GTX 1060 Vs RX 480 - head to head showdown

[Greebo]

If NVIDIA somehow use their great financial resources to make Pascal get much more performance with Vulkan then great. I'll obviously be more than happy being currently tied to NVIDIA with Gsync.

Right now it sucks compared to AMD's gains and they are currently accountable to that unless they sort it out. I have a feeling it's a hardware problem that'll be sorted with Volta in a few years.

Its not really AMD gains though.

Look at it this way. Nvidia running at 95% of the cars maximum performance in both dx11 and dx12.

Up till now, AMD have only managed to get 75% of their maximum performance of their cards in dx11 but now with dx12/vulkan they can get 95% of their max performance and be the cards they always were rather than the disappointment in games they ended up being compared to the specs.

So I never expected Nvidia to gain anything from Vulkan/dx12. It will be later gen cards that will (may) have to go down AMD's route to gain benefits from it.

Of course AMD now have had several years head start on Nvidia.

On the other hand, perhaps dx12/vulkan and async wont take off and Nvidia will find it cheaper to bury it, rather than develop a completely different chip...............

 

[hominid]

So really it just boils down to the 1060 being a bit better but a bit more expensive.

The 1060 is not better in the majority of DX12 games, and ridiculously slow in doom Vulcan - so, on the whole, the 480 is the better card.

In reality AMD should have added more than just memory to the 8GB version and branded it the 480X, then it'd be more competitive and better received.

They did, they increased the memory speed from 7000MHz to 8000MHz

Just really disappointing that x80 and x60 cards that once fought it out at the £150 price now compete at well over £200, and they call it progress.

There are multiple factors for this, primarily being the cost of the process. Each smaller node is both more expensive to produce the chips and more costly to design the architectures. The cost of a few GB's of vram have a very small, but not inconsequential, addition tot he final cost for end users.

What do you think the 1060 would have cost if it were not for the 480? Here's a hint, more than it is now.

 

[FoxEye]

Nope. What you see now is all you are going to get from nvidia on dx12.

They have just done their usual trick of using superior brute force to achieve performance.

I always see Nvidia as a big 6 litre V8 and AMD doing supercharged, twin turbo'd 1.6l engines. Eventually the small engine catches up performance wise with the big V8 lump

Well that doesn't really ring true when you look at the actual resource being consumed here - power.

I know you mean brute force in terms of algorithms, etc.

But whilst a big V8 might burn a lot of fuel comapred to 1.6, it's actually AMD who are using more juice than nVidia to achieve the performance.

So not quite such a good analogy, eh?

 

[kalniel]

Its not really AMD gains though.

Look at it this way. Nvidia running at 95% of the cars maximum performance in both dx11 and dx12.

Up till now, AMD have only managed to get 75% of their maximum performance of their cards in dx11 but now with dx12/vulkan they can get 95% of their max performance and be the cards they always were rather than the disappointment in games they ended up being compared to the specs.

So I never expected Nvidia to gain anything from Vulkan/dx12. It will be later gen cards that will (may) have to go down AMD's route to gain benefits from it.

Unfortunately this argument is buried by the gains seen by Pascal in async/DX12. Why nVidia suddenly can't get the performance out of pascal that they could out of Maxwell from the get go would be a question with no obvious answer. Occam's razor instead suggests DX12/async is actually beneficial, not simply masking an inefficiency.

 

[Seanspeed]

If NVIDIA somehow use their great financial resources to make Pascal get much more performance with Vulkan then great. I'll obviously be more than happy being currently tied to NVIDIA with Gsync.

Right now it sucks compared to AMD's gains and they are currently accountable to that unless they sort it out. I have a feeling it's a hardware problem that'll be sorted with Volta in a few years.

Guys, not every Vulkan title is going to produce the same results as Doom.

AMD cards aren't producing any better results than Nvidia in The Talos Principle, for instance.

Doom was an OpenGL titles where we know Nvidia has a big advantage, so switching away from that was going to benefit AMD more than Nvidia. OpenGL is hardly the norm, though.

There's a lot you can do with low level optimization and how much one card or brand does better than another is going to come down to a case-by-case basis. It is not some 'constant'.

 

[Greebo]

Well that doesn't really ring true when you look at the actual resource being consumed here - power.

I know you mean brute force in terms of algorithms, etc.

