• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

DOOM benchmark Result Thread.


Its great to see (atleast on the face of it) some support for the title on launch from the developer - far too many studios (and even sometimes the people involved at the publisher sometimes) these days seem to go on holiday the moment its shoved out the door (not that I exactly blame them) resulting in massive problems if there is a big launch day bug, etc.
 
Mantle wasn’t what they envisioned Vulkan to be but by starting with some of the Mantle code as a baseline they cut out some of the work needed. Mantle was a starting point but Mantle itself was a very limited and a bias API. Large parts of it was changed to make it a very flexible and none bias API. But due to the large changes Vulkan no longer favors AMD and instead treats all GPU vendors as first class.

If Vulkan was just Mantle with add-ons it would be called Mantle 2.0 also if Vulkan was Mantle with addons then Tile based cards wouldn’t have benefited as much as they have. Both OpenGL and Mantle are very unfavorable towards Tile based GPU’s while Vulkan is very favorable towards Tile based GPU’s. In fact Tile based GPU 's gain more from Vulkan then AMD cards do. That just one example that proves Vulkan isn’t just Mantle with a few add-ons. Vulkan is far more then just mostly Mantle code.

Did you know that the majority of the Functional methods in Vulkan are named the same as with mantle but with VK appended as the class name. And that many of the Tile based functions were simply added to vulkan. Yes its not entirely what Mantle was to start with since they ripped out the front end and USed SPIR-V instead, but in terms of API structure and functionality it is essentially Mantle by another name.

Plus there was nothing Biased about mantle, AMD made it as flexible as possible while using GCN as the base level for what functionality the API should have. It was made to be a clean API that could be used with any hardware that supported the features required in hardware by the API, no different to how directx has been supported over the years.
 
Did you know that the majority of the Functional methods in Vulkan are named the same as with mantle but with VK appended as the class name. And that many of the Tile based functions were simply added to vulkan. Yes its not entirely what Mantle was to start with since they ripped out the front end and USed SPIR-V instead, but in terms of API structure and functionality it is essentially Mantle by another name.

Plus there was nothing Biased about mantle, AMD made it as flexible as possible while using GCN as the base level for what functionality the API should have. It was made to be a clean API that could be used with any hardware that supported the features required in hardware by the API, no different to how directx has been supported over the years.
No it wasn’t none bias. Mantle was just like DX and was heavily bias towards IMR architectures the only type of architecture AMD use. Tile based GPU’s where treated like 2nd hand citizens putting them at a large disadvantage. One of the key changes in Vulkan was a change in the pipeline, render passes, command buffers and hardware queues plus other areas to remove the bias so that all GPU are treated equally and to remove the disadvantages.

Vulkan is not just made up of code from AMD, Vulkan has a lot of got coding experience from the Metal API which is the most successful low level API so far. You make it sound like AMD did the majority of the work. What about the large amount of code from the other GPU company's?

Vulkan is far more than just essentially Mantle by another name and to call it that is a disservice to all the work that went into it by other companies and coders. Mantle doesn’t favor AMD and in fact it’s the other brand of GPU’s that gain more from Vulkan then AMD GPU’s.


"“but in terms of API structure and functionality it is essentially Mantle by another name.”"
No it’s not as the specification, API structure and functionality for Vulkan have been changed and designed to meet the needs of more GPUs than just Mantle. For example features like render pass which are critically important to the function of tile-based renderer GPU's but Mantel does not support. Vulkan is far more then just Mantle and to say otherwise is unfair towards the people who put all the hard work into making Vulkan as amazing as it is.
 
Last edited:
That's good to know, thanks. Can't wait to see Vulkan performance as well, hoping for good things. I hope AMD is really behind this new API and are working hard on it because some seem to think Nvidia are more keen on it despite it being based on Mantle.

I think Nvidia have gotten involved with it's development to ensure it is friendlier to there hardware than it would be if they didn't work on it.
It's like with DX12 Nvidia have worked on Just cause 3's DX12 patch and ensured it used DX12 stuff rasters which they fully support and did not use DX12 stuff like Async which they don't. It's a shame as we're immediately faced with a situation where we have two very different implementations of DX12 meaning it will always lean one way or the other depending on how games using it are developed which means someone will always have there performance hampered, Sadly I think the same caveat will cover Vulkan as Nvidia needs it for Maxwell & Pascal.
 
Plus there was nothing Biased about mantle, AMD made it as flexible as possible while using GCN as the base level for what functionality the API should have. It was made to be a clean API that could be used with any hardware that supported the features required in hardware by the API, no different to how directx has been supported over the years.

I thought it was very biased and had very specific requirements from the hardware which is why AMD could not get there own 6000 series architecture to work with it, It was okay with GCN 1.0 & 1.1 however AMD suggested Fury/Tonga owners (GCN 1.2) use the DX api in titles like Bf4 rather than Mantle because it did not play well with the updated GCN, and it's why Nvidia wouldn't go near it.
 
But also it seems the Ultra setting might have been more a late addition to the game.

Another test:

http://www.pcgamer.com/dooms-nightmare-graphics-image-quality-compared-and-benchmarked/

It appears Ultra requires 5GB of VRAM but only looks slightly better.

