Jump to content

Anandtech - DX12 Performance Preview

Damn it, dupe of a different thread. Please close.

http://linustechtips.com/main/topic/305586-anandtech-reviews-directx-12-star-swarm/

 

 

 

Anandtech just released a Win 10 DX 12 preview with some NVIDIA and AMD cards.

 

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm

 

From the benches, DX 12 is looking good so far.

 

71450.png

Link to comment
https://linustechtips.com/topic/305636-anandtech-dx12-performance-preview/
Share on other sites

Link to post
Share on other sites

What? This graph makes no sense to me. This shouldn't double framerates

Specs: 4790k | Asus Z-97 Pro Wifi | MX100 512GB SSD | NZXT H440 Plastidipped Black | Dark Rock 3 CPU Cooler | MSI 290x Lightning | EVGA 850 G2 | 3x Noctua Industrial NF-F12's

Bought a powermac G5, expect a mod log sometime in 2015

Corsair is overrated, and Anime is ruined by the people who watch it

Link to post
Share on other sites

it has fps for DX11, DX12, and mantle. how is that not clear?

It doesn't make sense that the performance gap is so huge. It really shouldn't have been that effective

Specs: 4790k | Asus Z-97 Pro Wifi | MX100 512GB SSD | NZXT H440 Plastidipped Black | Dark Rock 3 CPU Cooler | MSI 290x Lightning | EVGA 850 G2 | 3x Noctua Industrial NF-F12's

Bought a powermac G5, expect a mod log sometime in 2015

Corsair is overrated, and Anime is ruined by the people who watch it

Link to post
Share on other sites

If it shouldn't improve framerates, then what should it do?

I meant it doesnt make sense that it's so effective

Specs: 4790k | Asus Z-97 Pro Wifi | MX100 512GB SSD | NZXT H440 Plastidipped Black | Dark Rock 3 CPU Cooler | MSI 290x Lightning | EVGA 850 G2 | 3x Noctua Industrial NF-F12's

Bought a powermac G5, expect a mod log sometime in 2015

Corsair is overrated, and Anime is ruined by the people who watch it

Link to post
Share on other sites

I meant it doesnt make sense that it's so effective

that benchmark is designed to show how bad DX11 is with draw calls and how good mantle is at draw calls. so it shows that in a CPU/draw call limited scenario it will give that improvement.

if you want to annoy me, then join my teamspeak server ts.benja.cc

Link to post
Share on other sites

Damn it, dupe of a different thread. Please close.

http://linustechtips.com/main/topic/305586-anandtech-reviews-directx-12-star-swarm/

 

 

 

Anandtech just released a Win 10 DX 12 preview with some NVIDIA and AMD cards.

 

http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm

 

From the benches, DX 12 is looking good so far.

 

71450.png

 

DX12 main point is to have games that are dependent on CPU's be directed towards the GPU right? So that being said terrible fucking games like ARMA 3 which we all know run like crap on AMD cpus well because that game doesnt depend on your gpu as much as your cpu. My question is will i see a big improvement on those games with my AMD fx 8320? I sure hope so then i dont need to get an i5 4690k..

• FX-8320  GTX 970  M5A97 R2  Corsair H100I GT  500W PSU  RIPJAWS 8GB DDR3  SAMSUNG 840 EVO 120GB  1TB HDD 

 
Link to post
Share on other sites

DX12 main point is to have games that are dependent on CPU's be directed towards the GPU right? So that being said terrible fucking games like ARMA 3 which we all know run like crap on AMD cpus well because that game doesnt depend on your gpu as much as your cpu. My question is will i see a big improvement on those games with my AMD fx 8320? I sure hope so then i dont need to get an i5 4690k..

in short yes. the api more efficiently use the cpu so it will not bottleneck, and can use more cores/threads

if you want to annoy me, then join my teamspeak server ts.benja.cc

Link to post
Share on other sites

It doesn't make sense that the performance gap is so huge. It really shouldn't have been that effective

 

The benchmark is specifically designed to show the diffrences between high and low level API performance. AS such, results are super exaggerated

 

Kinda like an emissions test between a 1960s Diesel Engine and an electric engine rather than a 1960s and 2015 Diesel engine

CPU: AMD Ryzen 7 3700X - CPU Cooler: Deepcool Castle 240EX - Motherboard: MSI B450 GAMING PRO CARBON AC

RAM: 2 x 8GB Corsair Vengeance Pro RBG 3200MHz - GPU: MSI RTX 3080 GAMING X TRIO

 

Link to post
Share on other sites

DX12 main point is to have games that are dependent on CPU's be directed towards the GPU right? So that being said terrible fucking games like ARMA 3 which we all know run like crap on AMD cpus well because that game doesnt depend on your gpu as much as your cpu. My question is will i see a big improvement on those games with my AMD fx 8320? I sure hope so then i dont need to get an i5 4690k..

 

arma 3 runs on DX11, they need to upddate the game to dx12, or wait for arma 4

9800X3D - NH-D15S - MSI X670E  -  32GB 6000MHz CL30 - RTX 3080 FTW3 - 970evo 1TB - 980pro 1TB - 970evo 1TB - 970evo+ 2TB -  970evo+ 2TB - 850evo 250GB - WDblack 1TB - WDblue 3TB - HX850i -  27GR95QE-B.AEU - 28MQ780-B - 27GP850-B - 27GL850 - Meshify 2
Link to post
Share on other sites

I would say repost, but this OP is better.

