DX11 to show up soon?

http://twitter.com/catalystmaker

OK, ATI is supposed to show some DX11 working cards tomorrow

Agreed, I really can't take twatters seriously.

this stinks I was just about to buy a Dx10 card.

First, I believe DX11 is supposed to be here soon. (is this supposed to work on Vista as well, or is this Win7 only?)

I'm not surprised AMD has working DX11 silicon already. DX11 isn't much more then DX10.1 with a few more things. AMD has had DX10.1 cards for quite awhile, so tacking on the DX11 parts probably aren't that hard. They have also already made GDDR5 and 40nm cards as well, so again this isn't all that new for them. I wonder what Nvidia has, and did they really find a way to more then double the shader count?

Rumors are running that ATI may come out with a full cross section, from top to bottom for DX11.
Rumors on nVidia side are somewhat more obscure, tho, theres reportedly several DX11 cards in the pipeline, and its said theyre several to 3-4 months behind ATI.
As to any specs on any cards, its all speculation, nothings really confirmed yet, or close. It may have alot to do with ATIs surprise with their R700 family release, and the 480/800 shader thing, besides them just normally holding their cards tight to the vest. Also, LRB will be here eventually, and thats just another competitor to keep things from

Link? After all, DX11 is a new API in Win7. We never saw DX10 in XP, so why would MS make such a massive SP for Vista?

Unlike DX11 isnt supposed to be as much of a paradigm shift as DX10 was from DX9 so it likely wont be a massive SP.

According to this
http://blogs.msdn.com/ptaylor/arch [...] ement.aspx
most of DX11 will run on DX10 hardware fine except for a few parts of it.

The hardware tesselation and SM5 are probably two of the most important parts however, and those will need a dx11 card.

Is this info from Micro$oft good enough?

http://www.xnagamefest.com/presentations.htm
http://www.microsoft.com/downloads [...] laylang=en



It's a different launch from DX10

Very different... Hard to believe as well, but now I have several different links saying the same thing, I guess I'm going to have to believe...

OT, I thought you were headed down to Kiwiland?

Hmm let's see. 959m transitors on the 4870 compared to possibly 1.2bn on this. The 3870 had 666m so yep that wafer is probably the top end 58xx gpu's assuming the writers math isn't way off.

http://anandtech.com/video/showdoc.aspx?i=3573
heheh

Yes I did know that, but i felt it was fair to pretend it didn't exist seeing as Nvidia doesn't have it and it never got used because of that anyway.

Well, nVidia hasnt been aiming in that direction, as its gpu related, and not gpgpu related heheh

Jennyh, remember we thought the same thing about the G80 vs R600, so while it looks daunting you never know until the final product ships.

Well nvidia is a big corp which should have a good r&d department I have good faith they will deliver on their refinement release of DX11 cards. Nvidia didn't skip dx10.1 but didn't want to change their processes for something that they know wont last that long always.

ATI always promotes them self as the best though innovation while nvidia takes the approach as best though refinement and making what works better.

ATI can take that approach due to the fact they aren't leading in nvidia vs ati and they have a smaller processing facility which cost less to change their practices from one card to the next.

edit:G80 vs R600 was rape cake T_T i hope it doesn't turn out like that again ATI did horrible vs nvidia making nvidia charge *** loads for their premium produce because ati couldn't deliver and made it worse with the G92 release.

Also rape does not trigger the censer lol

They were a big company 'with a good R&D dept' when they made the FX series too. Doesn't mean much.

Yes they did, and went so far as to downplay it and say their reasoning was transistor budget.

They each do that when it fits their own current situation, that wasn't the party line during the GF6800/X800 era , and nV didn't expect that during the R9700/FX era before it either.

ATi and nV use the same facility, TSMC.

As for who's leading, nV loses money making and selling their GPUs while ATi is AMD's main profit source to stem the flow of red from the CPU division, so really who leads whom right now?

