DX12 versus DX11 Gaming Performance Video Card Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,532
DX12 versus DX11 Gaming Performance Video Card Review - We play latest games with DX12 support and find out which is faster, DX12 or DX11? We use the latest drivers from NVIDIA and AMD to find any advantages in this GPU focused review. We’ll get to the bottom of the question, "Should I be running this game in DX12 or DX11 in order to get the best real world gaming performance?"
 
I have been wondering this, do you notice any quality difference between DX 11 and 12?? Like do certain effects look better, textures, particle effects, etc. etc.
 
The problem with that thinking starts when you look at the results and DX12 is not helping anything to begin with.
 
With Sniper Elite 4, is it more likely though that the game for AMD was just better optimised for DX12 rather than DX11 and their drivers under that API?
I notice that there is roughly a 20% difference on AMD between DX11 and DX12 for this game, however the DX12 performance of the GTX1060 6G seems to match pretty well to the 480.

I appreciate some of this is beyond the scope of the article, but it does influence the context/meaning of how well DX12 is seen to improve per game; yeah at its core it is still an improvement but separately this could be down to lack of investment (resource wise) to optimise with DX11 and associated drivers both from the developer and AMD - understandable from AMD perspective as they need to manage their finite resources well and I would assume focus on DX12 over DX11.
Part of that separate context and beyond the original scope; how much effort and resources are needed to optimise the game for DX12 compared to DX11, meaning do game devs put equal amount of effort-resources for both GPU manufacturers (personally I think this can swing either way depending upon DX12 game and engine functionality/options).

Thanks
 
Maybe Fury X should be involved in this review as well? I know it's older card but still AMD's fastest and it shows the differences. Polaris is low end and I wouldn't expect much of a difference (Plus it sucks hehe). If we go back and look at some of these old reviews for Fury X it does show differences in some games, sometimes really big differences.

This is from a [H] review comparing DX11 with DX12 on a Fury X in AOTS on the 6700k.

"Under our Heavy Load benchmark with the Fury X, we see results that we are much more used to seeing. We see a huge 37.3% increase moving to DX12"

In Battlefield 1 a Fury X is about 7%-10% faster in DX12 vrs DX11.

In Deus X, the Fury X is about 3fps-5fps faster in DX12 vrs DX11.

So it appears that we have no differences at all with a little Polaris (and sometimes even slower) but with the Fury X, we have better performance and shows the differences.
 
Game Developers 2012 - DX11 is too bloated and over complicated, we need to get 'closer to the metal'!

Game Developers 2017 - Working closer to the metal is too complicated for us! We need to go back to DX11!
 
Some aspects I hoped to see in this.

1. Lower end cards tested at 1440 P with DX 12 and different CPU's to see if there is a benefit to DX12 as advertised. Will this bring the framerate up for a 1060 or 1070, or whatever the AMD series are.

2. Does CPU matter in this case.. are intel i7 6700 or 7700s better or is there a greater benifit to being on a Ryzen 8 core 16 thread system?

Just curious.
 
The entire point of DX12 was to alleviate CPU side bottlenecks. Why is it surprising that in GPU bound tests theres basically no change in performance?

I wouldn't say it's the entire point, it's a point, but not the entire point of DX12, there are other points as well. This is a common misconception.

For just one example, DX12 allows closer to the metal programming, which potentially can mean improved performance on GPUs in GPU dependent situations, compared to DX11, and that's just one example and point about DX12, there are many, many more.

The GPU dependency of DX12 needed to be tested, and I would argue is very important as you are mostly going to be in GPU limited situations while gaming.
 
Last edited:
Maybe Fury X should be involved in this review as well? I know it's older card but still AMD's fastest and it shows the differences. Polaris is low end and I wouldn't expect much of a difference (Plus it sucks hehe). If we go back and look at some of these old reviews for Fury X it does show differences in some games, sometimes really big differences.

This is from a [H] review comparing DX11 with DX12 on a Fury X in AOTS on the 6700k.

"Under our Heavy Load benchmark with the Fury X, we see results that we are much more used to seeing. We see a huge 37.3% increase moving to DX12"

In Battlefield 1 a Fury X is about 7%-10% faster in DX12 vrs DX11.

