• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Deus Ex: Mankind Divided

Heh. Check out another prime example of DX12 excellence

 
Ieldra said:
I have no doubt they will work wonders in two weeks ;)

Come on, think about this, if it's in so bad a state they didn't feel like releasing it, what could they possibly do in two weeks to change that ?
Click to expand...
Can't say. They stated some last minute polishing.
 
Ieldra said:
I have no doubt they will work wonders in two weeks ;)

Come on, think about this, if it's in so bad a state they didn't feel like releasing it, what could they possibly do in two weeks to change that ?
Click to expand...
Two weeks later. Just enough time for everyone to run their DX11 benchmarks to show Nvidia in the lead.
 
MavericK said:
You realize that DX:MD is an AMD title, right?
Click to expand...
Yeah. Doesn't mean one IHV or another didn't request an extra week or two to optimize DX12 drivers after release. One other possibility is that SM6.0 is releasing in that timeframe, but I can only speculate on that.
 
Anarchist4000 said:
Yeah. Doesn't mean one IHV or another didn't request an extra week or two to optimize DX12 drivers after release. One other possibility is that SM6.0 is releasing in that timeframe, but I can only speculate on that.
Click to expand...

Something tells me we won't be seeing SM6 exclusive features in Deus Ex.

As for that DX11 nvidia in the lead remark, what makes you think AMD will have the lead in the DX12 version? It ain't set in stone that AMD will outperform NV, just that they usually get a performance bump for DX12. If the performance in DX11 wasn't trash there would be no huge increase
 
Ieldra said:
Something tells me we won't be seeing SM6 exclusive features in Deus Ex.

As for that DX11 nvidia in the lead remark, what makes you think AMD will have the lead in the DX12 version? It ain't set in stone that AMD will outperform NV, just that they usually get a performance bump for DX12. If the performance in DX11 wasn't trash there would be no huge increase
Click to expand...
I love how every NVidia fanboi comments on AMD DX11 drivers being trash. Funny how they were able to compete with NVidia the whole time. DX12/Vulkan is exactly what devs and gamers alike have wanted for a very long time. Yet somehow because it threatens NVidia position in benchmarks it isn't anything special. Most of you condemn AMD for their questionable pre-release benchmarks and the inflating of truth, the same could be attributed to your deflating the positives of DX12. It's early and all that DX12 brings is yet to come.
 
JustReason said:
I love how every NVidia fanboi comments on AMD DX11 drivers being trash. Funny how they were able to compete with NVidia the whole time. DX12/Vulkan is exactly what devs and gamers alike have wanted for a very long time. Yet somehow because it threatens NVidia position in benchmarks it isn't anything special. Most of you condemn AMD for their questionable pre-release benchmarks and the inflating of truth, the same could be attributed to your deflating the positives of DX12. It's early and all that DX12 brings is yet to come.
Click to expand...

What ? Precisely because it's early you shouldn't expect DX12 to be a magic bullet.

I didn't comment about DX11 drivers being trash, I commented about DX11 performance being trash specifically in the games where AMD gets huge gains from an API transition.

AotS DX11 is hilarious for GCN. Huge gain going to DX12, still matches NV DX11 performance.

Hitman DX11 runs very well for GCN, small gain going to DX12, still matches NV DX11 performance with no difference in IQ.

I've always maintained that DX12 is going to lift bottlenecks, you've always contended that DX12 is the best thing since sliced bread and you don't have the evidence to support your contention.

Until a DX12 title has DX12 exclusive effects, it doesn't matter which API you're running, what matters is the performance.

If Deus Ex DX12 brings a 300% performance boost for GCN, yet still ends up being behind NV's DX11 performance for competing products, who cares if it's DX12, it runs worse. You can't accept this fact, neither can you accept the fact that DX12 entails IHV-specific optimization, and you always bitch and whine about any IHV-specific work being for NV hardware, but when DOOM Vulkan ships with shader intrinsics being used exclusively for GCN you have no trouble presenting it as evidence of AMD's mythical "low level API superiority".

DX12, Vulkan games just like DX11 and OGL will not inherently run better on AMD/NV, it depends on the game, as always. The main difference is CPU and command submission bottlenecks should not longer hold GCN back.
 
Ieldra said:
What ? Precisely because it's early you shouldn't expect DX12 to be a magic bullet.

I didn't comment about DX11 drivers being trash, I commented about DX11 performance being trash specifically in the games where AMD gets huge gains from an API transition.

AotS DX11 is hilarious for GCN. Huge gain going to DX12, still matches NV DX11 performance.