I'm running Ultra at 4K with everything maxed, including x8 TSAA and memory usage peaks at around 3.9GB. There is no stuttering, so the memory management on the Fury X is working well, 4GB is sufficient.

Lowest FPS I've encountered is this location here.

379720_20160514123854_1.png

FreeSync is a god send in this title, it's amazing the difference it makes and make 4K gaming possible on just a single GPU.

EDIT

My apologies Cat, didn't realise there was a nightmare mode image quality setting, i have not tried or unlocked that yet.
 
Last edited:
I thought it was very biased and had very specific requirements from the hardware which is why AMD could not get there own 6000 series architecture to work with it, It was okay with GCN 1.0 & 1.1 however AMD suggested Fury/Tonga owners (GCN 1.2) use the DX api in titles like Bf4 rather than Mantle because it did not play well with the updated GCN, and it's why Nvidia wouldn't go near it.

That is not being biased, that is like saying dx12 is biased because I has the same limitations as mantle in terms of hardware requirements. It was made for general compute based architectures. All of AMDs nom GCN based hardware still have a lot of fixed function hardware so were not feature capable with what they wanted to achieve with the api.

You might as well call vulkan bias as well since it has the same limitation by your standard, as well as every version of dx since older hardware could not use them.

And Nvidia not wanting in is since Mantle was being made by AMD, nothing more.

And in terms of everything pottsey said, the majority of that was already in mantle, Render pass was then added to allow Time-based renderers to be used. Ever UT thing else you mentioned was already a core part of mantle. Vulkan would have hit version 1 far sooner than it did and many vendors had demos running within a few weeks of AMD donating the mantle api. The reason it too so long to release was more due to the front end changes and using spirv more than any changes to the backend which mantle made the majority of the foundation.
 
I'm currently using a R9 380 4GB and I got a craving for some demon hunting. :) Not the fastest card out there, but it's good enough for 1080p at high settings (shadows at low). On some levels it drops down below 50 fps, but I have to say it's pretty smooth. I'm really enjoying this game, lot of fun! :D
 
I'm currently using a R9 380 4GB and I got a craving for some demon hunting. :) Not the fastest card out there, but it's good enough for 1080p at high settings (shadows at low). On some levels it drops down below 50 fps, but I have to say it's pretty smooth. I'm really enjoying this game, lot of fun! :D

I think for people with newer cards the game seems to run fine,IMHO.

Will give it a go with my GTX960 once the game is a little cheaper.
 
Think I will wait until it gets Vulkan support, hopefully 3440x1440 locked at 98fps, ultra settings on a 1080!!
 
“It was made for general compute based architectures.”
No it wasn’t. It was made for a single type of general compute based architecture. Not general compute based architectures. It didn’t work very well if at all on the other architectures.

Mantle was made for IMR general compute based architectures only and had a negative bais against all other general compute based architectures.



“And in terms of everything pottsey said, the majority of that was already in mantle,”
Yes a lot of it was already in mantle but was written in a way that favors IMR architectures and puts other architectures at a disadvantage. So the code was rewritten hence why I say Vulkan isn’t Mantle with a few changes. Mantle was the baseline but large parts of it got updated and changed to remove the bias Mantle had against other GPU architecture’s.

Mantle speed up the process of the creation of Vulkan but that doesn't mean most of Vulkan is based on Mantle. What about all the other none Mantle low level API code that was given to Vulkan and ended up in it?
 
Two posts of actual benchmarks, over a hundred of the usual AMD versus Nvidia nonsense :D

Does Doom have a built in benchmark or is there standard point in the game people are using?

I seem to be getting 120fps most of the time, everything maxed out @1440p
 
Last edited:
No it wasn’t. It was made for a single type of general compute based architecture. Not general compute based architectures. It didn’t work very well if at all on the other architectures.

Mantle was made for IMR general compute based architectures only and had a negative bais against all other general compute based architectures.




Yes a lot of it was already in mantle but was written in a way that favors IMR architectures and puts other architectures at a disadvantage. So the code was rewritten hence why I say Vulkan isn’t Mantle with a few changes. Mantle was the baseline but large parts of it got updated and changed to remove the bias Mantle had against other GPU architecture’s.

Mantle speed up the process of the creation of Vulkan but that doesn't mean most of Vulkan is based on Mantle. What about all the other none Mantle low level API code that was given to Vulkan and ended up in it?



Manlte didn't even work very well on other GCN architectures, the Fiji results were far lower than Hawaii for example.
 
Just read the link in OP, why are they so puzzled about the 5GB thing. That's how much the consoles use.

Doubt its related to consoles.

One of the things they did with the advanced "ultra high" quality settings option in RAGE (and while somewhat unrelated I think doom 3 had similar) was to disable texture compression which is largely pretty pointless if you are using a good algorithm but massively increases the actual VRAM required which is probably the same here.
 
well im on a 970 until the 1070 comes out. Im running at 3440x1440, mixed high and medium settings.

lowest been about 45fps and a high of 70fps. averaging at about 55.

No doubt if i lower the res it will run much faster but for most part its smooth!
 
it would be nice for amd to sort out the crossfire issues with doom.

mate has 295x2 cant use top settings as apparently his card doesn't have 5gb vram :p
 
Back
Top Bottom