But whilst a big V8 might burn a lot of fuel comapred to 1.6, it's actually AMD who are using more juice than nVidia to achieve the performance.

So not quite such a good analogy, eh?

I was basing it on performance not juice used

 

[Freddie1980]

To be honest both are great cards. Overclocking wise, Tom Logan managed 2100mhz with the strix. The 480 we really need to see if AIB cards make any difference.

The 1060 is the clear winner on DX11. RX480 currently winning on DX12.

Will we see any major DX12 implementations before next gen?

It would be better to wait for pascal A-sync support before jumping to conclusions on pascal Vulkan performance though.

Battlefield 1, the new Deus Ex game, Civ 6 are all on the way out this year.

 

[akarypid]

Why stay away from the 1060? If you are saying that because of DX12, do you not think that NVidia will get to grips with it?

I can't say that with 100% certainty, but my sense is yes, so that's my opinion.

Here's the thing (and I honestly am trying to get this straight as I've put money down for a Nitro+):

There's DX12 as an API. This is all the multi-core/multi-engine/multi-GPU etc stuff and the new paradigm.

Then, there's DX12 features (the stuff that MS packages into feature levels and even moves around back and forth).

In the former (the API, not the features), AMD will always get a 5-10% advantage as demonstrated several times already (and yes, it's only up to 10%, the rest of the gains have nothing to do with the ability to run async in parallel instead of using preemption). It's not huge (like some people make it out to be), but it's significant (especially for otherwise closely matched cards such as the 1060 and 480) and it's something all DX12 games will have.

The second thing that leads me to picking the 480 is its extra power: the 1060 is fewer teraflops worth of power, which doesn't show in DX11/OpenGL. But the DX12/Vulkan titles can and will tap into it. The AMD drivers will no longer stand in their way.

In a price-sensitive segment, one's got to think about resale value down the line and stuff like that. Two years from now, my bet (and that's why I put my money for a 480) is that I will be able to sell it for more in order to upgrade. Plus I pay a little less now. Sure I've got to tolerate 10% fewer FPS in DX11 titles (not even that much with my non-reference card), but I can live with that for a year.

Where it gets complicated is with features.

For example, DX12_1 feature set contains Conservative Rasterization and Rasterizer Ordered Views. This stuff is currently only in NVidia cards as AMD don't implement them (and have not found anything hinting whether and when they intend to). Once games start utilising these (I think RotTR does) they will give some advantage to NVidia. How much remains to be seen, but you can't rule out the possibility for something as significant as async and the 10% AMD gets from that. It also remains to be seen whether AMD will at some point support them.

A similar thing exists with DX12_0 where NVidia is lacking in support but AMD is already providing that stuff so that's actually another plus for AMD.

In any case, all indication is that the more DX12/Vulkan games arrive, the less attractive the 1060 will look over the 480. That's my main bet.

 

[D.P.]

I can't say that with 100% certainty, but my sense is yes, so that's my opinion.

Here's the thing (and I honestly am trying to get this straight as I've put money down for a Nitro+):

There's DX12 as an API. This is all the multi-core/multi-engine/multi-GPU etc stuff and the new paradigm.

Then, there's DX12 features (the stuff that MS packages into feature levels and even moves around back and forth).

In the former (the API, not the features), AMD will always get a 5-10% advantage as demonstrated several times already (and yes, it's only up to 10%, the rest of the gains have nothing to do with the ability to run async in parallel instead of using preemption). It's not huge (like some people make it out to be), but it's significant (especially for otherwise closely matched cards such as the 1060 and 480) and it's something all DX12 games will have.

The second thing that leads me to picking the 480 is its extra power: the 1060 is fewer teraflops worth of power, which doesn't show in DX11/OpenGL. But the DX12/Vulkan titles can and will tap into it. The AMD drivers will no longer stand in their way.

In a price-sensitive segment, one's got to think about resale value down the line and stuff like that. Two years from now, my bet (and that's why I put my money for a 480) is that I will be able to sell it for more in order to upgrade. Plus I pay a little less now. Sure I've got to tolerate 10% fewer FPS in DX11 titles (not even that much with my non-reference card), but I can live with that for a year.

Where it gets complicated is with features.

