Bf4 at 1080p ultra settings on my 780 got 100+ odd average fps with a 3570k at 4.2 so as I Said nvidia's competing cards on dx11 were just as good as mantle on amd's in bf4.
I have nothing to say about mantle helping Amd's own cards because it did.
-
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.
NVIDIA Publishes DirectX 12 Tips for Developers
- Thread starter Boomstick777
- Start date
Bf4 at 1080p ultra settings on my 780 got 100+ odd average fps with a 3570k at 4.2 so as I Said nvidia's competing cards on dx11 were just as good as mantle on amd's in bf4.
I have nothing to say about mantle helping Amd's own cards because it did.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,493
Hang on... Using anecdotal results for one half of the equation doesnt make it true, lol
Anand didnt show a 40fps jump in mantle for a 290X, so you aren't replicating their tests, so its irrelevant to use their results for one side and not the other
Anand showed BF4 going from 72 to 78fps, less than 8% improvement, on the same settings a 980 got 80fps... If you are starting from 80 then that is already above Anands 290 scores... I really cant believe... Oh I give up
Anand didnt show a 40fps jump in mantle for a 290X, so you aren't replicating their tests, so its irrelevant to use their results for one side and not the other
Anand showed BF4 going from 72 to 78fps, less than 8% improvement, on the same settings a 980 got 80fps... If you are starting from 80 then that is already above Anands 290 scores... I really cant believe... Oh I give up
There was something seriously wrong with your system if you got a 40 fps jump using Mantle. I ran Mantle on a 290X and frames were roughly even on both Mantle and DX11 (3930K @ 3.8) and then ran it on the FX and Mantle was roughly 20 fps down on DX11 (AMD no longer coding for Mantle).
I went from 80-90 dips to 120 cap pretty much solid. Is there something wrong with my system?
(I suppose I have cf though.)
You of all should see a difference in those result between working, and not working crossfire profile. Single Fury X is command processor limited on mantle around 22M.
Mantle does have some tweaks in it that improves drawcall performance at the expense off cpu overhead. That is one of the reasons mantle is slower on low thread counts. You have that same tweak in starswarm (you can toggle it on/off there).
Soldato
- Joined
- 14 Aug 2009
- Posts
- 3,262
Mantle made a difference in CPU heavy parts of the game (the weaker the CPU, the bigger the jump). Of course, reviewers don't always play the games they're testing to know this or don't care. But they "review" stuff.
Last edited:
Well it sounds like, (using the reasoning in this thread) that AMD and NVidia graphics cards are actually the same thing, seeing as they both produce nice pretty pictures on a monitor.
Before people shout this down as stupid, (which of course it is) just think about the reasoning here, it has been suggested that Mantle and DX12 are based on the code.
OK so here's is a bit of code for undoing a bolt.
Place spanner on nut.
place spanner on bolt.
hold on bolt still.
turn spanner on nut anticlockwise.
So please feel free to rewrite that code without using any of the same terms, because we cannot have it sounding similar, or looking similar in any way.
Now of course there is a massive difference of the above scenario and the coding for a low level API, but I'm sure you get the idea. What does the coding look like for DX11, or the Xbox API, I'm sure there will be similarities in there as well. It certainly doesn't mean they were all made by the same team using a single piece of base code.
Another aspect to think on, isn't NVidia the big bad evil corporation, if so why did they let AMD's Mantle spawn DX12, Vulkan and any others that might come along. Surely they wouldn't allow that to happen, being that they are so controlling and will stop at nothing to destroy and discredit AMD.
Of course either they didn't need too because Mantle isn't the same, or they are not the big bad evil corporation that they are made out to be.
Disclaimer: this post is probably a load of bovine excrement, but it is no less valid than some of the other theories being banded about in this thread.
Before people shout this down as stupid, (which of course it is) just think about the reasoning here, it has been suggested that Mantle and DX12 are based on the code.
