Gtx 1060

[Zenny]

I don't know if the game as a benchmark build in, so could be they used different settings.

The german site specifically states they have it raining during their benchmark run. Could be gameaxis does not have it on.

Maybe the other site is looking at minimum frames? Guess it might depend on the track as well.

[dadev]

The short of it is, Vulkan is a better default platform to develop around, and it is in Nvidia's best interest to optimize around it. Yes, I fully understand that Vulkan was based off Mantle (AMD clearly contributed in a major way).

However, even Nvidia admits/flaunts Vulkan as the next-generation, cutting edge API. What's going to happen in a few years when most modern games support/optimize around Vulkan? Will there be excuses then, or will Nvidia adapt their technology to better support it? The answer is clear, and it's not valid whatsoever to compare GPUs and then fail to acknowledge the current advantage AMD has on futureproofing.

Anyway, even if we toss out every other bit of data regarding these GPUs, the rx 480 4gb version is by far the cheapest 60fps 1080p ultra GPU on the market.

1440p makes for an interesting comparison between the 8gb 480 and the 1060 since they are similarly priced. But again, there is futureproofing to consider, and by that alone I would opt for the 480.

I'm not an AMD fanboy; years ago I had the 8800gtx and was happy with it, although it was pricey. I just think there is a more compelling argument to be made for the 480.

Putting marketing material aside, what makes Vulkan a better "default" API?

Assuming there really was such a thing as "default" API then lets look at the facts for a moment, the absolute majority of games are developed for consoles first. So the "default" modern API is either DX12 or GNM. I would also like to point out that those APIs have accompanied development tools, something I cannot really say about Vulkan.
In the odd case that I'd like to the develop for PC first, in my opinion (as a developer, not a marketing guy) there's no clear winner. DX12 has the development tools and good support from Microsoft which ensures at least decent support from vendors, while Vulkan can reach more audience. So defaulting to DX12 will cheapen development costs, while defaulting to Vulkan might net bigger revenue. It's a numbers game in which bigger revenue likely wins, but don't expect developers being too happy with an API for which development tools are far inferior.

Now, back to the "default" thing. Aside from GNM (and I'm really not going to comment more about it as NDAs explicitly prohibit discussing details) the modern APIs are fairly similar so there's really no point to pigeonhole your engine to one API. Make it general enough to adhere to principles of DX12 and Vulkan at the same time (which is not hard at all) and then work out the details separately.

Then there's the second thing you mention, "futureproofing". What is that exactly? I do not mean literally, but rather what people think of as future of graphics? Professionally, the answer to that would lie in the technologies that are going to be used in the future, rather than a simple choice of API.
For instance, will future games be made with VR in mind or continue the tradition of higher resolution on desktop? What will next-gen consoles offer? My point is, technology shouldn't fit itself to the API, it is backwards! API should fit itself to technology.
If there's a danger (and I really don't believe there is) that Vulkan and DX12 will not be able to utilize NVidia's hardware in the future, then NVidia will break off and create their own API. This was done before, once very recently by AMD with Mantle, and another time much longer ago by 3dfx with Glide. NVidia shouldn't adapt their technology to fit DX12/Vulkan, just as AMD shouldn't have (and in fact didn't) adapt their technology to fit older APIs.

[Deleted]

If there's a danger (and I really don't believe there is) that Vulkan and DX12 will not be able to utilize NVidia's hardware in the future, then NVidia will break off and create their own API. This was done before, once very recently by AMD with Mantle, and another time much longer ago by 3dfx with Glide. NVidia shouldn't adapt their technology to fit DX12/Vulkan, just as AMD shouldn't have (and in fact didn't) adapt their technology to fit older APIs.

Mantle was developed by AMD, not because DX11 misused or did not work properly on AMD hardware. It was made because there was a demand for a closer to the metal API like on the consoles.

[dadev]

Mantle was developed by AMD, not because DX11 misused or did not work properly on AMD hardware. It was made because there was a demand for a closer to the metal API like on the consoles.

How would you utilize compute queue in DX11? And no, LiquidVR doesn't count.

[Deleted]

How would you utilize compute queue in DX11? And no, LiquidVR doesn't count.

I will be honest, I don't know the ins and outs of DX. But I was under the impression mantle was developed due to demand out of the industry.

[dadev]

I will be honest, I don't know the ins and outs of DX. But I was under the impression mantle was developed due to demand out of the industry.

It was a joint interest. Developers wanted API that closer resembled the APIs on consoles, not just in usage philosophy but also in how you apply it (low overhead was key factor). While for AMD DX11 became problematic, theoretically they could cover it in drivers, but this would've introduced even more overhead and would go against the same principal I noted before, technology shouldn't fit itself for APIs (very minor stuff are an exception of course).

[Remilia]

Why not? We know Async works on Pascal, we have 2 examples of it so far.

Sort of, as I noted before and as Pascal's own white paper stated. Pascal is and isn't able to do async compute, more so it's not really capable of multi-engine all that well. It's fine in accepting asynchronous compute load but doesn't gain much when it takes in graphics + compute all that well. Pascal's GPCs can only take in one type of load, either compute, or graphics. During a context switch via preemption, if for example it takes in a graphics + compute, then a GPC could be segmented (in GP104's) for example, 3 graphics and 1 compute, however once one is done it needs the rest to finish. This is different than Maxwell where it flushed the entire thing where it did one task then context switch to the other load type.
This is different to GCN because in GCN it takes in whatever load it can fill.

