Ashes of the Singularity comes with a canned benchmark which makes it easier to compare the performance delta between DX11 and DX12, though actual gameplay may differ in performance it does make things much easier. [H]ard|OCP set the graphics to Crazy and tried out the two top cards from NVIDIA and AMD in both APIs and found some very interesting results. The AMD cards performed well above expectation, the Fury X happily sitting at the top of the pack but the 390X was more impressive, matching the performance of the 980 Ti. The AMD cards also increased in performed when running underDX12 compared to DX11, a feat the NVIDIA cards were not able to replicate.
It is still early days for the new DirectX and we should expect to see performance changes as drivers and game engines are refined but for now if you are looking to play this new RTS AMD is the way to go. Check out the full performance details as well as VRAM usage in [H]'s full review.
"The new Ashes of the Singularity game has finally been released on the PC. This game supports DX11 and the new DX12 API with advanced features. In this Day 1 Benchmark Preview we will run a few cards through the in-game canned benchmark comparing DX11 versus DX12 performance and NVIDIA versus AMD performance."
Here is some more Tech News from around the web:
- Baldur’s Gate – Siege of Dragonspear Out Now @ Rock, Paper, SHOTGUN
- Wot I Think: Dark Souls III @ Rock, Paper, SHOTGUN
- Bethesda releases Fallout 4 Wasteland Workshop trailer video @ HEXUS
- Someone Is Trying To Fix Aliens: Colonial Marines @ Rock, Paper, SHOTGUN
My favorite part of the
My favorite part of the article…
“One other thing before we conclude, AotS was not 100% stable for us in testing. We had crashing issues on both AMD and NVIDIA GPUs in both DX11 and DX12.”
move along…. nothing to see here
In other news… i read that they Dark Souls 3 might actually run best on the PC?!?!? can it be true… biggest story of the year that a AA1/2-A game actually runs best on PC at launch? Someone should verify if this is true.
I think in terms of good,
I think in terms of good, honest implementation of DX12/Vulkan API Ashes is going to the “best case scenario” for at least a year.
I mean look at Tome Raider for example, DX12 saw no performance gains because incompetent developers have to learn how to manage GPU memory which was a task that was given to DX11 to (inefficiently, but good enough) perform.
I think for most games that were built with DX12 API in mind at most you could expect a 10% performance increase from DX11 from average developers until next-gen APIs take hold of the market and developers learn how to properly utilize the tools at their disposal.
Can anyone explain why such a
Can anyone explain why such a huge lead for AMD?
The game is using a DX12
The game is using a DX12 feature called “Asynchronous Compute” something that Nvidia doesn’t support.
I thought Nvidia said that a
I thought Nvidia said that a driver was all that was needed to fix that?
No, AMD GPUs have certain AC
No, AMD GPUs have certain AC units that Nvidia lacks. Nvidia attempts to replicate it via software, effective but slow. That said, I believe the Nvidia driver just turns AC off.
Drivers can not fix the fact
Drivers can not fix the fact that Nvidia lacks the Async-compute implemented in hardware on its pre-Pascal GPUs, so Pascal is fixing some of that issue for Nvidia, but still Pascal may not be up to the same level of async-compute in the hardware level of AMD’s current GCN and Polaris(greatly improved in the GPS hardware Async-compute ability) that AMDs has had since GCN 1.0. Nvidia has improved its thread GPU processor dispatch and scheduling granularity down to the instruction level with Pascal so that alone with help on any Pascal based SKUs. Drivers can not make up for lack of async-compute fully implemented in hardware as the hardware methods are the fastest and most responsive way to manage a processor’s thread management.
It’s impossible to do GPU/GPU processor thread management in software as in the time it takes for there to even be actions to begin to fetch the code from memory, there can be thousands of thread process events that can not wait for even a memory fetch to occur let alone the necessity to execute any software based thread management code. Processor thread management is best done at the hardware level inside specialized hardware functional blocks on CPUs and GPUs that are system clocked multiplied to operate at minimal latency and have the ability to complete the management tasks faster than even one CPU’s/GPU’s instruction’s clock cycle. The functional blocks that manage the CPU/GPU processor’s threads are clocked much higher than the other parts of the CPU/GPU in order to stay ahead of the instructions being decoded/executed/fetched on any processor and that functionality can not be replicated in software that has to itself be fetched/decoded/scheduled and executed, as thousands of hardware events can occur in the hardware’s specialized thread execution blocks(many specialized blocked per CPU/GPU core).