In Deus X, the Fury X is about 3fps-5fps faster in DX12 vrs DX11.

So it appears that we have no differences at all with a little Polaris (and sometimes even slower) but with the Fury X, we have better performance and shows the differences.

Polaris was used due to its latest iteration of GCN architecture, thus the best potential for any DX12/Async Compute hardware benefits.

I can appreciate that people desire more cards tested, naturally doing so takes more time, and really we got the answers we were looking for with the cards used.
 
Some aspects I hoped to see in this.

1. Lower end cards tested at 1440 P with DX 12 and different CPU's to see if there is a benefit to DX12 as advertised. Will this bring the framerate up for a 1060 or 1070, or whatever the AMD series are.

2. Does CPU matter in this case.. are intel i7 6700 or 7700s better or is there a greater benifit to being on a Ryzen 8 core 16 thread system?

Just curious.

This was never intended to be a CPU test. But we do appreciate your feedback.
 
If only developers would just flip the switch to enable DX12. Then they'd see a 20% performance boost. DX12 is going to save all! /s

I can't even count how many times I read basically that post for the last 2 years. DX12 was to be the savior of poorly optimized or low frame rate games.
 
If only developers would just flip the switch to enable DX12. Then they'd see a 20% performance boost. DX12 is going to save all! /s

I can't even count how many times I read basically that post for the last 2 years. DX12 was to be the savior of poorly optimized or low frame rate games.

Certain users will continue to say stuff like "but oh if they actually implement it correctly in the future!!!!" but so far there has been zero proof of the benefits of DX12 to end users.
 
Certain users will continue to say stuff like "but oh if they actually implement it correctly in the future!!!!" but so far there has been zero proof of the benefits of DX12 to end users.

DX12 is not a magic wand. The only limitations that DX12 has are the programmers themselves.
 
Wow, so why does DX12 suck so badly after all this time? Maybe it's time for DX13 ???

It does not suck it is just not used in an optimal way. API will take time for programmers to learn and get the most out of it , if you take an avg development time of 4 to 5 years and games that have been featured now were not made for DX12. In a few years time that cycle is over and in that time you will see better use of DX12.
 
Hi All

Nice review.

Question, Going forward which API shows the most promise? DX12 or Vulkan
 
Last edited:
DX12 is backed by Microsoft and lots of tools exist to work with it. Downside is it's only available on Windows 10. Vulkan can be used on different operating systems, including older Windows versions (7 and 8.x). Since I use different operating systems, I'd like Vulkan to take off (would be very nice to not use Windows on my home system at all finally).
 
i'd like to see how it compares on older cards.. sure i can good 'dx12 r9 390' but I mostly just get ashes of the singularity with old drivers.. seeing if dx12 breathes new life into these older cards would be nice!
 
So, what I gather from this is, DX12 in its still-early stages is about as good as a very mature DX11. Which is a positive sign for the future of DX12.
 
DX12 reminds me a lot of DX10 at this point, we've been through this before. You all know what happened to DX10, it was quickly replaced in favor of DX11, and DX11 has been great for games. I'm not sure this will happen to DX12, but no one can deny it's got itself off to a rough start. One thing DX11 really helped with was improving game visuals, image quality got better, they were able to do more with graphics. DX12 hasn't even reached this point yet, right now it's all about just making it at least perform as good as DX11, or a little better. It will take another leap to actually use the performance benefits to improve image quality with new graphical features and higher quality geometry and whatnot. It has a ways to go still.
 
Ignore the Vulcan API behind the curtain.

Wait till we get Game Mode, then you will have to run all these tests again and report a percentage increase that falls within the margin of error.

Is this the straw that loosens the death grip that microsoft has on pcgaming.
 
Ok, since I have two of these games (RotTR and DX:MD) and less than latest high-end hardware (Xeon E3-1230v2 and AMD R9 285 2G card) I think I'll run the canned benchmarks later today and see if my "lower but not necessarily low end" system shows any difference. I don't have time or setup to do the type of testing [H] normally does!
 
