The First DirectX 11 GPUs
- 3
- Add a Comment
I have always wondered why Microsoft has insisted upon codifying graphics unit capabilities in such strange ways. Though it is certainly a good thing to have a listing of what features are necessary for a certain level of performance, it has always seemed as though the Microsoft people who make the pronouncements of new features have not gotten together with the designers of the graphics chips themselves, or those who design the programs (usually games) that will use the feature set.
If there was more communication between these groups of people, it would be easier to pick a graphics card for a specific purpose, and perhaps one measurement alone could identify a card’s fitness to purpose. I envision this as something similar to gasoline, and the octane rating. Major manufacturers of gasoline put different compounds into gasoline, to enhance its antiknock capabilities, and make the drivability of your car better, but you don’t really need to know the fact that one uses aniline, or another uses something else to take the place of the tetraethyl lead now banned by the government. No, all that the motorist needs to know is the octane requirements of his car, and once procuring it, needs to think of it no more.
Would it not be nice to be able to look at GPUs from nVidia, ATi, and any other player (come on S3, you can do it again), know what your software required, and then decide according to the dictates of your budget?
Anyway, Microsoft has come up with yet another specification, DirectX 11, and of course, it will be used a pushing tool to get people to give up their beloved Windows XP, as it is almost certain that DX11 will not be supported on Windows before Vista.
Adrian Kingsley-Hughes has a note up this morning, citing a German site having a few specifics for the next generation of GPUs, all 40 nm process, and using impressive amounts and types of graphics memory. Also prodigious are the claimed performance numbers, in gigaflops, but alas, still no single octane-like number. Perhaps another day.
AMD/ATI
Radeon HD 5870
GPU clock: 900MHz
Shader clock: 900MHz
Memory clock: 4,400MHz
Memory: 1GB GDDR5
Manufacturing process: 40nm
Peak GFLOPS: 2,160Radeon HD 5870 X2
GPU clock: 950MHz
Shader clock: 950MHz
Memory clock: 4,600MHz
Memory: 2GB GDDR5
Manufacturing process: 40nm
Peak GFLOPS: 4,560NVIDIA
GeForce GTX 3xx (G300)
GPU clock: 700MHz
Shader clock: 1,600MHz
Memory clock: 2,200MHz
Memory: 1GB GDDR5
Manufacturing process: 40nm
Peak GFLOPS: 2,457
AKH also notes that, in each case, the GPUs are set to be, if the numbers are to be believed, at least twice as powerful as current zeniths in each manufacturer’s stable.
The thing is, we have all seen the raw power of cards go up, yet certain very powerful cards, in a rush to get them to market, omit one or two specific features, needed to fall in line with a given DX level.
It’s probably too early to tell, but these cards are most likely out of the buying range of most who will drool at the cards claiming those specs.
§
.
.
 |
Nothing is illegal if one hundred well-placed business men decide to do it. |
•
Californians – Don’t forget to vote today!
3 Comments
Minute Flourishes in History » Blog Archive » Quick scan of the net - radeon hd
May 19th, 2009
at 9:18am
[...] http://www.lockergnome.com/theoracle/2009/05/19/the-first-directx-11-gpus/Radeon HD 5870. GPU clock: 900MHz. Shader clock: 900MHz. Memory clock: 4400MHz. Memory: 1GB GDDR5 Manufacturing process: 40nm. Peak GFLOPS: 2160. Radeon HD 5870 X2 GPU clock: 950MHz. Shader clock: 950MHz. Memory clock: 4600MHz … [...]
Kitty
May 19th, 2009
at 9:37am
This is overtly cynical, and I think you’re really missing the point what they do.
Windows XP is going away, for better or worse. At least Microsoft does actually push for better technologies, unlike other companies who shall remain nameless and lock their users into low end cards.
Your gas anology is ironically appropriate, as the reason we pay so much for gas is that the cost is so much higher to produce different types of gas. In order to produce fuel, you have to put it through a catalytic cracker and feed it through a distillation tower. The same barrel of oil can be split into multiple parts, or simply constituents like kerosene and diesel.
The reason gas costs as much as it does it we have so many octane levels. Diesel is every bit as clean as the unleaded gasolines you see, and if they produced only diesel, you would see the price of gas get cut in half. However, since the oil companies have to use more energy intensive processes to break them down, store them seperately, ship them seperately, and then the stations themselves have to pay to use detergent packs at the end point and have several tanks for each type, the price gets passed on to you. Btw, all the gasoline shipped around the country, is the same, and detergents like ‘techron’ are identical to what you get at valero, exxon and everywhere else. Ethanol content used to be a factor, but everyone has it now.
This is precisely what happens when you ask manufacturers to produce, stock, store and ship multiple levels of GPUs as rated by purpose.
I find the way microsoft conducts direct x far better than the way opengl works. Not only is opengl rarely updated, it never even attempts to keep up with the times. Coding the same application in opengl takes roughly 3x as long as coding it in direct x on a good day, because a lot of the menial coding tasks and indeed some of the more complex tasks (procedural synthesis anyone?), you end up coding yourself.
The one thing they could do better is make updating firmware easier, but the problem with the way you’re viewing this is that you’re assuming that the hardware manufacturers talk to one another. It’s far easier for cards to be tailored to a unified spec, then for software to tailor itself to multiple manufacturers. I don’t think anyone would ever accuse ATI and NVIDIA of colluding.
the oracle
May 19th, 2009
at 10:29am
You make good points, but I do think that, in general terms, one knows what the other is working on. There is just too much human nature in this for any ‘loose lips sink ships’ mentality to prevail.
Look , for example, at how much alike each series is, then think about the logical jumps to the next thing changed. By doing a little work on the drivers, the competitor can be fairly certain what is going on in the silicon. Reverse engineering is a wonderful thing.