 

Looks good.

i7 6700K - ASUS Maximus VIII Ranger - Corsair H110i GT CPU Cooler - EVGA GTX 980 Ti ACX2.0+ SC+ - 16GB Corsair Vengeance LPX 3000MHz - Samsung 850 EVO 500GB - AX760i - Corsair 450D - XB270HU G-Sync Monitor

i7 3770K - H110 Corsair CPU Cooler - ASUS P8Z77 V-PRO - GTX 980 Reference - 16GB HyperX Beast 1600MHz - Intel 240GB SSD - HX750i - Corsair 750D - XB270HU G-Sync Monitor
Link to post
Share on other sites

Is it possible to deduce anything at all from how much different result we get between AMD and nVidia cards? The difference from expected values between nVidia and AMD is big in both DX11 and DX12. Or is it purely from bad optimization of AMD drivers for this synthetic test?

Link to post
Share on other sites

That is a very large gap in performance. I honestly did not expect to see such a radical change.

|  The United Empire of Earth Wants You | The Stormborn (ongoing build; 90% done)  |  Skyrim Mods Recommendations  LTT Blue Forum Theme! | Learning Russian! Blog |
|"They got a war on drugs so the police can bother me.”Tupac Shakur  | "Half of writing history is hiding the truth"Captain Malcolm Reynolds | "Museums are racist."Michelle Obama | "Slap a word like "racist" or "nazi" on it and you'll have an army at your back."MSM Logic | "A new command I give you: love one another. As I have loved you, so you must love one another"Jesus Christ | "I love the Union and the Constitution, but I would rather leave the Union with the Constitution than remain in the Union without it."Jefferson Davis |

Link to post
Share on other sites

Is it possible to deduce anything at all from how much different result we get between AMD and nVidia cards? The difference from expected values between nVidia and AMD is big in both DX11 and DX12. Or is it purely from bad optimization of AMD drivers for this synthetic test?

People will try to, you can't though.

 

The benchmark is purely to see the comparison between DX11, Mantle and DX12. It isn't supposed to compare graphics cards.

i7 6700K - ASUS Maximus VIII Ranger - Corsair H110i GT CPU Cooler - EVGA GTX 980 Ti ACX2.0+ SC+ - 16GB Corsair Vengeance LPX 3000MHz - Samsung 850 EVO 500GB - AX760i - Corsair 450D - XB270HU G-Sync Monitor

i7 3770K - H110 Corsair CPU Cooler - ASUS P8Z77 V-PRO - GTX 980 Reference - 16GB HyperX Beast 1600MHz - Intel 240GB SSD - HX750i - Corsair 750D - XB270HU G-Sync Monitor
Link to post
Share on other sites

What? This graph makes no sense to me. This shouldn't double framerates

 

 

It doesn't make sense that the performance gap is so huge. It really shouldn't have been that effective

 

 

I meant it doesnt make sense that it's so effective

 

That feeling is that of your mind being blown at how inefficient DirectX has been up till now :P

 

You might have some side effects of realising that consoles can do what they do with limited hardware, because they run on software that is even more effient that DX12

Link to post
Share on other sites

The benchmark is specifically designed to show the diffrences between high and low level API performance. AS such, results are super exaggerated

 

Kinda like an emissions test between a 1960s Diesel Engine and an electric engine rather than a 1960s and 2015 Diesel engine

Diesel engines are actually bad in terms of emission reductions compared to petrol cars. Also the fact that DX12 is comparable to Mantel in terms of fps puts me off a bit, as in BF4 when using Mantel, it makes the game unplayable. It does increase maximum, but it cause frequent drops, and stuttering. Is this a problem that could be in DX12, or is that just a mantel thing.

Link to post
Share on other sites

It doesn't make sense that the performance gap is so huge. It really shouldn't have been that effective

Lol how do you think a game like GTA V can run on a ps3.

The difference is huuuuuge.

The only reason why games like Battlefield 4 had a small difference with Mantle is because the game was designed for DX11.

RTX2070OC 

Link to post
Share on other sites

A quote from the Anandtech article

Under DirectX 11 the bulk of Star Swarm’s batch submission work happens under a single thread, and as a result the benchmark is effectively bottlenecked by single-threaded performance, unable to scale out with multiple CPU cores. This is one of the issues DirectX 12 sets out to resolve, with the low-level API allowing Oxide to more directly control how work is submitted, and as such better balance it over multiple CPU cores.

 

 

71448.png

Yes the above is an edge case scenario with over 100,000 draw calls per frame. But it does demonstrate that even on an overclocked i7 4960X the Nvidia drivers becomes CPU limited much more easily in DirectX 11. The Nvidia DirectX 12 driver was entirely GPU limited, only becoming CPU limited when you went down to 2 cores, and even at that configuration the 980 was able to pump out double the framerate of DX11.

 

In real world applications we are going to see a plethora of different bottlenecks/complications so obviously you will not see gains of the same magnitiude. However devs seem to agree that achieving this kind of CPU load distribution was much much more difficult if not impossible using the old graphics APIs. So this is an important step forward for the industry.

 

The fact that some of you are not currently CPU limited in most games that you play does not mean that it's not important. Good optimization means doing things in the most efficient manner at every stage of the process, regardless of whether or not you can brute force it. Not to mention freeing up CPU resources means that devs can use that CPU power for other sub systems such as AI, for example think about the number of RTS and MMO games which get CPU limited so easily today as the number of units ramp up even on powerful Intel CPUs. An API which can render the same scene with less CPU time is clearly an example of superior engineering. It's clearly doing something right.

 

If DX12 and OpenGL Next can do this it is good for everyone. GPUs are becoming faster and faster at a great pace, it doesn't show any signs of slowing down. On the other hand single core CPU performance seem to have stagnated over the last 4 years both from Intel and AMD. What this trend means is that CPU bottlenecks are going to become more and more significant if we continue to do things using legacy techniques.

 

Try to look beyond how many frames per second more you can get today in BF4 and think about the industry direction...

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×