous and it cannot be allowed to happen.

Cheers!
 
from what i can see many people come with their own interpretation at what actually the GPP are. some said it is similar to what intel done in the past some said nvidia just want separate brand on their own. but the issue is most likely about nvidia and their partner. also AdoreTV videos sometimes can be misleading even when he talk about the fact. take 5k series and nvidia 400 series for example. back then AMD actually sell quite a lot of the 5k series and take some market share from nvidia. he said how good the 5k series is and yet said majority of people only buy nvidia cards. the proof is he showed how nvidia is making money in certain quarters and AMD is not. but he never showed how nvidia lost market share at that generation and what nvidia do to take back their lost market share. just pointing the financial situation of both company at the time is a big flaw in itself to prove "people only buy nvidia" since AMD debt situation is not the same as nvidia.
 
I was thinking deeper about that GPP program yesterday, and it sounds so dumb that it's incredible nVidia would go ahead with it.

So, if Asus were to sign it, they just need to create a new gaming brand called "nGreedia" (yeah, why not? haha) and put all nVidia products in it and leave ROG for everything else. This would definitely hurt nVidia instead of helping it, so nVidia wouldn't allow it, so it has to be the way AdoredTV described it. They will take over ROG and give Intel and AMD a real hard time to promote (indirectly) their own "gaming" stuff.

Food for thought really, but I'm still on the "get rid of it" side over this GPP stuff.

Cheers!
 
Nvidia might take over ROG (or coming up with their own branding to differentiate themselves) but what stopping AMD and intel to come up with similar program with their own branding? It kind of reminds me of Asus Mars and Ares. Mars for nvidia and Ares for AMD. but for majority of board partner they probably did not have the resource to do this. As i said this kind of problem no doubt will going to have it's affect on consumer but this "battle" probably is more about nvidia vs their partner. GPP is about nvidia exert more control on their product and board partner starts getting uneasy about this control that nvidia slowly enforce over the years. Just days after the issue about GPP being publised by Hardocp Digitimes talk about nvidia right now ask their partner not to promote mining and not to sell the GPU directly to miners (the one that buy their gpu directly in large quantity from factory in china). That's directly asking board partner not to take guaranteed sale. Just look how it was with the restriction nvidia did in regards to overclocking with kepler. In the end all boaed partner still agree to nvidia terms so they did not lose their warranty from nvidia. So to me this is not a battle board partner can win. they need us. Think again why when intel make their "unfair" move before OEM maker did not make a noise about what intel did will harm consumer in the end? To me it's because intel never try to force what their partner (OEM) can or cannot do with their product. Of course most of this is just my opinion. It will be interesting to see how will this going to pan out.
 


And now MS is announcing the DirectX 12 part: https://www.anandtech.com/show/12547/expanding-directx-12-microsoft-announces-directx-raytracing

Interestingly, it says that AMD has (or will have?) "intermediate" support for the feature. Maybe one of the "hidden, yet not used" features of GCN?

And annoyingly, anything pre-Volta will be full software for it. That will suck for any nVidia owner with current gens, I guess?

Cheers!

EDIT: Typos.
 
i think anything besides volta will do it purely on software. this is done more tightly with nvidia than AMD so it reminds me of FL 12_1. nvidia have FL12_1 since maxwell and AMD did not have them until they have Vega. so i believe on AMD side they probably will not going to have that "ray tracing" hardware until Navi generation. if GCN already have the hardware for it they most likely have talk about it by now. heck even showcasing some demo running on their current GPU to prove it. but even done on hardware it might end up being expensive in performance. remember how expensive tessellation on performance on first gen DX11 GPU? so AMD not having the hardware until Navi probably will not mean much. in the mean time this thing still in prototyping stage. i'm not surprise if actual games using this feature won't be coming out until 2019.
 
Status
Not open for further replies.