Interesting results. This is what I expected to be honest. DX12 helps CPU bound PC's. I would like to see you benchmark the same results with say an older horrible FX 9590 or something. Maybe even a Phenom II. That would be interesting to see if DX12 even helps.

Although that Sniper Elite results kind of surprises me.

Overall thanks for that review Brent. Was nice to see good [H] facts.
 
The entire point of DX12 was to alleviate CPU side bottlenecks. Why is it surprising that in GPU bound tests theres basically no change in performance?

That's pretty much all I heard being hyped for the longest time. So far it is still far short of what MS and others made it out to be.

The thing is though, is anybody that short of CPU power these days? I'm still running a five year-old Core i5 and still don't feel any need to upgrade. Sure, some things I would like to get done faster (A/V editing and converting in particular) but my GPU is still the bottleneck when it comes to gaming. Granted, my GPU is old too, but I bet if I ran out and bought a 1080 Ti I still wouldn't be short on CPU horsepower.
 
That's pretty much all I heard being hyped for the longest time. So far it is still far short of what MS and others made it out to be.

The thing is though, is anybody that short of CPU power these days? I'm still running a five year-old Core i5 and still don't feel any need to upgrade. Sure, some things I would like to get done faster (A/V editing and converting in particular) but my GPU is still the bottleneck when it comes to gaming. Granted, my GPU is old too, but I bet if I ran out and bought a 1080 Ti I still wouldn't be short on CPU horsepower.

I bet you are wrong. I made a quantitive leap moving from an i7 2600 to my i7 7700. In my game of choice the impact was 100fps at 1080p.
 
I really wish you would include the 1070 in your testing. Far, far more people own a 1070 than a 1080 or 1080Ti.
 
Awesome review! Quality content as always

myth-busted.png
 
if performance is basically the same using DX11 and DX12 why wouldn't you just use and recommend DX12?...
 
Thanks for the review as always. You guys rock!

I recently moved my 1440p/144hz G-Sync monitor over to my older build and tweaked the OC on the GPU's a little further. Had a few minutes to spare and did a quick bench w/ high settings, no aa, no v-sync but w/ g-sync.
The only game I have on the list is ROTTR. The 970's are now running stable on air at 1506mh core, 3903/7806mhz ram and temps holding 54-68c. Using the same driver I averaged 110fps(min ~30 max ~175. I was surprised its the best I've seen yet.
 
I would love to see a card like the 390x 8Gb card tested as to see what 512Mb memory bus looks like as that card was 980GTX area last I seen one tested as Hawaii has never stopped amazing us over time or CX 390Xs in Eyefinity using DX 12.. also what about Surround testing as we know AMD did a lot of work in that area as to play higher settings and has anyone really tested Eyefinity under DX12 under say 3 x1080s in Eyefinity .
 
also what about Surround testing as we know AMD did a lot of work in that area as to play higher settings and has anyone really tested Eyefinity under DX12 under say 3 x1080s in Eyefinity .

None of the mainstream GPU review blogs do multiple-monitor testing anymore. [H] gave it up a good while ago. Most new AAA games do support it, I believe all of the games in this test do. I bounce between 3x1080P Surround and 4k these days, depends on the game, support, performance and how it feels to me. As 3x1080P is only 75% the resolution of 4k so performance tends to be higher and if the framerates stay well over 60, I tend to still prefer the triple monitors.
 
Ok, got a chance to do my benchmarks. Really first time I've done anything like this. I did three passes of the built-in benchmarks in DX11 mode, then switched to DX12 mode. Latest Win10 and game patches. Hardware, as mentioned before is Xeon E3-1230 v2, so not overclocked; Gigabyte Windforce Radeon R9 285 at factory settings, 8GB DDR3 1600 RAM. I had the default "high" settings, but turned off lens flare, motion blur, Chromatic Aberration (DX:MD), and Film Grain (RotTR). Screen resolution is my monitor's native 1600x1200.

First, these are the settings I have played through DX:MD, and gameplay seemed fine. Probably some stuttering, but not distracting from gameplay. I played through the game in DX12 mode. I haven't played RotTR yet, but figured I could use similar settings. Second, without the numbers, the benchmark seemed a bit smoother in DX12 than in DX11.

