nvidia can do async nvidia doesnt regress kek

File: qrOyVG6.png (51KB, 1053x449px) Image search: [Google]

51KB, 1053x449px

>>55679576
kek

File: 56qckDx.png (35KB, 1053x278px) Image search: [Google]

35KB, 1053x278px

>>55679582
kek

File: EXdsnbr.png (35KB, 1047x278px) Image search: [Google]

35KB, 1047x278px

>>55679588
keek

File: shutitdown3.jpg (59KB, 300x477px) Image search: [Google]

59KB, 300x477px

>MAD loses in game made specifically for their gpus

>>55679576
I have to admit, the 1060 result for AOTS was quite surprising.

>>55679582
Nice to see that Nvidia can still gain extra performance from Vulkan.

>>55679576
>nvidia can do async
Nvidia can do async.
>In fact, next to the addition of GDDR5X, I’d consider the changes to work scheduling to be the other great change to the overall Pascal core architecture. With Pascal, NVIDIA has significantly improved their ability to allocate and balance workloads, which in turn has ramifications in several difference scenarios. But for the AnandTech audience the greatest significance is going to be in what it means for work concurrency when using asynchronous compute.
http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/9

It can but not as good as AMD's.

File: frogmech2.0.jpg (84KB, 800x655px) Image search: [Google]

84KB, 800x655px

Reminder that nvidia also has frogs

File: Screenshot_2016-07-21-10-04-42.png (498KB, 1920x1080px) Image search: [Google]

498KB, 1920x1080px

>tfw DOOM only runs better on amd hardware if you're using a top end i7 with hyper threading
>1060 blows 480 out the water when using actual hardware a lot of people buying these cards will be using

Welp, now we know that the 1060 is undoubtedly the superior card for the market it's being sold to. Check mate amd. Spread this, btw.

>>55679975
>i5 750
Kek who the fuck uses that

>>55679975

>tfw DOOM only runs better on amd hardware if you're using a top end i7 with hyper threading

I have an i5 4670 and I noticed a huge boost in performance on my fury x when switching to vulkan

4k, maxed stable 50+ fps

>>55679988
People who are poor enough to use a sub $300 card. The main point here is the 480 is only getting the massive advantages we're seeing due to reviewers using i7s with full hyper threading support or other $2000 processors which enable great scaling across multiple cores which vulkan/dx12 thrive off. When put in a situation where there are a limited amount of cores and no hyper threading support the 480 implodes. At least the truth is finally getting out.

You don't seriously believe the people who are going to buy these mainstream cards all have $2000 or i7 hypertheading cpu, right?

>>55680027
Please don't ever compare the fury x and 480 ever again. The fury x was doomed from the start with all those 4000+ cores being heavily underutilised. The 480 isn't even close to having a problem like that. What were seeing here is totally different.

>Geforce GTX1060 (1080p, D3D12) 37fps

>Geforce GTX1060 (1080p, D3D11) 39fps

K famalam.

>>55679944
or, amd is just bad in fully utilizing their resources and can finally make those idle cores work with async compute, i mean look how terrible they are in ogl and dx11

>>55679598
>nvidiots literally admit their cards are only capable of using apis from 2009

File: 1468684252750.png (56KB, 1652x1242px) Image search: [Google]

56KB, 1652x1242px

>>55679576

>>55680169
we've already established that time spy is not real dx12 or async

nvidia are just using cpu power for their preemption meme

>>55680178
>it's n-not based on AMD mantle so it d-doesn't count!

https://youtu.be/dTJPp7WdEBs

>>55680093
but you said that AMD gpus run vulkan properly only with the top end cpus which is totally not true because I noticed a huge increase in fps on my i5 4670

>>55680206
Fury x =/= 480

I doubt your story but nevertheless, the fury x has a lot more to gain compared to the 480 especially since it has more mature drivers.

And don't be a retard and say vulkan/dx12 don't require drivers. Drivers still play a massive part.

>>55680206
>>55680292

Oh and read the benchmark again. No one is saying amd doesn't get gains. We can ass it gets 20 fps extra with the i5. It just doesn't gain a massive amount like with the i7 hypertheading.

