Preemption; and a lot of removed constrains, so your code gets less (less checks, less management, etc.), the driver handles these cases then as it sees fit, probably in a much better way than a 3rd party from outside the black-box.
AMD: Sea Islands R1100 (8*** series) Speculation/ Rumour Thread
Preemption; and a lot of removed constrains, so your code gets less (less checks, less management, etc.), the driver handles these cases then as it sees fit, probably in a much better way than a 3rd party from outside the black-box.
This attitude is very very strange. Doesn it mean that all DX11 videocards will support DX11.1 by the means of a simple Windows Update.
No it doesn't, HD5/6 and GTX4/5/Fermi-6 won't support it, it requires features from the hardware side that just isn't there. Some of the new features could be supported I believe, but not all, and since DX looks for featureset you support, you can't really use those you could support, either.
The question is if DX11 hardware will be able to run DX11.1 applications (like DX10 hardware could run DX10.1 applications). If so, wouldn't it be possible to write a driver, which would report full DX11.1 compatibility, but perform the DX11.1 specific code on DX11 hardware with some additional performance cost? E.g. old Radeon 7000 with late drivers reported to be fully DX7 compatible (including HW TnL), but the driver actually emulated TnL on CPU.
DX10 hardware can't run DX10.1 applications, if there's no fallback path to just DX10, it won't run on DX10.
Same goes for DX11.1 for sure, if some dev decides to require DX11.1 support, then that's what you need to have. Any sensible dev, ofc, won't do that, though.
(And remember to keep API and Feature Set separate, all cards with modern drivers support DX11 or 11.1, but Feature Set can be anything from DX9 up
Officially we have been told that Kepler supports D3D 11.1. Whether that works in any current driver build I'm not sure.
As for why NVIDIA is being so cagey about it, I suspect NVIDIA wants to downplay it since only the Kepler members of the 600 series would support it.
DX 11.1 has one killer feature, at least for the developer - proper shader debugging support, you know, with breakpoints and all. That's what the new tracing API is all about. The best thing about this is that people with old GPUs get the benefits too - the developer is able to code better shaders in less time with the debugger, and the shaders then work on DX 11 hardware too.
For conventional shaders, a debugger isn't that big a deal, but when you start using compute for anything - AI, physics, texture compression, etc., the shader complexity goes way up, and a debugger is critical to getting the thing to work at all.
Bouncing Zabaglione Bros.
Legend
So isn't it about the time of year when we should start getting an inkling of AMDs next-gen cards coming in Autumn/Winter time? Is there anything up, or are they really going to coast through the back to school season on the GHZ editions alone?
Coming from a 6950 with unlocked quads, there's really nothing compelling about the 7xxx range, especially given it's price. Heck, I'm likely to build a new PC before the end of the year, and I'm intending to keep the old graphics card, rather than drop a lot of cash on an incremental upgrade.
Come on AMD, try and tempt me with Sea Islands cards! Don't you want my money?
Coming from a 6950 with unlocked quads, there's really nothing compelling about the 7xxx range, especially given it's price. Heck, I'm likely to build a new PC before the end of the year, and I'm intending to keep the old graphics card, rather than drop a lot of cash on an incremental upgrade.
Come on AMD, try and tempt me with Sea Islands cards! Don't you want my money?
Given AMD's usual schedule these days, I wouldn't expect anything before December, November at the earliest. But there could be enticing price drops before that. I would argue that the 660 Ti alone warrants some.
Right, but the hardware feature that enables this is being able to R/W UAVs from every shader stage (not just pixel/compute). I imagine most unified shader hardware does already support this though.
There are a few other minor features like being able to set the rasterization sampling rate separately from the currently bound render targets, etc. but it's pretty specific stuff. It's totally possible that those features might not be possible to support on some feature level 11 hardware, but it's likely that most hardware is fine.
Dunno why it would, unless the 7950 pricing here is completely scewed, as currently 7950's can be had for about the same price as GTX660 Ti's supposed MSRP+VAT (in reality GTX660Ti is ~50€ more expensive than what it should here), and the performance is similar or tad better on 7950
Depend what MSRP we are talking about, i have yet not see any review of the 660TI reference .. and so most are 30-40$+ more in reality.
(The only one is PCper who have underclock an MSI OC card ) ...
Supposed MSRP for Europe is 259€, without VAT, which puts the price with VAT somewhere around where 7950's can be had at the moment with VAT
Most of the reviews, even if they're from "factory OC'd cards", include reference version scores too
Kyyla
Veteran
They wont entice me! I'm holding out with my trusty 6950 until the 2nd gen of cards for this node comes out.
The 6950 looks to be able to play all of the 2012 releases just fine. I may not even buy a sea island card at this point unless a big pc game comes out that i can't play
Try upgrading to an Eyefinity setup, you'll see the need for a new graphics card appear much sooner.
eyefinity is not for everyone, all i want is 1600p at 30".
What can SI bring to upgrade on 7970 without bloating the perf/mm/watt? I am resisting the 7970 even though it is about 40-60% faster than my 6970. Will 8970 offer the same 40% increase over 7970?
I think one area Nvidia lead in benchmarks is tessellation. Yes 7970 caught up in 'standard' DX11 tessellation and unigine runs, but i feel some developers went all out on the extra polygon units offered by nvidia gpu.
What can SI bring to upgrade on 7970 without bloating the perf/mm/watt? I am resisting the 7970 even though it is about 40-60% faster than my 6970. Will 8970 offer the same 40% increase over 7970?
I think one area Nvidia lead in benchmarks is tessellation. Yes 7970 caught up in 'standard' DX11 tessellation and unigine runs, but i feel some developers went all out on the extra polygon units offered by nvidia gpu.
Try upgrading to an Eyefinity setup, you'll see the need for a new graphics card appear much sooner.
3 24 inch 1080p monitors for eyefinity.
The thing is , the games all run fine without AA and the next release i'm interested in is x-com followed by bioshock. I don't think either of them will give me problems.What else is there ? assasins creed ? The others worked so that should work.
I guess if you can stand gaming without AA, yes, it's fine. I personally cannot, so I'll be upgrading sometime in the nearish future.
Similar threads
