High DX11 CPU overhead, very low performance.

Sorry I should have mentioned in my previous post that I had vsync enabled on GTAV. It seems to be capping me at 55 instead of 60. My 280x capped at 60. Anyways, completely smooth and doesn't drop even driving. I turned off vsync just for the hell of it when driving and fps was in the 60-70s.

 

That's even better, so a 970 was the best possible choice for your cpu imo, in terms of performance and price. You couldn't do such a thing with a 390x.

Edit: AMD just lost a potential customer, but oh well, if they don't want to fix their drivers, then f*ck them!

 

I understand this is mainly a DX11 related discussion but how does this affect DX9c in CPU bound scenarios? as I've been considering going green for higher min FPS in said scenarios (UE3 based mmo).

My system tanks to ~ 35 FPS during the final scenes of part 1 of this benchmark whereas this guy with a lower clocked CPU but with a 970 holds 10 FPS higher mins.

I've tried with HT on/off with no discernable difference and even Win 7 vs Win 10 scores came within margin of error but with same mins.

 

i bought 4K monitor and now i dont really care about CPU bottleneck my 290X/390X 1100/1500 have so much work with resolution that i can feed it with celeron 300A

 

I prefer higher framerates so mostly run my 1080p 144Hz display @ 1440p 100Hz.
In BnS (UE3 DX9 w/CPU PhysX) I get 70-100+FPS when GPU bound but I'm forced to cap at 66 to lessen the extremity of drops when CPU bound. :bang:

 

there are games which never will work 144fps even with best available HW as there is just not fast enough single core CPU to feed it

 

It appears as though AMD doesn't play well with UE3 games. I game on a UE3 based game called Americas Army: Proving Grounds and also experience similar situations compared to other users who use NVidia GPUs in the same game with similar specs. Min FPS spikes plague the game on bigger maps and sustaining 122 FPS so that it can match my monitors refresh rate is almost impossible. Kinda makes me want to switch to the green side but then again this is an outdated engine.

Hopefully, AMD reads our suggestions and analysis of what they are lacking compared to the competition. It wouldn't hurt if they added a more complete coverage of well-optimized games, not just the latest or Mantle like engines.

 

LOL, active Mantle support seemingly lasted roughly a year after it's public release and I think it's completely dead by now. uke2:
It technically gained a sort of "legacy" maintenance status before the first consumer products (both hardware and software) which it debuted with (R9 290X cards and a few -mostly Frostbite 3 based- games from late 2013 or early 2014) officially dropped their own active maintenance statuses (the cards were still in the market with frequent driver updates and the games still received frequent updates and/or even story expansion packs).

I am not sure if there were at least 30 consecutive days in history when somebody could simply install the latest driver + game patch and the Mantle mode worked just fine without additional "tweaking".
They got it released relatively quick in a very promising early state, then started to fix/optimize it but soon they seemingly dropped any kind of interest in continuing to do so (like somebody hit "The KillSwitch").
You are forced to update the game (if you want multiplayer access or play the latest story expansions), so it technically boils down to experimenting with the driver versions (like placing mantle DLL files in the game's folder for ease) but in the mean time Windows 10 came out and not only Mantle got worse (it's sort of expected if AMD stopped caring about Mantle long before the Win10 release...) but also DirectX11 got better with it (a nice surprise). So, Mantle practically became useless for most people (the exact time depends on the hardware specs and game).

 

Since i was curious about tha discussion about AMD Dx11 CPU overhead, i put my 770 againt my 390 in the hardest place in DX11 to CPU which is BF4 shanghai with 64 players.

CPU used was a i7 HT disabled at 3.5ghz.

Mantle was tested aswell for 390. Unfortunately Youtube destroy quality too much in this fast motion footage but yet performance is visible.
To those curious to see what happens in DX11 heavy CPU scenario:

https://youtu.be/-J9yqdVWIOM

 

Interesting yeah. I get roughly 100fps on ultra @ 1200p in mp. My cpu of course is clocked to 4.4 ghz. And my gpu is 1050mhz

 

Yeah -> Mantle RULEZ !! (I play wit' it all the time in Bf:4 and Hardline + CIV)
Only one thing about Mantle -> ATI/AMD shows Great Programing Skill + They give strong Foundations for New APIs to come: DX12 and VULCAN
Now we need to wait for H/W build for DX12 (i mean Full Spec !) and Games.
And we can Just now Dream about CGI like Games in the Future !

 

While it is not direct comparison. It is nice work.
It shows 2 things. GTX 770 is mostly fully utilized. And r9-390 is mostly underutilized at 75~85% load and on top of that clocks often drop down to 920MHz range.
When you play competitive online shooter games, disable power efficiency in 'Radeon Settings'.

 

390 would need a beastly overclock like 4.7Ghz on that i7 to actually utilize it properly.

But then tell, why is the Battlefront so different? Its the same engine, same overhead and 390 beats even the 970 by large margin.

 

The engine is the same, but it's not the same revision. A lot of things can be done better even with the same engine (see the Witcher 3 expansion and how they reduced draw calls on the design level).

 

Yep RULEZ!

http://support.amd.com/en-us/kb-art...are-crimson-edition-16.6.1-release-notes.aspx

 

Mantle is so perfect that it will crash your game <3 <3 <3

 

The most amazing thing I noticed is that you run CF 290x with a 850W power supply. :3eyes:

 

@PrMinisterGR it's close to the minimum according to HH (https://www.guru3d.com/articles_pages/radeon_r9_290x_crossfire_vs_sli_review_benchmarks,4.html), but it should be fine if he's not going crazy with voltage on the 3770K

 

When it comes to PSU's, one should not get one that is "just enough", but should be rated higher than what you need. The reason being is that the harder you stress a PSU to max capacity, you will have lower efficiency, you will generate a lot more heat (indicative of lower efficiency), and you will lower the lifespan of the unit.

This is why I invested in my PSU. I caught it on sale for lowered cost, and when I OC my system, my power draw is no where near max capacity, and heat production is still minimal. A PSU running at 50% capacity is going to have higher efficiency and lower heat than one running at 90%+. I wouldn't be surprised to see efficiency ratings at 90%+ when at 1/2 load. Getting back to point, the unit has lasted me a long time, even with SLI setups, and I've only had to replace the cooling fan once. Plus I prefer single rail 12v models, especially for multi-GPU setups.

 

In the PSU dep better to have in excess than be short.

That's why i have a 1600 one.