Here are the average numbers from the three passes:

DX: MD - DX11 - Min 32.67, Max 60.40, Avg 47.57. One of the passes had an abnormally low minimum, the other two had 37fps for minimum
DX:MD - DX12 - Min 37.60, Max 58.30, Avg 47.57. Surprised me that the averages were exactly the same!
Conclusion - no perceptible difference based on the numbers between DX11 and DX12, but DX12 felt smoother.

RotTR - DX11
Mountain Peak - Min 21.85, Max 131.15, Avg 56.37
Syria - Min 6.55, Max 92.52, Avg 54.29
Geothermal Valley - Min 6.47, Max 73.86, 50.52
Overall - 53.67
RotTR - DX12
Mountain Peak - Min 41.87, Max 74.71, Avg 58.04
Syria - Min 29.62, Max 89.25, Avg 58.01
Geothermal Valley - Min 5.25, Max 82.67, 55.91
Overall - 57.28
Conclusion - Lower highs, but better averages in DX12. Some of the minimums were way higher, and there was a definite feel of smoother motion. I think the extreme low on Geothermal valley is due to one area of texture loading where the screen stuttered bad, nearly frozen, near the end for about a second.

So in my case, with these two games, there is noticeable improvement using DX12, on what would be considered good hardware for circa 2013. I can't afford to upgrade now, though something may be on the "hoRyzen," so I am certainly glad there have been some DX12 developments.

Now you all have given me the bug to try out different things and benchmark my system :p. Next thing you know I'll be downloading 3dMark, and MSI afterburning and tweaking like crazy!
 
if performance is basically the same using DX11 and DX12 why wouldn't you just use and recommend DX12?...
Recommending DX12 would essentially be recommending W10.

If both were the exactly the same, I'd recommend DX11 since it means that DX12 brings nothing new to the table, and thus no advantage of migrating to W10.
 
Recommending DX12 would essentially be recommending W10.

If both were the exactly the same, I'd recommend DX11 since it means that DX12 brings nothing new to the table, and thus no advantage of migrating to W10.

DX12 does bring new things to the table as far as better CPU core utilization and overall CPU improvements...it might not have any flashy new GPU tech like tessellation but it's still an upgrade...that way you can get the improvements in the games like Hitman where DX12 is better but you can also revert to DX11 in other games if you choose...you get more options when using DX12
 
DX12 does bring new things to the table as far as better CPU core utilization and overall CPU improvements...it might not have any flashy new GPU tech like tessellation but it's still an upgrade...that way you can get the improvements in the games like Hitman where DX12 is better but you can also revert to DX11 in other games if you choose...you get more options when using DX12

More options only really matters when it actually translates to more options to the end user, which is not the case with DX11/DX12 games right now, in fact, IIRC, there are less options in DX12 than in DX11 (only catching up now).

It's all well and good saying that DX12 uses more of your computing resources, but if that still do not translate to better experience for the end user, then DX12 has completely failed, in that even while using more to the metal coding, it still can't get more performance out of your system, and DX12 has greater software limitation (IE Win 10, as opposed to DX11) for essentially nothing.

My conclusion is drawn from that, if DX11 and DX12 is indiscernible from each other at the end result, then DX11 is still the better solution due to DX12's limitation, regardless of whether DX12 mode actually uses more of your hardware or not.
 
Recommending DX12 would essentially be recommending W10.

If both were the exactly the same, I'd recommend DX11 since it means that DX12 brings nothing new to the table, and thus no advantage of migrating to W10.

Whatever the debate is over Windows 10, while it's lost a bit of market share in gaming the last two months, it's still the #1 PC gaming OS according to Steam's survey at almost half the market. 10 isn't going away and in gaming particularly Windows 7 and 8.1, while they'll long be compatible with most games for Windows, aren't the future. Vulkan may become popular in PC gaming but DX 12, by virtue of the Xbox, isn't going away.

Bottom line, be it DX 9-12 or Vulkan or whatever else that's coming out in the intermediate future, Windows 10 is going to support it in the PC gaming space, period. People can feel how they feel but that's just the reality of it for now.
 
Last edited:
Back
Top