Slashdot Mirror
Intel To Integrate DirectX 11 In Ivy Bridge Chips
angry tapir writes "Intel will integrate DirectX 11 graphics technology in its next generation of laptop and desktop chips based on the Ivy Bridge architecture, a company executive revealed at CES. AMD has already implemented DirectX 11 in its Fusion low-power chips. Intel expects to start shipping Ivy Bridge chips with DirectX 11 support to PC makers late this year. Ivy Bridge will succeed the recently announced Core i3, i5, and i7 chips, which are based on Intel's Sandy Bridge microarchitecture."
does it still contain the DRM restrictions capability ?,
because Intel can forget all about CPU sales from us and from any of our customers until its removed
i dont care if it promises a free pony
contains DRM==No sale
period
Will Intel provide documentation so that other OSes will be able to make use of this feature ?
UPS Sucks
And what use is this to those that do not use Microsoft Windows? And what use is it when a bug is found in DirectX, you can change software, but hardware?
Take Nobody's Word For It.
Goes to 11!
(I'm sorry)
I'd rather they made their integrated graphics fast than simply support new DirectX capabilities. I don't really see the point of supporting certain features if the whole thing is going to be slow. I suppose it's easier to implement something than it is to implement it well.
They have always waited for other graphics companies to lead this charge. It's a huge effort on the hardware side to be Microsoft's partner on this. The benefit is that you get out the door first, but Intel has never pushed for that leadership position in graphics (so long as you don't count volume).
For every benefit you receive a tax is levied. - Ralph Waldo Emerson
1. Will this in any way benefit OpenGL?
2. Will this hinder future versions of DirectX or are they backwards compatible in a way that there would be large chunks in hardware and new changes made as firmware revisions or software implementations?
Learning HOW to think is more important than learning WHAT to think.
The Sandy Bridge chips are the first in which Intel has combined a graphics processor and CPU on a single piece of silicon.
I thought Intel already did this a while ago with the newer Atom chips:
http://en.wikipedia.org/wiki/Intel_atom#Second_generation_cores
Escher was the first MC and Giger invented the HR department.
It will still be too slow to use it for anything that is DirectX 11. Why do they even bother?
Those new texture mapping algorithms will really make outlook load fast.
Evil people are out to get you.
That's what I am worried about, I want my Minecraft landscapes to be rendered better.
He who knows best knows how little he knows. - Thomas Jefferson
For the foreseeable future you can have your pick of ARM and x86.
On the plus side, x86 has been pretty much RISC internally for a long time now. And a lot of the ISA has been changed over too. Once they tack on one or two more ISA extension you'll be able to have 100% of your code avoid the x86 path.
"Who is the Journal of Quantum Physics going to believe?" --Stephen Hawking
AMD has already implemented DirectX 11 in its Fusion low-power chips.
As has nVidia in GTX 400.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
You can find Sandy Bridge GPU benchmarks at http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/11
"Intel's HD Graphics 3000 makes today's $40-$50 discrete GPUs redundant. The problem there is we've never been happy with $40-$50 discrete GPUs for anything but HTPC use. What I really want to see from Ivy Bridge and beyond is the ability to compete with $70 GPUs. Give us that level of performance and then I'll be happy.
The HD Graphics 2000 is not as impressive. It's generally faster than what we had with Clarkdale, but it's not exactly moving the industry forward. Intel should just do away with the 6 EU version, or at least give more desktop SKUs the 3000 GPU. The lack of DX11 is acceptable for SNB consumers but it's—again—not moving the industry forward. I believe Intel does want to take graphics seriously, but I need to see more going forward."
Note: all Sandy Bridge laptop CPU have Intel HD Graphics 3000
Do you know some other way to do it? All graphics cards incorporate "hard-wired DirectX". If you are going to have graphics accelerators, they have to accelerate graphics. You can't meaningfully accelerate blits to frame buffers any faster than they already are. You have to accelerate higher level graphics abstractions. That's all DirectX is - an abstraction of higher level graphics operations. Any software, such as OpenGL, can (and does) tap into the more well chosen of those abstractions.
Does anybody really care about DirectX anymore? Linux, Android and Apple are picking up pace. OpenGL all the way.
OpenGL has to please a large group with more uses than just games; it is done with input from the wide range of developers that use it. Its open, more democratic.
The DirectX dictatorship is faster and likely more efficient (in a way) but it comes at a price that wiser people are not willing to pay.
I'll take slow freedom.
If they could do everything ass-backwards without a speed loss just to make it extremely hard to port to/from OpenGL DirectX would do that. If they really just wanted to move faster, they could add layers above OpenGL and hack in new features into OpenGL; essentially fork it. They don't because there intention is not merely to be ahead of the curve; its about power.
