GTX 1060 Vs RX 480 - head to head showdown

[hominid]

Explain why the 1060 is crippled?

192 bit memory bus on the 1060 compared to 256 on the 480, 6GB on the 1060 compared to 8GB on the 480 are two very obvious deficiencies.

It makes the 480 much more future proof than the 1060, not withstanding NVidias habit of dropping support for all but the latest generation - and I forgot the multple empty promises that came to 0..

 

[akarypid]

I bought a 1060..very happy with it.

do I wish I had bought a 480..nope,either one would have been perfect for me,i bought what was in stock.

I don't care about 2 years time,i might be dead or not even gaming by then,i care about now.

the 480 may be light years ahead of the 1060 in 2 years who knows..do I care,no

just be happy with what you have got..i bet 90% of the people arguing over these cards don't even own one lol

Do as you like mate, so long as you're fine with it. I do get this. I actually understand someone going 'I live in the here and now, 1060 is better for the games I have, I got it and am enjoying them. Tomorrow is another day.'

I myself won't upgrade till Navi and I want to get the best possible experience from the card I buy NOW, but in one year's time. So I'm going with the 480 and a few FPS lower for the Witcher won't hurt me.

PLUS: I haven't played Doom yet! I resisted buying it till I get my new card. And now the Vulkan patch is out!

 

[humbug]

I can see some reviewers and people pointing increasingly to old games when benchmarking GPU's in the next few years.

Toms Hardware have already started in that trend.

 

[Mr Latte]

I am not out to convince anyone that one is better over the other. I just don't get why some are putting so much stock into DX12, when even on AMD hardware, DX11 is working better. Vulkan is part of FM's Time Spy and I got an 11% uplift from that with Async on as opposed to off, so why won't Pascal see an uplift in other Vulkan games?

People keep ignoring these basics and just keep telling me that in the future, the 480 is the better choice.

You dont take into account that early DX12 titles maybe havn't fully utilised it. Surely it has to take time for the developers to improve as well.

The 1060 seems faster in most DX12 games than the 980. Although in Tomb-Raider it isn't (according to Vortez review). Clearly some DX12 titles work differently than others. Perhaps some people just dont get why others like yourself hang on to DX11 and downplay future API other than it demonstrates Nvidia's superiority.

 

[akarypid]

The problem with Vulkan for Nvidia is that AMD gave the team Mantle which the team then developed into Vulkan, I imagine Vulkan is always going to favour AMD which is fair in a unfair way.

That's not necessarily true. Fermi used to be closer to GCN in that regard and I'm sure Volta will have some sort of hardware support for increasing the parallel execution capacity on the hardware side.

Pascal seems to me like it was a node-shrink of Maxwell, BUT with two important patches (fast preemption + dynamic load balancing) so as to have some kind of speedup on DX12/Vulkan. Not as good as AMD's but in any case, something to compete with.

Imagine NVidia sticking with Maxwell for 2016 while more and more DX12 games come out. It would've been a disaster for them.

 

[Gregster]

But we know that already. In most cases the reference RX 480 is less than 10% slower. The partner aftermarket 480s will probably close that to almost equal performance. But let's say an aftermarket partner 1060 (e.g. MSI Gaming) is still 5-10% faster.

Do you really find this so important as to recommend the 1060 over the 480 for someone who is on a 3-year upgrade cycle? People who will play more DX12/Vulkan games than DX11 games over that period? People who will sell their cards in 2018 when there's very few DX11 titles?

I really don't understand this logic.

You dont take into account that early DX12 titles maybe havn't fully utilised it. Surely it has to take time for the developers to improve as well.

The 1060 seems faster in most DX12 games than the 980. Although in Tomb-Raider it isn't (according to Vortez review). Clearly some DX12 titles work differently than others. Some just dont get why others like yourself hang on to DX11 and downplay future API other than it demonstrates Nvidia's superiority.

Ahhh now we are getting somewhere. Mr Latte is saying now what I have been saying and the fact that devs are just getting to grips with it and this is what I have been saying. Dx12, Vulkan and even Mantle takes time to get right. Give NVidia time and they will get it right, as will devs and that is that. Doom is a massive showcase for AMD and a real selling point for the 480 (they should have used that over the daft AoTS Crossfire example when showcasing the 480.

192 bit memory bus on the 1060 compared to 256 on the 480, 6GB on the 1060 compared to 8GB on the 480 are two very obvious deficiencies.

It makes the 480 much more future proof than the 1060, not withstanding NVidias habit of dropping support for all but the latest generation - and I forgot the multple empty promises that came to 0..

Spot on points that and the extra 2GB of VRAM and a 256 bit bus is indeed a better scenario for future proofing. Don't confuse early optimisations as dropping support though and I did explain it to you earlier.

 

[akarypid]

The problem with Vulkan for Nvidia is that AMD gave the team Mantle which the team then developed into Vulkan, I imagine Vulkan is always going to favour AMD which is fair in a unfair way.