Technically speaking, there's no such thing as Asynchronous Compute, it's called Multi-engine, but Asynchronous Compute became a name that's here to stay.

[Sorshen]

Strix GTX 1060 being sold for the equivalent of 500$ in Romania by the country's biggest online retailer. roflmao

[Artorius]

Strix GTX 1060 being sold for the equivalent of 500$ in Romania by the country's biggest online retailer. roflmao

$610 here. Romania is fine.

[coprax]

https://docs.google.com/spreadsheets...WVo/edit#gid=0

[Fascinate]

Did you make that spreadsheet thats a lot of work lol. I counted, of the 56 games the 480 only had a lead in 5 of them most being single digit%'s. WoW the 1060 has a 30% lead and overwatch 21%, crazy stuff! Cant wait to get a 1060 im being patient, i want a dual fan EVGA model EVGA makes the best fans they last forever (which are basically the things that determine video card lifespans).

[coprax]

Did you make that spreadsheet thats a lot of work lol. I counted, of the 56 games the 480 only had a lead in 5 of them most being single digit%'s. WoW the 1060 has a 30% lead and overwatch 21%, crazy stuff! Cant wait to get a 1060 im being patient, i want a dual fan EVGA model EVGA makes the best fans they last forever (which are basically the things that determine video card lifespans).

No, I just found it on a different board where nvidia vs. amd flamewar happened.

[Fascinate]

Hah ya, its kind of ridiculous both are good cards people should stop whining and buy the card that performs better in the games they play (i play WoW and overwatch, no brainer for me).

[Vash The Stampede]

No, I just found it on a different board where nvidia vs. amd flamewar happened.

That is very convenient and neat. BTW, where's that flame war?

[coprax]

That is very convenient and neat. BTW, where's that flame war?

german site, called www.computerbase.de

[hellhamster]

So according to the spreadsheet (assuming it's on the up and up), the reference 1060 is like 13% better than the 480 on average. So in theory, for maximum value, the 1060 is a better buy if it's less than 13% more expensive than the 480. Too bad that pricing in Europe is completely retarded compared to the USA.

Here's 2 of the supposedly better reference cards

Greece:
http://www.visionstudio.gr/prodinfo.asp?id=52500&cat=5 (ATI Radeon RX 480 8GB DDR5 PCI-E HDMI Triple-DP (Sapphire) = 302 €)
http://www.visionstudio.gr/prodinfo.asp?id=52778&cat=5 (nVidia GeForce GTX 1060 SC Gaming 6GB DDR5 PCI-E DVI HDMI Triple DP (Evga) = 358.09 €)

vs USA:
http://www.newegg.com/Product/Produc...82E16814202221 (ATI Radeon RX 480 8GB DDR5 PCI-E HDMI Triple-DP (Sapphire) = $239.99)
http://www.newegg.com/Product/Produc...82E16814487261 (nVidia GeForce GTX 1060 SC Gaming 6GB DDR5 PCI-E DVI HDMI Triple DP (Evga) = $259.99)

So TL;DR, pricing and stock is all fucked up right now, and there's still the aftermarket cards, so the better option is to simply wait until the dust settles.

[Denpepe]

Do take into account that european prices normally include VAT while USA ones don't. That and the fact that the US Dollar has been stronger in recent years compared to the Euro does not help much.

Once supply get's to normal amounts prices may settle a bit but I would not count on them to drop a lot.

[coprax]

So according to the spreadsheet (assuming it's on the up and up), the reference 1060 is like 13% better than the 480 on average. So in theory, for maximum value, the 1060 is a better buy if it's less than 13% more expensive than the 480. Too bad that pricing in Europe is completely retarded compared to the USA.

Here's 2 of the supposedly better reference cards

Greece:
http://www.visionstudio.gr/prodinfo.asp?id=52500&cat=5 (ATI Radeon RX 480 8GB DDR5 PCI-E HDMI Triple-DP (Sapphire) = 302 €)
http://www.visionstudio.gr/prodinfo.asp?id=52778&cat=5 (nVidia GeForce GTX 1060 SC Gaming 6GB DDR5 PCI-E DVI HDMI Triple DP (Evga) = 358.09 €)

EVGA SC Gaming: 289€
http://geizhals.eu/evga-geforce-gtx-...-a1479077.html

Radeon: 269€
http://geizhals.eu/sapphire-radeon-r...-a1450841.html

so this seems to be more of a greece problem.

[Deleted]

it's even more fucked up here
rx480 costs ~298-310 eur
reference 1060 can be found for 293 eur
palit aftermarket 1060 can be found for 318 eur
with decent aftermarket 1060s starting at 345eur

basically making rx480 more expensive with worse cooling and worse performance...

[Asmekiel]

it's even more fucked up here
rx480 costs ~298-310 eur
reference 1060 can be found for 293 eur
palit aftermarket 1060 can be found for 318 eur
with decent aftermarket 1060s starting at 345eur

basically making rx480 more expensive with worse cooling and worse performance...

Your country is an Nvidia fanboy!
Sorry, had to. Nvidia put a nice card against the 480, it's just the vendors being weird with their pricing.