Why you should not buy AMD GPUs for DX11 games

[Monarch]

Most people think AMD's drivers are horrible. On this, and some other forums, it is generally thought that it is a myth. But is it? I've heard many complaints from people with AMD graphics cards, especially now with GTA V. Now, this is what I've experienced first hand, since I own a 290X.

 

First example, BF4.

 

 

I can confirm that this benchmark is accurate. I'm getting the same performance on my rig. Notice how R9 290X + 4670K with DX11 produce only 61 fps average. Compare that to 4670k + GTX 780, that produce 80 fps. That is a massive difference. If you look at the CPU and GPU usage, 290X is bottlenecked by the 4670K because its usage drops a lot, down to even 50% sometimes. R9 290X is perfectly capable of rendering 80 fps, as we can see with Mantle. 

So how come is 4670K able to feed GTX 780 with Nvidia's DX11, but not 290X with AMD's DX11? It's obvious that AMD's DX11 driver causes much more CPU overhead than Nvidia's. More CPU cycles used by the driver, less you have for framerate.

 

 

 

Now let's look at the GTA V.

 

 

Same thing. 4770k + 290X produce only 54 fps average, 4770K + 970 72 fps average. Here's how 290X's usage looks like.

 

 

It's obvious that AMD's DX11 driver is far less efficient than Nvidia's. You might get a faster card for less money, but you're going to need a much faster CPU to get the same framerate as Nvidia's equivalent. 

 

Next up, Alien Isolation. Compare the 2 CPU tests, one is with R95X2, and the other is with 780ti SLI. Notice how each CPU yielded more performance with NVIDIA. For example, FX 6300 with AMD 114 average, with NVIDIA 152 average.

 

 

 

 

 

Far Cry 4. Same thing.

 

 

 

 

Watch Dogs.

 

 

There's a a collection of benchmarks, as well as a discussion about this over at OCN, so head over there to learn more: http://www.overclock.net/t/1528559/directx-driver-overhead-and-why-mantle-is-a-selling-point-bunch-of-benchmarks

 

 

 

 

 

 

(...)the greatest gains and the highest absolute performnace are achieved by AMD’s Radeon R9 290X. As we saw in Star Swarm and continue to see here, AMD’s DirectX 11 throughput is relatively poor, topping out at 1.1 draw calls for both DX11ST and DX11MT. AMD simply isn’t able to push much more than that many calls through their drivers, and without real support for DX11 multi-threading (e.g. DX11 Dirver Command Lists), they gain nothing from the DX11MT test.

 

Source link: http://www.anandtech.com/show/9112/exploring-dx12-3dmark-api-overhead-feature-test/3

 

 

R9 290X: 674 796, GTX 970: 1 273 203. Almost 100% difference. 15.4 beta and 350.12. 

 

 

 

 

[Thebman712]

amd is doing a weird thing with their drivers called omga drives and are not supporting games as well

 

Plus, its day one bro, wait a week and everything will be diffrent.

[Ace Lean]

Well its hard to argue against numbers..

[Monarch]

amd is doing a weird thing with their drivers called omga drives and are not supporting games as well

 

Plus, its day one bro, wait a week and everything will be diffrent.

 

AMD will not optimize their DX11 driver. It has nothing to do with games, but with AMD's implementation of DX11.

[GBELILREESE]

I'd get Nvidia but i can't afford that shit. 

 

Sooo......

[Monarch]

I'd get Nvidia but i can't afford that shit. 

 

Sooo......

 

YOLO

 

Well at least you know what you're getting into with AMD.

[kuzimins]

No, I'm not going to keep quiet, because the truth has to come out. This is not to cause fanboy war, but to explain what you are dealing with if you go with AMD graphics. No one talks about this. If I had known how inneficient AMD's driver is, I would've never purchased their GPU.

You can hope to change minds of those who havent got amd gpu's yet, but no hope for those who do. Not here, at least

[Raudi_]

Seeing as I currently run an AMD GPU I can tell you this is bullshit.

[GBELILREESE]

Well, it's not like I'm going to get 20 fps with the games I'll be playing. 

 

Its not like I'm getting 20 fps every game I play, lel

[Thebman712]

AMD will not optimize their DX11 driver. It has nothing to do with games, but with AMD's implementation of DX11.

look at the gaps, they are 1-2fps most of the time, its day one and nvidia has better day one drivers. THERE WILL BE IMPROVEMENTS. the 970 is a better card than the 290x.

 

but most of it is day one issues, look at how buggy the game is right now, we cant tell how it will play out

[Majestic]

look at the gaps, they are 1-2fps most of the time, its day one and nvidia has better day one drivers. THERE WILL BE IMPROVEMENTS. the 970 is a better card than the 290x.

 

but most of it is day one issues, look at how buggy the game is right now, we cant tell how it will play out

 

The difference is, if you want that day-one experience, this is what the premium buys you.

For some that's enough, it's alteast an argument to make when buying a new card.

[jayctech]

Well... Good thing I have a 970 then.

[Monarch]

 

Well, it's not like I'm going to get 20 fps with the games I'll be playing. 

