Slashdot Mirror
NVIDIA Launches GeForce GTX 560 Ti 448-Core GPU
MojoKid writes "NVIDIA has just launched the GeForce GTX 560 Ti with 448 cores. Though perhaps a bit unimaginative in terms of branding, the new GeForce GTX 560 Ti with 448 cores is outfitted with the same GF110 GPU powering high-end GeForce GTX 570 and GTX 580 cards, but with a couple of its streaming multiprocessors fused off. The card has 448 CUDA cores arranged in 14 SMs, with 56 texture units and 40 ROPs. Reference specifications call for a 732MHz core clock with 1464MHz CUDA cores. 1.2GB of GDDR5 memory is linked to the GPU via a 320-bit bus and the memory is clocked at an effective 3800MHz data rate. Performance-wise, the new GPU proved to be about 10 to 15 percent faster than the original GeForce GTX 560 Ti and a few percentage points slower than the GeForce GTX 570."
I bought a 560 Ti just a month ago and now this? FFFFFFfffffffffff...
They couldn't call it a 561 Ti, or a 560 Pt? Or Au, or Ir, or whatever other element is "better" than Titanium.
"but with a couple of its streaming multiprocessors fused off."
fused off? Really?!
I don't know how they would do that, except than rather not connecting them in the blueprints.
Or are they just "defect" 570 and 580 relabeled.
Yay gotta get me one o' those!
The summary doesn't make it clear, but... how many cores does this new video card have?
What is with the branding scheme on these things? I see a summary with lots of letters and numbers and almost no useful information as to what the hell good they all are.
I may have to share this planet with animals, but I'm doing my damn best to eat every last one of them.
They already implied that these are salvaged 570 parts for a limited holiday production run. If you snag one with a free game you want, not a bad deal. Think carefully if you might go SLI in the future, since they'll be hard to find later.
all parts have defects. although sometimes companies don't test throughly enough to find the defects. and usually the defects don't impact normal operation. But when the potential exists for problems, they are forced to either scrap the part or bin it as a lower spec part. Binning improves yield and helps keep prices down on the higher-end parts that do pass tests.
But here's the problem with binning from a marketing standpoint: "and a few percentage points slower than the GeForce GTX 570". This binned 570 is about $60+ cheaper and will likely slide down to the old 560 Ti (naming is confusing!) prices. So now they've created a cheaper version that is almost the same performance, and run the risk that customers will choose the cheaper product over the more expensive (and I assume higher margin) product.
“Common sense is not so common.” — Voltaire
As a whole, it's impressive that we can build such a thing. It's equally impressive that the number one reason for such an advanced piece of technology is so people can virtually shoot the current unfashionable eastern europeans by using more polygons.
I still can't fathom spending $300 on a video card....and feeling like I got a slammin deal in the process.
What happened to the red-hot competition of 2008, when I built my first modern system and got a newly released Radeon 4850 for $150? That card was maybe the fourth most powerful you could get; there was no serious improvement to be had without adding more dies, via either X2 cards or crossfire.
Now today the 560 Ti and the 6950 occupy the same relative position in the hierarchy of GPUs that my 4850 held in 2008...yet rather than being brand new and $150 those two cards are almost a year old and $250-$300.
Ouch.
Nice summary, more redundancy please!
Never let a lack of data get in the way of a good rant.
That's pretty impressive.... It takes the current crown for fastest video card... all so I can play a dumbed-down Skyrim with its dumbed-down console interface and low-res console textures. What is it *for*?
More interesting news is that Nvidia's next high end card will probably arrive in late 2012, or even 2013 - while AMD's high end card Tahiti is expected in January 2012.
http://www.edn.com/article/520175-Nvidia_Kepler_GPUs_to_trail_AMD_s_next_generation.php
Bitcoins anyone? Seriously though, A lot of people don't quite settle for the onboard GPU, especially if you like to game with a very high resolution, or multiple monitors. I wouldn't even consider myself a gamer, but I run discrete in my desktop, though it's a now aging ATI Radeon HD 5770, it still works well enough for my needs. My biggest reason is multiple displays supported well. Most discrete graphics don't have multiple digital connections, so one will be via VGA, and invariably the colors won't match between two of the same monitor, which is far more annoying than anything to me, and calibration was still too far off for my tastes.
The point is far more people are willing to buy these. I happen to favor mid-range cards for myself, and usually will start with onboard graphics for anyone that doesn't play 3d games.
Michael J. Ryan - tracker1.info
Impressive! Imagine what a Beowulf Cluster of these things could . . . wait a minute! This chip IS it's own Beowulf Cluster! =)
There is most certainly a use for high-end GPU nVidia cards. Like in supercomputing applications,. . .
Does it make my speedy bird go really really REALLY fast?
Vote monkeys into Congress. They are cheaper and more trustworthy.