No mate. It's not a matter of software. Even though Vulkan came from Mantle, DX12 did not: Microsoft simply imitated the approach/paradigm but it was a clean-room implementation. So the mighty NVidia with the better software team suddenly can't write proper software for both Vulkan and DX12?

That's not it: NVidia need a new architecture to address this. And this has already happened.

Pascal is Maxwell with hardware patches (fast context-switch + dynamic load balancing) for enabling some sort of reasonable level of parallel execution. If they had gone with a pure Maxwell node-shrink the DX12/Vulkan performance would've been embarrassing to say the least. So they released Pascal as a stop-gap to Volta which will really solve this problem for them (perhaps even better than AMD).

Async APIs allow you to keep feeding work to the card even from a slow CPU and even from multiple cores. The drivers are primitive and simply communicate the work to the card. The real performance gain/loss comes from the hardware. And GCN has better-suited hardware for this (even compared to Pascal).

 

[Marine-RX179]

Spot on points that and the extra 2GB of VRAM and a 256 bit bus is indeed a better scenario for future proofing. Don't confuse early optimisations as dropping support though and I did explain it to you earlier.

The message from Nvidia is very clear tbh, with the bus size being smaller than even the GTX760 and memory bandwidth no more than it, plus no support for SLI, Nvidia simple don't want the gaming beyond 1920 res crowd to get a card any lesser than a 1070

But the thing is even for 1920 res gamers, some would want to use DSR or Super-Sampling, and the chop on the bus size and the average memory bandwidth would definitely hurt the performance more when those features are used.

 

[melmac]

I am not out to convince anyone that one is better over the other. I just don't get why some are putting so much stock into DX12, when even on AMD hardware, DX11 is working better. Vulkan is part of FM's Time Spy and I got an 11% uplift from that with Async on as opposed to off, so why won't Pascal see an uplift in other Vulkan games?

People keep ignoring these basics and just keep telling me that in the future, the 480 is the better choice.

But you are ignoring things as well. You have said it yourself several times, that Nvidia always release their cards with full performance right out of the gate. And that AMD catching up, isn't due to Nvidia gimping performance and more to do with AMD eventually getting full performance out of their cards.

If that is true and the past is anything to go by then the 480 will improve much more than the 1060.

On the other hand, If we ignore the past history of both companies, then if you say that Pascal cards will get better in Dx12 and Vulkan because they are new and it's a new architecture, surely the same can be said for the GCN 1.4 cards. Vulkan and Dx12 are as new to AMD as they are Nvidia.

See here for what I think about Mantle: Conversation moved so fast my post was missed!!

https://forums.overclockers.co.uk/showpost.php?p=29809949&postcount=296

 

[akarypid]

Ahhh now we are getting somewhere. Mr Latte is saying now what I have been saying and the fact that devs are just getting to grips with it and this is what I have been saying. Dx12, Vulkan and even Mantle takes time to get right. Give NVidia time and they will get it right, as will devs and that is that. Doom is a massive showcase for AMD and a real selling point for the 480 (they should have used that over the daft AoTS Crossfire example when showcasing the 480.

Though you explicitly reference Mr Latte, you quoted me as well: far be it from me to put words in anyone's mouth. I'm just quite vocal when it comes to what I think, but that's the end of it. And I do try to read what others provide and understand.

 

[Feoilban]

I can see some reviewers and people pointing increasingly to old games when benchmarking GPU's in the next few years.

Toms Hardware have already started in that trend.

Ill take it you watched the Analysis from our scottish friend ?

 

[Gregster]

The message from Nvidia is very clear tbh, with the bus size being smaller than even the GTX760 and memory bandwidth no more than it, plus no support for SLI, Nvidia simple don't want the gaming beyond 1920 res crowd to get a card any lesser than a 1070

But the thing is even for 1920 res gamers, some would want to use DSR or Super-Sampling, and the chop on the bus size and the average memory bandwidth would definitely hurt the performance more when those features are used.

I consider both the 480 and 1060 to be 1080P cards. The 480 has leg room though with being able to go crossfire, so anyone cosidering 1440P and above would be wise to take that into account. I personally would recommend mGPUs though with the current state of play but others who use it might.

 

[Nate75]

is there a TL;DR answer to this topic please?

 

[humbug]

Ill take it you watched the Analysis from our scottish friend ?

I did.

 

[Gregster]

Though you explicitly reference Mr Latte, you quoted me as well: far be it from me to put words in anyone's mouth. I'm just quite vocal when it comes to what I think, but that's the end of it. And I do try to read what others provide and understand.

Yer, I was rounding up my point really to both of you and felt it answered both posts.

I also am able to read what others are saying also and Hominid and Marine make very good points regarding bus sizes, VRAM amounts and throw in lack of SLI on the 1080 and a good argument in favour of the 480. I won't be swayed on the future of DX12 though, as nobody knows what is what with either AMD or NVidia as of yet and I certainly wouldn't bank on Pascal not getting any DX12 goodness in the long run, as will AMD.