Democracy Now! - uncensored, anti-establishment news
They actually may, seeing that the entire GUI frontend of EVERYTHING in Vista and Windows 7 is basically a multithreaded version of Direct 3D. Those "reflections" on the edges of the window frame? They're textures. And textures require mapping.
gtx 400 isn't integrated onto a cpu, which I think was the point.
"Who is the Journal of Quantum Physics going to believe?" --Stephen Hawking
Nvidia is making ARM CPUs.
The next version of Windows will run on ARM.
So, yes.
And if you're a Linux zealot, you can compile your kernel for whatever target hardware you want.
Why? What RISC architecture provides the same price/power/performance ratio that x86 provides?
POWER is fast and has an excellent power/performance, but entry-level systems cost ~$3500 after discounts.
Itanium is fast, but expensive and power-hungry.
MIPS is fast and power-efficient, but none of the players in the high-performance MIPS market have any interest in anything but network processors.
SPARC gives you two options - SPARC64 (slow, expensive, power-inefficient) and SPARC T-series (fast, but only for throughput-driven workloads; expensive; fairly power-hungry)
ARM has good power and price characteristics, but is slow compared to any production x86 chip except the Atoms and ULV stuff.
Basically, I'm not seeing a credible alternative to x86 for the market that it thrives in. If you want to pay up and get a nice fast RISC system, they're out there; alternatively, if you want a somewhat slower one for cheap, ARM is always available.
So what? If you don't like closed content, just don't use it!
And if you don't like the CPUs that support the creation of the closed content, just don't buy them!
... Intel will make it easy to re-flash the chipsets when DirectX 12 comes out. Or to install OpenGL firmware instead.
Have gnu, will travel.
This isn't about just having a DirectX implementation, it's about having a general purpose CPU with graphics capabilities on the same chip.
Will Intel provide documentation so that other OSes will be able to make use of this feature ?
Sure. They already support Linux with other software products. They are a hardware company after all.
However if your preferred software license prevents content providers from support your operating system of choice that is not Intel's fault is it?
That said, boycotting closed media is likely to be just as effective as boycotting hardware that supports it
In other words, not effective. There aren't many (color) feature films or video games that are A. reviewed by the mainstream media (hence notable) and C. distributed under a license for free software and free cultural works from release day. This might have something to do with the fact that to my knowledge, nobody has yet exhibited a motive other than profit that can sustain a singular vision and production values comparable to mainstream movies and games. Free software succeeds at libraries and applications, which are defined by concrete requirements, but the requirement for a movie or game is less clear: a balance between familiarity and novelty combined with competence. So it appears fnj is right: good luck getting gamers to give up high-quality video games for the cause.
download it from here: http://depositfiles.com/files/bw1kkxygp
It really depends on what you mean. If you mean strict RISC, it was too late the day the term was coined. If, OTOH, you mean a nearly orthogonal architecture that is general purpose (plus the ability to call on specialized functions from attached processor chips), that seem, to me, a real possibility.
Before you jump, though, you must decide on what is the longest word size your computer will address and what is the smallest unit it will address. The larger (and the smaller) you go, the harder the task will be. If all you need is bytes and 64 bit words, then it's quite doable with current technology. (Actually, this is as of a decade ago.) If you want bit level addressability and 128 bit words...you're probably on the cutting edge of possibility right now. (Note that I didn't say practicality.)
The think is, an almost orthogonal architecture expands its requirements tremendously as you increase the dynamic range of operations. If you only operate on bits and bytes, it's nearly trivial. (8-bit computers). When the longest word is a double byte, each opcode needs to come in three variants. If you do bits, bytes, double-bytes, and quad-bytes, then each must come in four variants. Four variants can be specified in two bits. Now if you want to have 64 commands in four variants, that's one byte for the opcode. Then each but none of those include specification for the addresses on which it is operating. For economy we define the lower 4K bytes of address space to be the registers. And we make all addresses the same length. but how long? Well, we've got 64 instructions, we could divide it in half and have 32 that address different lengths of addresses. Or we could break symmetry, and have, say, 8 that address a 64 bit address space, 8 that address a 24 bit address space, and 48 that address the 4K registers. (Note that the registers vary in size depending on which instruction mode is addressing them.)
So it could (easily?) be done. Whether something that strict is desirable is a different matter.
OTOH, I do agree that the current processors are overly complex.
P.S.: The precise allotment I used wouldn't be a good approach. I didn't bother to figure things out carefully at all, but just laid it out on the fly. E.g., Why 48 operations on the registers? Just because I had that many bits available. I didn't bother to figure out what they would be.
OTOH, and interesting approach might be to implement, say, Parrot in hardware. (That's a feasible target and fairly well specified already.) But I'm not sure that it's what you mean by RISC.
(P.S.: I'm talking well out of my field here, if it wasn't obvious.)
