Test Shows Over 100fps Difference Between DX12 and DX11

[ComradeHX]

No one saying anything about how low 13fps is?

He had AMD 8 core cpu; it is to be expected.

 

---

 

I bet he took minimum fps of DX11 result and subtracted it from average(or even max) fps of DX12 result.

[Faceman]

There is too much that is sketchy about this.  If it is true, then this is awesome, but I don't think nearly enough information has been provided to take this to court.

[Xorbot]

I am not that hopeful that the stacking VRAM thing is true.  However, if it is true, then my GTX 680 SLI just got a breath of life.  Note, GTX 680 SLI still performs very well and has a lot of processing power, but each cards' 2 GB of VRAM is a limiting factor for today's games on Ultra settings.

[silberdrachi]

Id like to see what the testing methodology was.

 

There are a lot of huge issues in DX11 that DX12 is supposed to fix, by hammering away at one of those issues to drop the FPS to stupid levels just to see it jump up in DX12 is dumb. Id like to know if this was actually something in one of their engines that sucked so much that it only got 13fps or if they created some sort of canned benchmark to test the performance of a single function that saw improvement. I really hope its not the first, because that would just make them bad programmers (no one is going to release a  DX12 only game for a long time) and the second makes them just stupid for thinking thats any indication of average performance.

 

Considering they are literally right down the street from me, i know a few people working there now, going to have to see if I could find anything else about it. Sometimes i wonder if i should've taken a job there, but i just never want to work in gaming.

[Jerakl]

Sounds extremely fishy.

Weird, never seen you around before.

 

And you almost have 4K posts...

[Albatross]

VRAM stacking?

[sgloux3470]

I mean, I know that there was some increase in performance from DX11 to 12 and that multithreaded CPUs gain some performance, but I refuse to believe DX12 makes that much of a difference.

 

Unreleased GPU? MAJOR grain of salt for now.

 

DX11 had some serious bottlenecks that developers had to work around.  You could approximate certain effects at good framerates in DX11, but I think in cases where you wanted to more accurately simulate them it would cause huge performance hang ups.  That's likely what we're seeing here is the latter situation the bottleneck is cleared up.

 

It's not like they're unlocking super duper performance.  They're just clearing up bottlenecks that were bogging down the system.  (i.e Drawcalls is a big one.)

[EpicAdom]

Although this is good, I'm betting this is absolute best case scenario.

This^

[MeDownYou]

Why is this a surprise? when people said this about mantle (which is still slightly better than dx12 btw) everyone was like hey...cool but unless its optimized for 8 cores then the real world difference is wayyyy less. So, why don't we look at this with a new perspective eh?

[Opcode]

I highly doubt these results...

What's there to doubt about them? Look at Star Swam tests in comparison.

 

 

Brad Wardell already said that they are working on a new game (an actual game) that makes Star Swam look "primitive".

[Humbug]

They promise this now, but in actual games I fell like it'll be a much smaller jump. And certainly not a uniform jump across titles.

he did not promise anything.

Everybody knows(or should know by now) that starswarm is designed to push massive amounts of draw calls which cripples directx11 drivers with CPU bottlenecks. It's a scenario designed to isolate one aspect of GPU performance.

It's a good demo, people shouldn't be silly and expect similar gains in games. Also half the people on this forums seem to want to judge graphics APIs simply by how many more frames per second they can get in X game, which is a very narrow view.

[Keudn]

Ill bet $100 that the "test" he did was specifically designed for DX12 and when he ran it on DX11 it ran like shit so he was like "oh hey look how much performance gain you get!"

[GoldSrc]

Over 100FPS difference? yeah, sure...

 

 

[Opcode]

Ill bet $100 that the "test" he did was specifically designed for DX12 and when he ran it on DX11 it ran like shit so he was like "oh hey look how much performance gain you get!"

Believe it or not Star Swarm was optimized for DirectX 11 and look where it stands with Mantle.

[Misanthrope]

This is a tech demo, and a pretty bad one at that: Nobody is gonna be able to properly code for such heavily CPU bound multi threaded games for a good while. There's a reason why we don't have tons of copies of high end software that is multithreaded like Adobe tools and such being created left and right.

[Dezz]

 

Of course, Wardell’s followers had questions aplenty regarding his DX12 test, which he duly answered. He told one follower that he ran the test on a Crossfire system with an AMD processor, while he said to another, when quizzed on whether the system specifications compared to the Xbox One, Wardell replied that it was, “way beyond console stuff”.

 

 

Without the comment of him saying AMD based CPU it could of very well been a 5960X and GM200. Tho for the simple fact that we know he's working in AMD's booth at GDC this year to showcase a few of his games (Star Control I imagine being the Star Swarm replacement) that pretty much confirms that he's running all AMD hardware.

 

 

I think it was already answered.

[Opcode]

I think it was already answered.

Wasn't a question I'm stating even if he didn't disclose what he was running I would still know exactly what he's running.

[FakezZ]

There are 2 cases:

1) DX11 was good and DX12 is just gonna improve performance decently (10-20% with normal use)

2)DX11 was such a piece of shit and they fixed DX12 making it decent and are now seeing 1000% better performance.

I think we all know the chances the 2nd case is true are incredibly slim. But still... you never know.

[Opcode]

There are 2 cases:

1) DX11 was good and DX12 is just gonna improve performance decently (10-20% with normal use)

2)DX11 was such a piece of shit and they fixed DX12 making it decent and are now seeing 1000% better performance.

I think we all know the chances the 2nd case is true are incredibly slim. But still... you never know.

DirectX 10-11 were not optimized API's for the eras that they took place in. Number one on your list is completely out of the question. DirectX 12 stays up there with Mantle which brings over a 400% performance improvement in comparison to DirectX 11.

[FakezZ]

DirectX 10-11 were not optimized API's for the eras that they took place in. Number one on your list is completely out of the question. DirectX 12 stays up there with Mantle which brings over a 400% performance improvement in comparison to DirectX 11.

200% performance increase on what? Best case scenario with a crappy cpu or a normal use-case?

[Mooshi]

While I respect Brad, def need to try DX12 on a new card in Win10 to get a hands on feel!

[Mujteba]

Was he measuring that initial framerate spike when you just open a game I wonder?

[Ethnod]

Sounds amazing but... Microsoft, I'm keeping my sceptical hat on mofos, until there is proof I'm not downloading another untested and unproven OS from your skanky ass until at least SP1... probably SP2. Fell for that shit with WinME, learn that lesson 

 

[jecho]

If properly added to the Total War engine this could be HUGE performance gains. I would say it could mean at least 20% performance increases in battles. Maybe even proper cpu core scaling, which would be funny if a 10 core Xeon outperformed the i series chips. Otherwise DX 12 should only bring modest gains in performance. 

[Thony]

Probably a game that had terrible DX11 Crossfire support.

But if this means that DX12 can finally optimize crossfire/sli then im extremely happy! Might even pick up another 760 or see if 960 is compatible for sli.

Over 8 months to go and HYPE is getting way too strong already.