Hitman DX11 runs very well for GCN, small gain going to DX12, still matches NV DX11 performance with no difference in IQ.

I've always maintained that DX12 is going to lift bottlenecks, you've always contended that DX12 is the best thing since sliced bread and you don't have the evidence to support your contention.
Click to expand...
Evidence is intelligence and the ability to see beyond the little box most of you naysayers are stuck in. The biggest DX12 positive is AI and the shear number you can have. I love Skyrim and can easily see what DX12 can bring to it. Sad you can't or won't.
 
JustReason said:
Evidence is intelligence and the ability to see beyond the little box most of you naysayers are stuck in. The biggest DX12 positive is AI and the shear number you can have. I love Skyrim and can easily see what DX12 can bring to it. Sad you can't or won't.
Click to expand...

You're talking about drawcall submission right now, MMOs and Open World RPGs should benefit a lot from DX12, but what does that have to do with anything ?

Tomb Raider reaps these benefits in DX12. So does AotS. So what ? Do you have a point mate ?

Also, this is shearing

2000px-Shear_scherung.svg.png

I don't know how you shear numbers, but it's sheer numbers. Sheer numbers of what ? And how does DX12 affect AI lol ? Just because you have the compute queue exposed ? This is silly

"Evidence is intelligence"

No it fuckin ain't, evidence is evidence, and you have none.
 
JustReason said:
Evidence is intelligence and the ability to see beyond the little box most of you naysayers are stuck in. The biggest DX12 positive is AI and the shear number you can have. I love Skyrim and can easily see what DX12 can bring to it. Sad you can't or won't.
Click to expand...

How is that evidence of anything but your dubious taste in RPG's?
 
stashix said:
How is that evidence of anything but your dubious taste in RPG's?
Click to expand...
Do you have anything to add that is intelligent or do you lack anything tangible so you resort to infantile comments. By the by, Skyrim is still a top 10 game on steam most days. So seems the majority of the world doesn't share your asinine comment.
 
JustReason said:
Do you have anything to add that is intelligent or do you lack anything tangible so you resort to infantile comments. By the by, Skyrim is still a top 10 game on steam most days. So seems the majority of the world doesn't share your asinine comment.
Click to expand...

Take it easy, you'll pop a vein and will be left dragging a foot.
 
TaintedSquirrel said:
I'd also rather get 60fps in DX11 than 50fps (or whatever) in DX12.
At least with the delay, they're forcing everyone to play the superior DX11 version for a few weeks rather than people who don't know better getting sub-par performance in DX12.
Click to expand...
Are you saying Nvidia just sucks cock at playing dx12 in general or something else? funny thing is it actually improves frustrates on AMD cards lol! I could swear the new 10 series by nvidia improves in dx12 games right? Heck i thought it did.
 
Ieldra said:
Something tells me we won't be seeing SM6 exclusive features in Deus Ex.
Click to expand...
XB1 is the testbed for SM6, so the game likely uses them. Just not on PC. It would seem to be possible if they knew the timetable. It should benefit both IHVs, although AMD might get a bit more benefit from it.

Ieldra said:
As for that DX11 nvidia in the lead remark, what makes you think AMD will have the lead in the DX12 version?
Click to expand...
Never said they would be in the lead, but odds are the gap would be significantly narrower if all the reviewers used DX12. IF DX12 ended up being faster for AMD there would be reason to avoid reviews saying use DX11 for Nvidia and DX12 for AMD.
 
Skyrim is top seller there it is a great game. HAHAHA

Oh man.

dX4W9Mb.png

5WQy9nx.png

OX0m8Qn.png

Tell me all about AMD's inherent LL-API advantage and explain why it dissipates when you overclock the cards ? let me guess, overclocking makes up for the "missing hardware features" ?

Anarchist4000 said:
XB1 is the testbed for SM6, so the game likely uses them. Just not on PC. It would seem to be possible if they knew the timetable. It should benefit both IHVs, although AMD might get a bit more benefit from it.


Never said they would be in the lead, but odds are the gap would be significantly narrower if all the reviewers used DX12. IF DX12 ended up being faster for AMD there would be reason to avoid reviews saying use DX11 for Nvidia and DX12 for AMD.
Click to expand...


Oh I have no doubt it will narrow performance deltas if AMD was behind in DX11, I hate having these discussions because I always end up seeming like I'm downplaying the benefit of DX12 for AMD. I'm not, I recognize it brings huge gains, as did VK in DOOM, but these are always relative to lackluster performance using another API.