For example, DX12_1 feature set contains Conservative Rasterization and Rasterizer Ordered Views. This stuff is currently only in NVidia cards as AMD don't implement them (and have not found anything hinting whether and when they intend to). Once games start utilising these (I think RotTR does) they will give some advantage to NVidia. How much remains to be seen, but you can't rule out the possibility for something as significant as async and the 10% AMD gets from that. It also remains to be seen whether AMD will at some point support them.

A similar thing exists with DX12_0 where NVidia is lacking in support but AMD is already providing that stuff so that's actually another plus for AMD.

In any case, all indication is that the more DX12/Vulkan games arrive, the less attractive the 1060 will look over the 480. That's my main bet.

If you are going to make a bet you should at least understand the facts first, statements like "run async in parallel instead of using preemption" are just hilarious wrong.

 

[akarypid]

If you are going to make a bet you should at least understand the facts first, statements like "run async in parallel instead of using preemption" are just hilarious wrong.

Educate me!

 

[D.P.]

RTM'd your post, abstrusely no need for such rubbish on this forum.

 

[humbug]

RTM'd your post, abstrusely no need for such rubbish on this forum.

You may well have felt right to RTM his post but i also would rather you didn't go into lengthy arguments about Nvidia architectures in this thread.

By all means do that if you want but do it in your own thread.

 

[Griffildur]

RTM'd your post, abstrusely no need for such rubbish on this forum.

he does that a lot, just report him and move on. Eventually there will be a few days of peace once a ban is in place.

 

[0h_no_not_again]

Educate me!

Where's my popcorn?!?

But seriously, I'm curious as well. It is very difficult to get the truth of this matter through the insane noise that is the internet/forums/reddit...

EDIT: Hm, maybe not relevant to this thread, so linky for those inclined? Thanks!

 

[akarypid]

RTM'd your post, abstrusely no need for such rubbish on this forum.

Let's keep things civil. I'm all for discussing this but if we're going to have a debate on this we should really use a different thread as the focus here is whether someone prefers 1060 or 480.

 

[D.P.]

Educate me!

http://stackoverflow.com/questions/...currency-parallelism-and-asynchronous-methods

For starters, Asynchronous doesn't mean parallel. In fact, the most common examples of asynchronous execution occur in strictly in single-threaded applications, for examples the asynchronous execution of javascript code within a single thread.

here is some info on async in pascal:
http://www.bit-tech.net/hardware/graphics/2016/06/15/evga-geforce-gtx-1080-ftw-review/6

 

[fs123]

Educate me!

I wouldn't bother with D.P. He will only educate you on why AMD is rubbish and Nvidia is the best.
If you want proper knowledge of how all the async/pre-emption stuff works then go to Overclock.net.

http://www.overclock.net/t/1572716/directx-12-asynchronous-compute-an-exercise-in-crowd-sourcing

 

[akarypid]

http://stackoverflow.com/questions/...currency-parallelism-and-asynchronous-methods

For starters, Asynchronous doesn't mean parallel. In fact, the most common examples of asynchronous execution occur in strictly in single-threaded applications, for examples the asynchronous execution of javascript code within a single thread.

here is some info on async in pascal:
http://www.bit-tech.net/hardware/graphics/2016/06/15/evga-geforce-gtx-1080-ftw-review/6

I get that. You submit a request, get a callback. All that you said is correct and so is the link. From a client point of view it's all about asking for something and moving on to do other things while that something gets done. You can do it from a single thread like you said, or from multiple threads (if you are already multi-threaded).

But I'm not talking about the client side of things.

I'm referring to the driver+hardware on the other side that actually runs asynchronously.

AMD can dispatch work from driver to hardware in parallel, whereas NVidia cannot: it can only interrupt one thing for another very fast (in Pascal).

So that where you get the steady 5-10% from in favour of AMD.

It's two legitimate ways to implement an async service. Both are valid. One just generally runs faster than the other in most cases (not all).

 

[Gregster]

Give over Simon. Whilst D.P might not always be correct, he knows his stuff and far more than me and you put together.

 

[akarypid]

I wouldn't bother with D.P. He will only educate you on why AMD is rubbish and Nvidia is the best.
If you want proper knowledge of how all the async/pre-emption stuff works then go to Overclock.net.

http://www.overclock.net/t/1572716/directx-12-asynchronous-compute-an-exercise-in-crowd-sourcing

Thanks for this. Glimpsed over it and it's the kind of technical depth I was looking for.