It's also unlikely that the RV8xx series leaves itself open for a repeat of the G80 vs R600 situation as it doesn't appear to have the second launching part, and also doesn't have to worry about a different node than the competition, both will be using 40nm and should experience the same issues if there's a repeat of the problems that were experienced with 80nm HS.

Except for the 6xxx series against the x8xx series. The 6xxx supported SM3, while the x8xx only supported SM2. At least AMD was big enough to admit they didn't support it this round, and they would the next. How many rounds has Nvidia skipped DX10.1? (I guess two as long as you don't count the 9xxx chips as new.)

Companies always put the best spin on things. If it suits the marketing dept to claim SM3 support before their competitors, then they'll do it. But if they fire back with something else, well then they will claim XXX excuse. What matters to me is AMD said next round and did it. Nvidia only seems to rename chip, and put spin on things.

Edit: fixed quote tag.

A 40nm g200 might outperform a 40nm r700, but will it outperform the bottom end r800?

ATI dont just hold all the aces, they hold all the cards (no pun intended), they are in complete control of the market except for the very, very top. They will get that back soon of course with the 58xx and the 58xx X2. I will be disappointed if the single gpu 58xx isnt at least as fast as a gtx295 however.

If we fast forward 3 months, the gpu market could easily look like this :-

5870x2 an absolute mile ahead of anything else (possibly 2x faster than a gtx295).
5870 in second place tied with gtx295

Now, assuming Nvidia shrinks the g200 and cuts prices as much as they can (dont forget the 448 and 512 bit buses), those 285gtx, 275gtx and 260gtx still have to be faster than the equivalent 5850 and 5830's. The difference again is, ATI will make money on these small chips while Nvidia keep losing it.

The good old g90's will be crushed by 40nm 4870 and 4850's.

Nvidia need a miracle to just stay in the game next year. They are 3-6 months behind and although I expect the g300 to be a real powerhouse, it could already be all over for them by the time it's released.

HardOCP has some DX11 screenies up. As expected, nothing there wows me, and I wouldn't have known it was DX11 if the pics weren't labeled.

But that's the same about DX8.1 vs DX9, DX9 vs DX10, etc.

A still screenshot does little to expose any differences, especially in things like efficiency, which is the name of the game with DX10.1 & DX11, not dramatic visual differences like DX7/8 to DX8.1m where suddenly you had shiny water/surfaces.

How are you going to appreciate the benefits of tessellation, memory optimizations or better multi-core support in a 2D image?

You could emulate almost any static visual in DX10 using DX8.1, so screenies of a sandbox isn't really going to do much. Even something like HDR, specular lighting wouldn't be truly appreciable until you had video, and a fleshed out scene with backgrounds and other objects to interact with.

I'm going to reserve judgment until there's motion video of DX10/10.1 alongside DX11 trying to achieve the same final output.

Also remember this is early hardware with next to no driver development, so I'm not expecting much for a while other than developers doing alot of crass demo stuff at first.

No not really. Think of Tesselation taking a standard 70,000 plygon model running at 2-3 FPS, and instead turning that into a 5,000 plygon tesselated model that looks the same which runs @ 14fps, it's a massive improvement, but you wouldn't SEE a difference in a screenshot, and wouldn't know 14fps is 'good'.

If tesselation works properly you shouldn't be able to tell the difference between a hi-poly model and a low poly tesselated model, so a still wouldn't tell you much, and fps alone wouldn't either, unless, like I said you had another card running DX9/10/10.1 beside it to compare.

One thing is for sure, it's far from an incremental update from DX9, it's long since been a complete split as to how things are done, both from a threading, memory and feature perspective.

But also like I said, you can make DX8.1 look just like something rendered in DX11 in a still shot (same with software/CPU render can look like a DX11 output), the tough thing is to make them similar in full motion video.

I would dispute that A.I comment, where is he getting his info that that cannot be done with only a cpu.

Not sure how its broken down, but maybe using the compute shaders vs cpu, its requiring only a single pass, and thus the latency is slashed to the point where the strain on either or both cpu and gpu would be able to make it work vs reducing the amount, or severe fps lags