There is also the fact that
There is also the fact that an AMD 390 actually has a greater amount of raw processing power than a 980 Ti. If DX12 allows them to make full use of that power, then it isn’t that surprising. It is surprising that they couldn’t get more performance out of AMD parts under DX11 though.
*390X that is, not 390.
*390X that is, not 390.
Async was someone AMD made,
Async was someone AMD made, it was something on gcn 1.0 cards aka 7000 series. SO can’t expect nvidia to support it properly right off the bad when it was added in to dx12 pretty late after maxwell was finalized and in production. Async is an optional part of DX12 as well.
Really, Nvidia has been
Really, Nvidia has been gimping down its hardware to sale its power usage metrics to potential customers, and Nvidia could get away with that pre DX12/Vulkan but not now, and Nvidia should have been beefing up its GPUs in the async-compute area a long time ago. Nvidia has made improvements with Pascal, but Nvidia has had no excuse for not doing this sooner other than ought-right greed on Nvidia’s part.
Nvidia has always had less compute ability on its consumer GPU SKUs at a much higher cost than AMD’s offerings. The very reason the bit-coin miners choose AMD’s GPUs where because of AMD’s better compute ability, and now even the games are accelerating and doing more non graphics gaming compute alongside the gaming compute on the GPU via DX12/Vulkan and Nvidia finds itself behind the curve once again.
No amount of your sponsored comments will defend Nvidia’s poor decision process or the over marketing of the power savings metric at the cost of performance that Nvidia and its friendly press outlets have been pushing to make the sales of its SKUs. Power usage is important but not power usage at the expense of gaming performance so Nvidia can not get away from the new DX12/Vulkan methods and the movement in the PC/laptop/devices market of doing/accelerating more gaming/non gaming workloads on the GPU.
The HPC/Workstation market never let Nvidia get away with the gimping on most of the SKUs for those markets, and Nvidia still does not use Maxwell much on server/HPC SKUs and now Pascal is fixing some of the issues, but Nvidia still has a ways to go.
I thought that myth had been
I thought that myth had been debunked already, it’s not that NVidia doesn’t support “Asynchronous Compute” as some of their cards do, it’s just AMD’s GCN is far better with “Asynchronous Compute”
My thought is that they have
My thought is that they have not implemented that yet???
Ouch, green with AC envy?
Ouch, green with AC envy?
Nice async boost for AMD vs.
Nice async boost for AMD vs. dx11 :)))) what a joke.
yep , 10% right from the
yep , 10% right from the start on cards that were not designed to run DX12 when they were conceived.
Let the jokes like this continue.
I made a choice with R9 270X and it got faster with times on the games I run ( I don’t play then new craps and trendy games ) .
So , joke you say ? Ok let the 10% free performance is a joke on me 😉
The home is on those who own
The home is on those who own cards without Async. XD
https://www.youtube.com/watch
https://www.youtube.com/watch?v=NJOyMtrb5kA
https://www.youtube.com/watch?v=0nMZ7PXTCv8
Comparison on r9 380 4g
Back when I puchased two
Back when I puchased two 7970ghz I was told 3GB of VRAM was overkill.. Where are the 680s now? Here again GCN is paying off. Seems like AMD cards get faster with time and my past Nvidia cards quickly slow down as soon as a new card comes out. Not saying there is any funny business going on with Nvidia unified drivers but Ive seen the same games get slower and slower with each new driver release when you dont have a current gen card from Nvidia.
You AMD Fag Boys are so
You AMD Fag Boys are so funny. You’re all poor as fuck, stupid as fuck, peasants. But what’s really funny, is that you’re PROUD of being poor and stupid.
verified mega asshole.
verified mega asshole.