Why are there so many AMD shills here? Everyone knows nVidia

Discussing shills is not technology. PhysX is technology. Too bad literally nothing uses it.

> I'm contrarian mamay

If they weren't stupid they'd know NVIDIA and AMD are almost identical on the GPU front. They both beg for their chips to be made on the Fountries of others because they don't have their own and as a result they are both overpriced.

The only reason that NVIDIA shows a difference is artificial, they simply make much bigger dies and therefore priced higher because it's easy to do that on GPUs, they are by definition parallelized.

But try to do the same on CPUs and the reality is shown. Intel destroys AMD because they have their own Foundries. AMD and NVIDIA are both fucked before they get their own Foundries.

>>58704721
Maybe it's because Nvidia is anti-competitive and cunty in the way they try to artificially lock technologies to Nvidia.

>>58704721
>Why are there so many AMD shills here?
There are?

>Everyone knows nVidia makes the best hardware
Actually, they make the best software (firmware and drivers). On the hardware front, especially for consumer grade stuff, it's pretty close. It's just that AMD fucks up their software stack over and over again.

>even /v/ agrees with me
Nobody cares what /v/ thinks, they probably don't even know how to program CUDA.

>and they know what they're talking about since they actually use their GPUs instead of ricing out loonix desktops.
See above. They play games, that's not using their GPUs.

>>58705088

I don't see much difference in software either. My R9 290 is extremely solid for years now. The main """""""""""""""""""""""""""""""difference"""""""""""""""""""""""""""""" they have is that NVIDIA orders bigger dies that cost more, which isn't any revolutionary idea, AMD could do exactly the same but they believe NVIDIA overdoes it on overpricing and they might have a point.

GPUs are extremely simple machines to make faster because they work by definition in parallel. AMD knows very well how to make gigantic dies and "destroy" NVIDIA but they'd be priced $1000.

CPUs is where it's at to see who really has technological advantages because single thread performance is still the king. Intel destroys them both because they have Foundries.

File: eMi74pw.jpg (219KB, 450x800px) Image search: [Google]

219KB, 450x800px

>>58704721
People hate Intel and that hate transforms into love of AMD and when they start comparing GPUs since people are already invested in AMD they go into defensive mode ignoring the fact that Nvidia is way better than Intel.

>>58704881
>>58705126
GP106 is ~15% smaller than Polaris 10 and is pretty competitive with it.
Nvidia are a bunch of yellow kike cunts, but Maxwell/Pascal is actually a decent architecture.

Fuck GoyWorks though.

>>58705126
It's not about the die size alone, that would simply make the GPU use a fuckton more power. Performance per watt is an issue, no matter how many parallell cores you have. AMD actually have more raw ALU/GFLOP for the same price levels, but async compute + scheduling is where it's at.

Nvidia GPUs have a superior scheduler by far.

>>58705135

Let's look at the main meme cards of both today,

AMD RX 480 Die Size: 232mm^2
GeForce GTX 1080 : 314mm^2

That's a hell of a difference ........in size only.

The notion they are all different is a stupid meme. They are almost identical technologically. NVIDIA just makes more parallelism which is extremely straightforward to do, AMD knows exactly how to do it, but they know they be too high priced too.

The only good message out of this is that at least we know they are honest about their costs because AMD wouldn't be stupid and not do the same, bigger dies.

It's all side-effect of them not having their own Foundries. They cost a shitton of money. We're talking in the vicinity of billions for a single upgrade.

>>58705184

see >>58705191

They have differences of course, but it's a meme they are that huge, GPUs are extremely parallelized and very easy to be parallelized further (VERY unlike CPUs) so "if you can't make it faster make it bigger" is a very easy thought in the GPU world.

>>58704721
>false consensus and strawmen being an argument

>2017

>>58705204
>more parallelized
Now this is the real meme, anon.

As I said, it's not simply about making it bigger. I already pointed out that AMD cards in the same price range yields more GFLOPs, that's not where the performance boost comes from.

>>58705233

Are you fucking kidding me that it's a meme? The very definition of how the GPU pipelines is in parallel. This has nothing to do with CPUS which are almost in reverse.

It's the very reason that sticking 2X GPUs together give at least 70% or more without much work from the programmer, something unheard of in the CPU world.

Remove the overheads, lag, bugs of the dual-GPU mess, and a bigger die goes to almost +96% with a single chip that is just double (as an example).

>>58705191

Look at this guy talking like he can make gpus himself. What a memer.

>>58705286

It's not anything hard to get. CPUs are extremely hard to make parallel and then get more performance because it requires a lot of work from the programmer, it requires software paradigms that need it, and by definition most software paradigms can't even use it that well to begin with, especially in interactive applications like gaymes.

In GPUs, it's the definition of how the Pipeline itself works for graphics. It's not just something that happens to work. It's exactly how the technology exists to begin with, in full parallelism.

You stick more of those transistors together, the programmer does fuck all, and you get enormous advantages. Do the same in CPUs and you are fucked, you'll get almost nothing, hence Intel.

>>58704721
nvidia doesn't make the best hardware.
they just make it up with software.
amd has more raw power but awful gayman performance.

>>58705366

Memes. NVIDIA has by definition more "machine power" on the main meme cards 4chan cares about. see >>58705191

It's pretty obvious on the software front they are not that much different.

There might be differences in the Enterprise GPUs but this is 4chan.

File: kek40.png (249KB, 358x358px) Image search: [Google]

249KB, 358x358px

>tfw my R9 390 destroys the GTX 970 and even comes very close to the 980

>>58705383
> my top of the line (at its time) card is better than previous generation almost-top one, and even comes very close to the previous generation top one
They are not housefire starters though.

>>58705380
RX 480 ($200)
5,834 GFLOPS

GTX 970 ($400)
3,494 GFLOPS

GTX 1060
3,953 GFLOPS

GTX 1080
8,228 GFLOPS

>>58705404
>house fire meme
only thing Novideo faggots can comeback is old memes, how are you liking your EVGA fire starter, or the 3.5 gigs of vram

>>58705426
old cards, old memes

what really matters is fps and framerate consistency, and all issues were fixed when 970 was relevant

>>58705417

I don't see anything surprising. Also, bigger dies fail more. Generally those Foundries have large failure rates and they include it in your price.

So, when NVIDIA orders bigger dies, they also order higher failure rates, and the consumer has to pay for them.

>>58705417
but still there are games where a rx480 is slower than a 970/1060