>>55680169
so 2.1% performance drop?
that's within the margin of error.

>>55680319
We can see *

File: amd dx12.png (2MB, 1450x3400px) Image search: [Google]

2MB, 1450x3400px

>mfw dx12

>>55680057
>poor people
>using 200-300$ cards
you must be retarded

I am fully expecting that over the coming weeks/month there will be driver updates or patches to Doom in particular that improve performance on the Nvidia side.
But honestly it doesn't mean anything to me because nothing I have to play has DX12 and the GTX1060 is a great card that isn't still drawing more power from the PCIe slot 12V line than it should be.

>>55680399
>people with older rigs ie non enthusiasts
>buying mainstream budget cards which make up 80% of all gpus purchased eg 960, 380, 1060, 480
>this retard somehow thinks the x60 and x80 are high end tier or some shit and not mainstream budget cards most people buy when running a toaster

File: 2832197.jpg (445KB, 763x775px) Image search: [Google]

445KB, 763x775px

>>55680169
>>55680178
>>55680324
DX12 requires seperate render paths for each vendor, something that 3Dmark didn't do. So timespy is rubbish for testing the DX12 capabilities of both Nvidia and AMD.

>>55680292
>Fury x =/= 480
Don't move the goalposts, you clearly said AMD hardware

>>55679975
>>55680027
>>55680605

All the people replying to this post are fucking retards. Read the original post again. On no place does it imply that doom doesn't get gains with vulkan which this spastic >>55680027 is trying to say. The original post clearly says it only runs better on amd hardware (vs nvidia) with an i7 but the nvidia card takes the lead with the older cpus. Nowhere was it said that doom doesn't gain anything on amd hardware. The fucking graph shows that it does. You need learn to read including >>55680206

>>55680595
>DX12 requires seperate render paths for each vendor
So does DX11, DX10, DX9, DX8, DX7, etc if you want to get the absolute best out of every card.
Not that developers actually put in all that effort all the time.

>>55680691
lol why are you so hostile?

>>55680756
You mad?

>>55680797
are you 10 years old or something?

>>55680802
Are you a failed abortion or something? Neck yourself buddy.

>>55680802
lol u mad

>>55680199
no its not based on what dx12 and vulcan dictates as async period

>>55680691
implying amd has an overhead on an api desinged for amd card
kek
literally being this dense

>>55680178
>we've already established that time spy is not real dx12 or async
"We" being AMDrones? FM has stated both AMD and Nvidia had to give the go ahead to release Time Spy and most everyone else, including noted reviewers like Anandtech, have accepted it.

>>55679975
so basically all of the benefit doesn't come from the actual card but the cpu instead

fucking ayylmao

>>55679975
>using 7 year old CPUs to benchmark hardware
seriously?

File: ayy1.jpg (34KB, 413x395px) Image search: [Google]

34KB, 413x395px

>>55679576
>Super heavily AMD optimized Shill game which exists only for Benchmarks still is beaten by a 1060 which costs the same and beats 480rx by 20 fps in DX11

AMD can't die quick enough. Also friendly reminder that Ashes don't even reach 1000 players on peak and every reviewer who even uses it is a shill paid to make AMD garbage look good. KYS.

>>55679576
apex kek

it's over, nvidia ISSSSSS toast

File: medish.jpg (83KB, 523x440px) Image search: [Google]

83KB, 523x440px

>>55679975
Doom is a buggy game. Turning ht on a 5960x cripples performance for example.

>>55683884
Buggy? No, I'd say its doing its job.

The reason its probably slow is due to the 16 threads it's managing and the CPU throttling @ 3Ghz. On top of that, its probably reaching the Amdahl's law limit on multicor performance gain.

Essentially double dipping these two would cause something like that. Probably.

>>55679598
>literally ignored every post above him to post a knee-jerk non-truth

>>55680720
>if you want to get the absolute best out of every card.
>Not that developers actually put in all that effort all the time.
but the previous API didn't clearly specified to do that. While DX12 does. In fact, if you don't do it they tell you to use DX11.