Right, gotta go to work now and an enjoybale couple of posts and well done to all for the civility and non rude replies

 

[fs123]

I am not out to convince anyone that one is better over the other. I just don't get why some are putting so much stock into DX12, when even on AMD hardware, DX11 is working better. Vulkan is part of FM's Time Spy and I got an 11% uplift from that with Async on as opposed to off, so why won't Pascal see an uplift in other Vulkan games?

People keep ignoring these basics and just keep telling me that in the future, the 480 is the better choice.

Timespy uses DX12 not Vulkan. The Async Method used by FM is not the one AMD has in hardware but is actually Pre-emption that is implemented by Nvidia.

Kronos may or may not add Pre-emption to their Vulkan API so it's a wait and see situation. AMD has seen a big boost in Vulkan and many DX12 games simply because of reduced driver overhead and not entirely because of Async Compute. Async can add another 10-20% on top of that.

The reference 480 is good enough in DX11 to achieve 60fps or more so does not matter if it is 10% slower. What does matter and what you are trying to play down is that it is better at DX12/Vulkan and can beat the 1060. The custom 480's will only extend that lead and match the 1060 in DX11.

The 480 is clearly the more future proof card that is also capable enough in curretn games but sadly we still get the usual Nvidia salemen trying to downplay it.

 

[Freddie1980]

Ahhh now we are getting somewhere. Mr Latte is saying now what I have been saying and the fact that devs are just getting to grips with it and this is what I have been saying. Dx12, Vulkan and even Mantle takes time to get right. Give NVidia time and they will get it right, as will devs and that is that. Doom is a massive showcase for AMD and a real selling point for the 480 (they should have used that over the daft AoTS Crossfire example when showcasing the 480.

There's not really a lot for Nvidia to get 'right' other then Tomb Raider (GNC regresses in this game as well) performance doesn't tank when going to DX12 from DX11 or from Vulkan to OpenGL. To put it another way Nvidia isn't faced with a problem like AMD does when a game comes out with Gameworks and cripples performance.

The only issue I see is if what Pascal cards Nvidia chooses to write Async software paths for, you have 3 cards with three different memory setups I'm not sure if Nvidia will want to invest the resources to program every DX12 game driver for all the Pascal cards.

Post Pascal I wonder if their next generation will feature a hardware Async engine like AMD cards have.

 

[adam98150]

The message from Nvidia is very clear tbh, with the bus size being smaller than even the GTX760 and memory bandwidth no more than it, plus no support for SLI, Nvidia simple don't want the gaming beyond 1920 res crowd to get a card any lesser than a 1070

But the thing is even for 1920 res gamers, some would want to use DSR or Super-Sampling, and the chop on the bus size and the average memory bandwidth would definitely hurt the performance more when those features are used.

It'll still do a decent job at 1440p, hell I've seen benchmarks where a 970 is running doom at 4k 30fps.

 

[Mr Latte]

Ahhh now we are getting somewhere. Mr Latte is saying now what I have been saying and the fact that devs are just getting to grips with it and this is what I have been saying. Dx12, Vulkan and even Mantle takes time to get right. Give NVidia time and they will get it right, as will devs and that is that. Doom is a massive showcase for AMD and a real selling point for the 480 (they should have used that over the daft AoTS Crossfire example when showcasing the 480.



Spot on points that and the extra 2GB of VRAM and a 256 bit bus is indeed a better scenario for future proofing. Don't confuse early optimisations as dropping support though and I did explain it to you earlier.

Yes but it just depends how much useful "Async Compute" becomes, sorry but having it on a hardware level is better and that gives at this time AMD an advantage. Their is no way Nvidia will not bring it on a hardware level too at a later time. You cannot rely on the "future driver" argument being something that will balance the issue or expect it to.

Just for a moment stop to think, what if the 480 was the 1080 competitor card with similar speeds. Just how much a difference would we then of seen in DOOM with Vulkan and how it would have had Nvidia rather concerned. It is only one title but still would make a rather large impact.

Doom is also not an AMD showcase title, clearly Nvidia used it at the 1080 launch and to showcase its huge performance and with Vulkan.

 

[Gregster]

Yes but it just depends how much useful "Async Compute" becomes, sorry but having it on a hardware level is better and that gives at this time AMD an advantage. Their is no way Nvidia will not bring it on a hardware level too at a later time. You cannot rely on the "future driver" argument being something that will balance the issue or expect it to.

Just for a moment stop to think, what if the 480 was the 1080 competitor card with similar speeds. Just how much a difference would we then of seen in DOOM with Vulkan and how it would have had Nvidia rather concerned. It is only one title but still would make a rather large impact.

Doom is also not an AMD showcase title, clearly Nvidia used it at the 1080 launch and to showcase its huge performance and with Vulkan.

Pascal does do ASYNC though at hardware level, it just does it differently to how AMD do it but the same effect is achieved, as can be seen from Time Spy.