OK so here's is a bit of code for undoing a bolt.
Place spanner on nut.
place spanner on bolt.
hold on bolt still.
turn spanner on nut anticlockwise.
So please feel free to rewrite that code without using any of the same terms, because we cannot have it sounding similar, or looking similar in any way.
Now of course there is a massive difference of the above scenario and the coding for a low level API, but I'm sure you get the idea. What does the coding look like for DX11, or the Xbox API, I'm sure there will be similarities in there as well. It certainly doesn't mean they were all made by the same team using a single piece of base code.
Another aspect to think on, isn't NVidia the big bad evil corporation, if so why did they let AMD's Mantle spawn DX12, Vulkan and any others that might come along. Surely they wouldn't allow that to happen, being that they are so controlling and will stop at nothing to destroy and discredit AMD.
Of course either they didn't need too because Mantle isn't the same, or they are not the big bad evil corporation that they are made out to be.
Disclaimer: this post is probably a load of bovine excrement, but it is no less valid than some of the other theories being banded about in this thread.
There was an nVidia driver update at some point that made quite a difference as well - with ultra settings I went from dips to 67fps in some places to holding 120fps (capped) solid with that driver.
Drivers are great when they work, I am still on 350.12 on my 780 as the new one i tried ( 355. something ) crashed every game i played, after that i could not be bothered so left this driver on.
Yes i agree the drivers Nvidia had around the 340's were top notch performance wise.
Depends on what part of BF4, some parts are not at all CPU bound while others are.
On the 290 with DX11 my Min FPS were about 70 in CPU bound parts, went up to about 110 in Mantle. (100% GPU bound)
They are about 90 on the same parts DX11 with the GTX 970.
On the 290 with DX11 my Min FPS were about 70 in CPU bound parts, went up to about 110 in Mantle. (100% GPU bound)
They are about 90 on the same parts DX11 with the GTX 970.
When I compared FPS in mantle and DX ages ago (SP that I did a run with max settings, the mini. FPS reading for mantle is wrong though, it was never that low when I was doing the play through), there was a huge difference, mantle was superior but with the 15.7 drivers and all the BF 4 patches, the difference wasn't as big, mantle still has a lower frame latency overall and higher min. FPS though, which are the most important figures. In game, I can easily tell the difference in smoothness between the 2. However, I imagine if you are using a g/free sync screen, the difference won't be as noticeable.
My latest comparison using the 15.7 drivers and all the BF 4 patches, MAX settings @ 2560x1080 doing the siege run in an empty server with a ping of 40-50:
Mantle:
DX:
And frame latencies, frametimelog001 (red line) is mantle:
My latest comparison using the 15.7 drivers and all the BF 4 patches, MAX settings @ 2560x1080 doing the siege run in an empty server with a ping of 40-50:
Mantle:
Code:
Battlefield 4 - Frame Time Analyzer Report
Logs Selected: All (Merge Logs)
FTS Selected: 2 ms
Game - FT CPU - FT GPU - FT
Average Frame Time: 13.889 13.89 13.13
StdDev (FT): 1.677 1.716 1.573
Minimum FPS: 49.764 49.116 52.329
Maximum FPS: 89.526 106.724 95.012
StdDev (FPS): 8.624 8.904 9.053
Time Spent: Game - FT CPU - FT GPU - FT
Above 200 FPS: 0% 0% 0%
Above 144 FPS: 0% 0% 0%
Above 120 FPS: 0% 0% 0%
Above 100 FPS: 0% 0.119% 0%
Above 90 FPS: 0% 0.907% 7.804%
Above 60 FPS: 93.89% 93.508% 98.449%
Above 45 FPS: 100% 100% 100%
Above 30 FPS: 100% 100% 100%
Total Data Points Read: 4190
Total Erred Points Detected: 1
Total Files Read: 1DX:
Code:
Battlefield 4 - Frame Time Analyzer Report
Logs Selected: All (Merge Logs)
FTS Selected: 2 ms
Game - FT CPU - FT GPU - FT