 

Its not like I'm getting 20 fps every game I play, lel

 

But you're not gonna get 60 stable either.

 

Seeing as I currently run an AMD GPU I can tell you this is bullshit.

 

You won't see this problem if you're GPU bound, i.e 1440p and above, unless you have multiple GPUs.

 

look at the gaps, they are 1-2fps most of the time, its day one and nvidia has better day one drivers. THERE WILL BE IMPROVEMENTS. the 970 is a better card than the 290x.

 

but most of it is day one issues, look at how buggy the game is right now, we cant tell how it will play out

 

You didn't get it. I'm talking about AMD's DX11 driver efficiency in general. It has nothing to do with individual games.

ANd the 970 isn't better than the 290X. They are about the same when neither is bottlenecked.

[Thebman712]

The difference is, if you want that day-one experience, this is what the premium buys you.

For some that's enough, it's alteast an argument to make when buying a new card.

well amd is driving toward dx12, and they will be more prepared for that.

 

hell im running a nvidia card and i am still getting issues, there are alot of bugs in the game right now and its up to r* to fix them

[GBELILREESE]

Yes, I'm pretty sure I'm going to get 60 fps. 

[Majestic]

I find this even more funny

 

 

/thread on all the AMD vs. Intel topics.

[Imakuni]

No, I'm not going to keep quiet, because the truth has to come out. This is not to cause fanboy war, but to explain what you are dealing with if you go with AMD graphics. No one talks about this. If I had known how inneficient AMD's driver is, I would've never purchased their GPU.

Sigh....

 

2 Benchmarks (with one of them being a launch title) on 1 setting isn't exactly "the truth". There are more games than those. There are more settings than those. Hell, there are more hardware specs than those!

 

Not to mention, you can get 290x for a lot cheaper than 970s. And the fact that the 970 is a much newer product, with AMD's new lineup coming very soon.

[Monarch]

 

well amd is driving toward dx12, and they will be more prepared for that.

 

hell im running a nvidia card and i am still getting issues, there are alot of bugs in the game right now and its up to r* to fix them

 

There are a lot of DX11 games now, and a lot of them will be made in the future. I didn't buy a GPU to play games in 2016/7 but now.

 

Yes, I'm pretty sure I'm going to get 60 fps. 

 

Not in CPU heavy games like BF4 and GTAV you won't.

[GBELILREESE]

How are you gonna tell me I'm not gonna get stable 60 fps without even knowing the card I'll be using, CPU, and the game and settings which I'll be using? 

[Tacitus]

I had a 970 and it was great at 1080p

Then I had a 290x, and at 1080p there is absolutely no noticeable difference whatsoever.

Now I have 2 290xs and it is beast.

Look at the high-res benchmarks, 290x crossfire is ahead of 970 sli and not far from 980 sli.

[Raudi_]

You won't see this problem if you're GPU bound, i.e 1440p and above, unless you have multiple GPUs.

No see the problem here is you have no official proof to back up these accusations your making besides some numbers on a graph that you expect use to believe. Do some realworld testing or use the API overhead test that is in 3DMark between a 290x and a 780 with a 4690k and you will see no difference even in games there is very very little difference and in most cases the R9 290x is half a frame behind with a day 1 launch. I went from running 2x GTX 770's in SLI to a single R9 280, and what is sad that is in all my games I only use the DX11 API and yet I still hold a better average and min frame rate with it then I did with the 770's.

[GBELILREESE]

Well, I'm not playing GTAV or BF4. 

[Raudi_]

I find this even more funny

 

-snip-

 

/thread on all the AMD vs. Intel topics.

That I can say is true, AMD's FX CPU's are doing terrible right now for GTA V, a friend of mine has a FX 8350 with a 970 and only holds an average of 50FPS in single player with it completely maxed out. Yet while the 280 I'm using was holding 40-45 average with the 15.3 Beta Drivers, now with the same settings with the 15.4 Drivers I'm averaging 70FPS in single player.

[SeanBond]

It's been a long time now, this isn't exactly new news. Nvidia released a dx11 update in their drivers that gave the single biggest performance boost to cpu bottlenecked games I have ever seen. We honestly haven't seen any new driver improvements since then on cards that were around when that driver came out around a year ago.

[Majestic]

It's been a long time now, this isn't exactly new news. Nvidia released a dx11 update in their drivers that gave the single biggest performance boost to cpu bottlenecked games I have ever seen. We honestly haven't seen any new driver improvements since then on cards that were around when that driver came out around a year ago.

 

Yeah, their DX11 drivers are on par with Mantle atm. Atleast, with current drawcall limitations.

 

 

That I can say is true, AMD's FX CPU's are doing terrible right now for GTA V, a friend of mine has a FX 8350 with a 970 and only holds an average of 50FPS in single player with it completely maxed out. Yet while the 280 I'm using was holding 40-45 average with the 15.3 Beta Drivers, now with the same settings with the 15.4 Drivers I'm averaging 70FPS in single player.

 

8350 + GTX 970 *cringe*