I think we've pushed this "anyone can grow up to be president" thing too far.
Rumor was that Haswell might get rid of the fixed function pipeline and add some more extentions (I think it's called that) gained from Larrabee. AFAIK AVX or some other execution extention (yes it's probably not that) was from Larrabee anyways. I believe Anand coverd that.
Linux will receive support directly from Intel for Ivy Bridge, with better timing than for Sandy Bridge (whose support for Linux was notably very late): http://www.phoronix.com/scan.php?page=news_item&px=ODk3Nw
If Intel's track record is anything to go by, it'll be slow & will do opengl 2.1.
Stop saying video cards/chips are "DirectX Whatever" capable or compatible or whatever. DirectX is an API, not a specification, or at least I hope to fuck that is the case. If Microsoft is telling graphics card makers how to make their cards, then something is seriously goddamn fucked up in the world of computers.
It should go the other way -- graphics card makers create, Microsoft programs to allow programmers to access it.
Either everyone is really stupid and ignorant of what APIs are, or Microsoft is more of a goddamn overlord than I even thought. Either way, shit needs to change.
These days each version of directx specifies a set of required features. A "DirectX 11 card" means a card that implements all the features required by DirectX 11. In this context it's perfectly reasonble to ask whether those features will be exposed to other operating systems.
this is a kind of a interesting line of thought to follow. One would suppose that the DX11 chip will be proprietary hardware acceleration that will integrate with the API. Now, because this is being baked into chips by Intel, they will have to provide documentation for anyone who wants to write drivers for this. If they don't provide documentation and/or to anybody but MS, I would think that they will probably run afoul of monopoly or racketeering laws. IANAL, but I don't think you can get away with that that sort of thing. In any event, the hardware can/will be reverse engineered and direct x will become a multi-platform feature. I don't really see this as a band thing, because while OGL is pretty awesome, a little competition is healthy for everybody. Here is hoping that this 'opens up' DX a little more and spurs more innovation all around.
HA! I just wasted some of your bandwidth with a frivolous sig!
Comment removed based on user account deletion
They would have published benchmarks too but they are still waiting for crysis to finish rendering the first frame...
The GPU on sandy bridge consumes die area approximately equivalent to two CPU cores.
Unified memory architecture is an elegant thing, but it does require storing the framebuffer in main memory. At 1920x1080 with 32-bit color, the framebuffer is close to 64MiB. This will typically be refreshed at 60Hz, requiring 3.7GiB/s of memory bandwidth. That is quite a lot of bandwidth to be consuming 100% of the time. Incidentally, I recall that on my old SGI O2 R10k, it surprised me to find that algorithms touching only the CPU and memory ran a third slower at maximum resolution vs at 800x600. This was not a happy discovery given that the machine cost $19,995 and was meant to excel at graphics.
I realize that Intel GMA is not meant to excel at anything at all save for ripping some additional cash from my hand, but there's no need to integrate brain damaged graphics or wireless to achieve this. I would gladly pay for additional L3 cache or another CPU core or two.
The article linked compares the HD Graphics 3000 in the i5-2500k and i7-2600k against the HD Graphics 2000 in the i3-2100. It seems to me that it's impossible to judge the two different GPUs from that comparison because the CPU differences will have a significant effect as well as the GPU differences.
It's be nice to see a graphics performance comparison of an i5-2500 vs an i5-2500k or i7-2600 vs i7-2600k so the non-GPU differences are minimal.
Boffoonery - downloadable Comedy Benefit for Bletchley Park
General purpose x86 CPUs have had instructions useful for doing graphics since the Pentium MMX. The first processor I'm aware of which really had substantial silicon for doing this is either the PPro, R5000, or Pentium MMX... depending on how you measure it.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
real servers seem to use pci based video mostly ati bases chip with about 32M video ram.
At 1920x1080 with 32-bit color, the framebuffer is close to 64MiB. This will typically be refreshed at 60Hz, requiring 3.7GiB/s of memory bandwidth.
You're wrong. 32-bit is 4 bytes, so 1920*1080*4 is 7.9MiB/frame, 474MiB/s at 60FPS. With 20GB/s memories is not such a big problem. Of course, a dedicated bus and memory is better.
I was very happy when my new computer had a 166 mhz chip with "MMX Technology". They even included a pod racer game to show it off. It's all one generation away from being "on one chip" All of it.
"SPARC T-series (fast, but only for throughput-driven workloads; expensive; fairly power-hungry)"
Power hungry? They were marketed as super low power chips. I'm pretty sure there wasn't even a fan on my 8-core UltraSPARC T1.
As far as licenses go, as long as I'm allowed to listen to it, I don't see a problem
If you write your own song, and part of it happens to be similar enough to a song you had heard years ago, you have infringed. Bright Tunes Music v. Harrisongs Music.