Average Frame Time: 14.429 14.428 13.933
StdDev (FT): 1.78 1.957 1.708
Minimum FPS: 44.823 37.154 45.351
Maximum FPS: 85.288 114.482 88.378
StdDev (FPS): 8.358 9.157 8.641
Time Spent: Game - FT CPU - FT GPU - FT
Above 200 FPS: 0% 0% 0%
Above 144 FPS: 0% 0% 0%
Above 120 FPS: 0% 0% 0%
Above 100 FPS: 0% 0.246% 0%
Above 90 FPS: 0% 0.761% 0%
Above 60 FPS: 87.448% 86.465% 92.041%
Above 45 FPS: 99.975% 99.681% 100%
Above 30 FPS: 100% 100% 100%
Total Data Points Read: 4071
Total Erred Points Detected: 1
Total Files Read: 1And frame latencies, frametimelog001 (red line) is mantle:
Last edited:
Look at the FuryX BF4 results running mantle, a good 25% slower than the 290 and similarly slower than DX11, purely because AMD don't provide updated driver support and optimization for EOL'd Mantle on Fiji cards.
that is much more telling than any performance gains you got running Mantle on a Hawaii card comparing optimized mantle drivers and less optimized DX1 drivers on a game designed to show off Mantle on Hawaii cards.
The take home message is even with mantle, and DX12, drivers are more important than the API to get the best performance. The terrible Ashes benchmark clearly showed what happens when drivers and software are optimized for one vendor or one API.
that is much more telling than any performance gains you got running Mantle on a Hawaii card comparing optimized mantle drivers and less optimized DX1 drivers on a game designed to show off Mantle on Hawaii cards.
The take home message is even with mantle, and DX12, drivers are more important than the API to get the best performance. The terrible Ashes benchmark clearly showed what happens when drivers and software are optimized for one vendor or one API.
Soldato
- Joined
- 5 Sep 2011
- Posts
- 12,876
- Location
- Surrey
Good middleware management and drivers are mutually important, sometimes very much so if exhausting sync points / draw calls
Since when was Ashes only optimized for one vendor? Can you post some sort of source/evidence to prove this or is this just some made up stuff you posted?
That is a search engine not any evidence to back up your claim. I am not going to look around to back up your specific claims. If you can't backup what you are saying in anyway don't post such things. From the bench thread i did not see that one vendor was optimised for more than another in anyway. And who is trying to troll? I asked you a simple question which i though was easy enough to answer seen as your making specific factual posts.
edit -
Here is some evidence to back up what i am saying
http://forums.overclockers.co.uk/showthread.php?t=18690671&highlight=Ashes
You can clearly see there isn't a certain vendor being optimised for more than another. So calling a bench terrible and vendor specific is only baiting and a troll attempt especially when you get called on it and you can't back up your claims.
Last edited:
You completely amaze me with your blind stubbornness
You take a single person's run with 4 cards and super clocked CPU as your argument and completely ignore various sites using different cards.
As I said, Mantle is/was 1.5 years on the market for AMD to work their magic in multigpu support. DX12 is out for couple of months and you want equally refined multigpu drivers?
Also from your link I see:
DX12 draw calls per frame-638 976
Mantle draw calls per frame-638 976
Completely different
And you DO realise that Matts 4x fury x are scoring same in DX12 as other sites are scoring with single 290x in DX12? Exactly the same: 18.5mln.
There probably is a bug/limiter on dx12 or something.
DX12 or Mantle Draw Calls has nothing to do with the GPU beyond compatibility.
DX12 is limited to 6 thread Draw Calls
Mantle is not, so DX12 does not scale beyond 6 threads while Mantle does, thats why Matt with his 16 Thread CPU score way higher.
Kaap is right they are different tho i don't think they are different in the way he thinks they are.