What's so difficult to understand with that ? That's the point of the criticism toward 3Dmark

>>55680346
in every thread there is that guy, that post this picture. And he never get answers, so I think it's my duty to finally give you a (you).

File: lol.png (36KB, 268x237px) Image search: [Google]

36KB, 268x237px

>>55679582
Holy shit is that sad, it already loses to a cheaper solution, dead GPU so soon.
Next months are going to be funny for the 1060, with the upcoming DX12 games. How can nVidia ever recover from that?

>>55680057
>6700k
>$2000

Hello, Nvidia

>>55679975
>Cherry picks Doom for the CPU overhead
It's not as cut and dried as that son. Apart from Doom and AotS it pretty much evens out.

http://www.youtube.com/watch?v=pjZ3MmxP0Fg

>>55679975
AMDPOORFAGS ON SUICIDE WATCH

>>55686319
>amdrones cherry pick doom as the be all and end all of benchmarks because of performance gains in vulkan (disclaimer: with high end i7's)
>amdrones suddenly write it off as being a bad game with cpu overhead when relevant benchmarks with realistic hardware don't favor them

like clockwork.

>>55683598
I can get 1060 and play modern games on my old q9550. Stay mad amdrone.

>>55680324
So is OP's:
>37 to 36
>31 to 30
>25 to 24
>39 to 33
>31 to 28
>26 to 24

>>55686379
To be honest though if you are putting a new GPU in a system that old you should probably kill yourself.

>>55686695
>need high end CPU
>need high end MB so poo in loo card won't fry your PCIe
>budget build GPU
HAHAHAHAHAHA

>>55686750
Show me a case of a fried PCI-E because of the 480.

>>55686761
Even if we disregard the PCIe problem, you still need high end build to match 1060. 480 is a failure for the price margin it's intended.

>>55686796
Not really if you are a AIB partner brand slut. I would personally want the MSI GTX 1060 Gaming rather than the cheaper partner cards. But that means I would have to pay £299 instead of £249 for a Nitro+ OC. If the MSI card had come in at the same price or even a tenner more I would have gone for it. I know I can buy other partner cards but I just don't like their aesthetics and worry how they will perform.

>>55686796
>if we ignore the part where I lied

>>55686976
>lied
kys amdrone

>>55686987
>lies
>gets caught
>gets angry

>>55686902
>he hasn't seen the beast zotac 1060 for £233

zotac are top notch when it comes to nvidia. great overclockers and brilliant cooling solutions.

>>55679975
>x4 955

I though it was Athlon, i was wrong.
now I wanna know how Athlon does with rx480

>>55683884
isn't that normal for Intel CPUs?

>>55679975
>not having a i7 6700k

lets be totally honest, this is a cpu you wont need to change for the next 10 years.

>>55679576
>1200MHz GPU competing with 2000MHz GPU

>>55687002 (You)

File: capture.jpg (45KB, 755x426px) Image search: [Google]

45KB, 755x426px

>>55679988
I'm literally using a 750 fuck you man
it's good shit

File: 1469086294876.png (17KB, 506x292px) Image search: [Google]

17KB, 506x292px

>>55679975
All this tells me is how optimized AMD's drivers are right now and that the true performance of the 480 hasn't been shown yet. If you want the card that will last then the 480 is it. I guess if you plan to sell the 1060 in a few months then go for that.

>>55687070
>>55683884


Depends on the workload - if the hardware threads are being slammed turning on hyperthreading gives negative scaling because (by definition) SMT is just a clever hardware scheduler, the performance comes from utilising unused cpu resources. In a very broad way thats how AMD's implementation of async compute works gpu wise (its more complicated than that for gpus, but whatever).

Even in an ideal scenario hyperthreading is only around 20-30% as good as an additional core. /g/ needs to learn that hyperthreading is not some pixie magic and (and does) give negative scaling at times. I wouldn't be surprised (not that I know) that hyperthreading gives poor scaling in the likes of prime 95.