Seriously though, A lot of people don't quite settle for the onboard GPU, especially if you like to game with a very high resolution, or multiple monitors.
I sure as hell wouldn't be able to settle for an onboard GPU, they're simply too slow. I already cringe whenever I'm away from home for longer periods and have to kill some time by trying to play on my laptop, and these days I'm starting to feel my overclocked Geforce 460 on my desktop is starting to get slow and needs a replacement soon.
I'm in the same boat. I run a discrete graphics card on my desktop because for about $50 I got a video card that runs all of the games I play (I only buy video games when they hit the $10 bin at [big box store]), while the on-board graphics card won't handle it. Of course, my desktop is about 8-10 years old now so that might be a bad example.
Check out my lame java blog at www.javachopshop.com
Still trying to figger out the method to the madness of going back wards. They had a 9000 series and then a 500 series. Then a 200 series. I'm waiting until they come out with the zero series. And it's not just Nvidia. Seems a lot of companies are doing it.
Wuddooeyeno? IITYWYBMAD? Like nuts? eclecticallyincorrect.com
Its pretty difficult to get precise power figures on graphics cards - reviews always rate against 'total system' but never give us for reference the system power use without the card (or perhaps an onboard video solution). In any event, all modern cards are total power pigs. At a time where Intel and AMD try are trying very hard to reduce cpu power consumption, graphics cards are using up many multiples of those savings. I'm not sure where 'it has a wall outlet plug' gave these card and gpu producers license to subsidize the power companies.
Meanwhile, at AMD/ATI Headquarters:
"Well, fuck it. We're going to 449 cores."
For a very few specialist problems. Just like FPGAs.
On the whole, are they're useless for most applications requiring performance. A lot of people bought into the hype.
.
ATi redefined what they call a "core" some time ago to basically mean each sub processing unit of their cores so they say their Radeon 6970s have "1536 Stream Processors". I'm sure at some point nVidia will redefine theirs to mean each bit of an operator or something and we'll have cards with millions of "cores" before long.
nVidia and AMD do it different. nVidia counts their processing sub units as a "core". They aren't quite a core as you'd think of it on a CPU, but similar. AMD counts each execution sub unit in their "cores" as a "stream processor". So roughly speaking for the 4000 and 5000 series there are 5 SPs for each core.
That doesn't quite tell the whole story though as what each core can do is different between the different vendors. More or less, to the extent it is useful information at all, it is only comparing within products of one vendor.
Also there are other things not advertised as much that matter. That would be things like ROPs (Render Output Unit) and TMUs (Texture Mapping Unit). Those deal with rasterizing pixels and textures respectively. So those relate to rendering speed as well. Those used to be all that graphics cards talked about, before they had the processing section that makes them a GPU.
Really the thing to do is this: Decide how much you can spend, decide what you wish to do, then go read reviews that show how the cards actually work with that software. For me as a gamer, I read HardOCP. They take the cards and actually play games with them, and you can see how well they fair.
Solidworks plays fine with nVidia cards. Probably better than an ATi card since nVidia's OpenGL drivers are top notch (just as fast and features as their DX drivers) and ATi's are not quite as optimized. In terms of a nVidia 210, well those are very low end, lower than a 4650 so I don't know how much it'd help. Still probably be head and shoulders above the integrated graphics though.
Is you pay by the wafer when making chips. A wafer will cost a certain amount to make depending on the process, the size, the fab and so on. That is how the company that makes GPUs is charged. So the more chips that come off the wafer, the more the cost of that wafer can be spread out. That means not only having smaller chips, which of course can fit more per wafer, but having less defective chips.
Hence binning based on units that work (or don't). As you say, it brings up yields and thus brings down unit cost.
This is particularly useful for large chips. Defects are the kind of thing that occur at a certain frequency per wafer. So if you have a wafer that has, say, 4 defects roughly evenly spaced around it, you probably have 4 chips with failures. If your chips are small and you fab 2000 of them per wafer, no big deal, just toss them. However if your chips are huge and you fab 20 of them per wafer, that is a massive amount of failures. If you can instead simply shut down the damaged section and bin it as a lesser chip, you can bring yield back up.
That's also why chips that are designed smaller can cost less. That is more or less what lower end GPUs are. They just design a smaller chip, with less of the units form the bigger one. You can then put more of them on a wafer, and thus they cost less.
That's starting to change with things like the AMD Fusion chips, but the key phrase is "starting to."
"[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz
Or all these companies (AMD, nVidia, Intel) slapping more "cores" on their products and only yielding 10 - 15% performance improvements kind of lack luster? I think if companies can't at least double performance then they shouldn't bother releasing a new product. The trickle release of half-hearted improvements in performance has to end.
I haven't thought of anything clever to put here, but then again most of you haven't either.
maybe the 560 Ti's will come down in price to that magical 150-175$ threshold to which I refuse to spend over for a video card.