There's nothing inherent in DX12 that makes it better for either IHV, DX12 games that are using optimized codepaths for one IHV and not the other will naturally favor one IHV and not the other, just like any API. It's not a magic bullet as it is often portrayed, and the absence of performance improvements for NV cards means nothing except that the DX11 path was running just fine...
 
primetime said:
Are you saying Nvidia just sucks cock at playing dx12 in general or something else? funny thing is it actually improves frustrates on AMD cards lol! I could swear the new 10 series by nvidia improves in dx12 games right? Heck i thought it did.
Click to expand...

The only game with big performance recession like that is the Thailand episode of Hitman, if ass you say NV sucks with the DX12 versions, yet their DX11 performance using the same IQ is higher even than AMD's in DX12, what does that say to you ? I'll tell you what it says to me, it says that the DX12 path is far less efficient than the DX11 one, and that's 100% on the developer, not NV.
 
Ieldra said:
There's nothing inherent in DX12 that makes it better for either IHV, DX12 games that are using optimized codepaths for one IHV and not the other will naturally favor one IHV and not the other, just like any API. It's not a magic bullet as it is often portrayed, and the absence of performance improvements for NV cards means nothing except that the DX11 path was running just fine...
Click to expand...
Async shading would make a difference IMHO. While not DX12, I think Doom and Vulkan would be a prime example. ID's TSSAA looks like an asynchronous timewarp. Allowing them to reduce AA and postprocessing costs as they only occur when a vblank is encountered. That should benefit both IHVs, although obviously Nvidia has some timing issues with the async shading part. While it might not show in raw FPS, I have a feeling that's the cause of the stuttering and input lag they're encountering.
 
Anarchist4000 said:
Async shading would make a difference IMHO. While not DX12, I think Doom and Vulkan would be a prime example. ID's TSSAA looks like an asynchronous timewarp. Allowing them to reduce AA and postprocessing costs as they only occur when a vblank is encountered. That should benefit both IHVs, although obviously Nvidia has some timing issues with the async shading part. While it might not show in raw FPS, I have a feeling that's the cause of the stuttering and input lag they're encountering.
Click to expand...
I've had no stuttering at all in vulkan doom, and pascal has no problems with async, on maxwell it's disabled entirely so... Doubt it's the cause of any stuttering you've encountered

Async rendering tasks still benefit Pascal less than Fiji for example, even in timespy, I don't think it will have a major effect on the performance landscape
 
Ieldra said:
I've had no stuttering at all in vulkan doom, and pascal has no problems with async, on maxwell it's disabled entirely so... Doubt it's the cause of any stuttering you've encountered

Async rendering tasks still benefit Pascal less than Fiji for example, even in timespy, I don't think it will have a major effect on the performance landscape
Click to expand...
Pascal can likely use preemption to get around the issue. Once every other frame or so depending on vblanks would be a huge issue. Point being it's an advantage that DX12 would have over DX11 beyond the usual benefits expected of async. Reducing the actual workload as opposed to increasing performance by filling gaps.
 
Anarchist4000 said:
Pascal can likely use preemption to get around the issue. Once every other frame or so depending on vblanks would be a huge issue. Point being it's an advantage that DX12 would have over DX11 beyond the usual benefits expected of async. Reducing the actual workload as opposed to increasing performance by filling gaps.
Click to expand...

Sorry, I don't follow, you'll have to explain what Id is doing to reduce workload as opposed to filling blanks.

Well Pascal preemption is more granular than GCN's so if anything it can interrupt a running shader faster than GCN can but, as usual, the question is should it be interrupting a shader to run another task?
 
Ieldra said:
Sorry, I don't follow, you'll have to explain what Id is doing to reduce workload as opposed to filling blanks.

Well Pascal preemption is more granular than GCN's so if anything it can interrupt a running shader faster than GCN can but, as usual, the question is should it be interrupting a shader to run another task?
Click to expand...
In one of the interviews with the devs he mentioned how he liked to decouple rendering to improve performance. My understanding is that under TSSAA each frame is rendered with no AA or postprocessing. On consoles even the resolution of each frame may change. Then during a vblank a compute shader runs that intreprets all the rendered frames. Upscaling, AA, postprocess etc. So the game could be rendering 200 fps and only running the compute shader that does the SSAA and postprocessing 60 times. The compute shader would be fixed to your refresh rate while the game was likely rendering much faster.
 
So the resolution of every frame was changed dynamically to meet a time constraint ? Vblank is just the time between the end of reading the last line of a frame from FB and the beginning of reading the first line of the following frame right ? Essentially they're applying postprocessing only on the frames that are displayed ?
 
Ieldra said:
So the resolution of every frame was changed dynamically to meet a time constraint ? Vblank is just the time between the end of reading the last line of a frame from FB and the beginning of reading the first line of the following frame right ? Essentially they're applying postprocessing only on the frames that are displayed ?
Click to expand...
Not necessarily every frame, but on consoles they could be rendering 900P just to boost performance a bit. Yes they could adjust resolution on the fly to maintain a stable framerate though. On PC the resolution was fixed according to the devs. Correct on the postprocessing/AA only on displayed frames. It's a bit different than that with the temporal AA part. It was some sort of bicubic interpolation of the last few frames on top of reprojection to arrive at the presented frame. So the rendered frames are never actually presented, just the output of a compute shader mashing them together.
 
Anarchist4000 said:
Not necessarily every frame, but on consoles they could be rendering 900P just to boost performance a bit. Yes they could adjust resolution on the fly to maintain a stable framerate though. On PC the resolution was fixed according to the devs. Correct on the postprocessing/AA only on displayed frames. It's a bit different than that with the temporal AA part. It was some sort of bicubic interpolation of the last few frames on top of reprojection to arrive at the presented frame. So the rendered frames are never actually presented, just the output of a compute shader mashing them together.
Click to expand...

I see... but vblank will be inversely proportional to the framerate so as that goes up you will have less time to perform the compute shader unless i'm not understanding something fundamental here, if you're playing with unlocked fps then it will be all over the place
 
Intel beating the pants off Nvidia with their new server technology, and now another "DX12 is the devil" invasion by the Nvidia defense force.
 
Anarchist4000 said:
Pascal can likely use preemption to get around the issue. Once every other frame or so depending on vblanks would be a huge issue. Point being it's an advantage that DX12 would have over DX11 beyond the usual benefits expected of async. Reducing the actual workload as opposed to increasing performance by filling gaps.
Click to expand...


No preemption will not get around that issue, by using preemption you will be loosing milliseconds of time, not even microseconds.

Get the whole preemption thing out of your head, to use preemption, you have to turn off a kernel, then turn another kernel on, then turn off that kernel, and then turn the other kernel back on again, That is 4 steps, this is why it can't be used, it can be used for time sensitive tasks and prioritize tasks, because you are planning for certain things to happen. So in that delayed time you can do other things as well.

Even with fine grained preemption with AMD cards if that was used it will still have noticeable negative impact on latency,
 
razor1 said:
No preemption will not get around that issue, by using preemption you will be loosing milliseconds of time, not even microseconds.

Get the whole preemption thing out of your head, to use preemption, you have to turn off a kernel, then turn another kernel on, then turn off that kernel, and then turn the other kernel back on again, That is 4 steps, this is why it can't be used, it can be used for time sensitive tasks and prioritize tasks, because you are planning for certain things to happen. So in that delayed time you can do other things as well.

Even with fine grained preemption with AMD cards if that was used it will still have noticeable negative impact on latency,
Click to expand...

Well the preemption latency is basically the time between giving the command for something to halt and it effectively halting and making space for (potentially) a new kernel, you're describing a context switch. I think what Anarchist was saying in the context of this DOOM discussion is that the preemption should help as you want to interrupt running shaders at vblank and have them work on the frame about to be presented, a little bit like ATW.

Preemption ensures your time critical kernel will be done in time, but it's effects on runtime of other kernels is unpredictable really (without looking at specific case on specific HW)
 
yes its extremely unpredictable in that executing it without understanding the needs of all the queues and what ever else they need you can't effectively predict what you might miss, Doom I don't think uses this at all because they were fairly specific on what they use async for, and they never talked about preemption. Even oxide stated that haven't used preemption for anything as of yet but are "looking into it".

To create that kind of code, to have that much control over the graphics pipeline, you will end up loosing performance as a result.
 
razor1 said:
yes its extremely unpredictable in that executing it without understanding the needs of all the queues and what ever else they need you can't effectively predict what you might miss, Doom I don't think uses this at all because they were fairly specific on what they use async for, and they never talked about preemption. Even oxide stated that haven't used preemption for anything as of yet but are "looking into it".
Click to expand...

Would they "use it" though ? I imagine driver is in charge, preempting only when some kind of priority flag is present on a queued grid
 
I would not leave something like that up to the driver. Doesn't sound feasible in a runtime environment.
 
razor1 said:
I would not leave something like that up to the driver. Doesn't sound feasible in a runtime environment.
Click to expand...

If I'm not mistaken AMD announced something somewhat related it was called priority queue or something, related to programmable ACEs and tweaking behavior for time critical things like ATW, you could program it explicitly (though with some quirks) or it would use some kind of default behavior, but I digress...
 
You must log in or register to reply here.
Back
Top