NVIDIA's GT300 specs outed -- is this the cGPU we've been waiting for?
By Darren Murph 
posted Apr 26th 2009 1:22PM
posted Apr 26th 2009 1:22PM
[Thanks, Musouka]
posted Apr 26th 2009 1:22PM
Estimated time drivers save by using real-time GPS traffic updates
Drivers that use GPS systems with real-time traffic info spend 18% less time behind the wheel than those who do not.
Now that we've thrown 'em off the trail, use the form below to get in touch with the people at Engadget. Please fill in all of the required fields because they're required.
Hey ATi - you hear that?
RELEASE THE 5000 SERIES ALREADY
4890 JUST came out...
I said 5000
If Intel releases motherboards with chipsets that can do what an Nvidia 280 can do (and I'm certain they can) Its gonna seriously put a dent in video card sales if they sell to manufacturers like HP and DELL to have these things "Crysis ready" off the shelf.
Graphics Processors made by AMD/ATI and Nvidia are big, powerful, and power hungry.
There really is no way around it.
Even cards with the smallest die size are in the 200+ watt range. There is processing power in a graphics card that Intel chips cant even touch.
Sure, they can manage to create a similar processor to Nvidia and ATI, but to have it integrated in all desktops? I seriously doubt it. It costs alot of money to manufacture a 2 teraflop processor.
And when Intel gets GTX 280 class processors so small and cost effective to be integrated in the weakest of devices, Nvidia and AMD/ATI will be well beyond that processing power in their discrete solutions, and games will be requesting power from those more powerful graphics chips. Why?
They dont have the limitation of size that is required for an integrated solution. And the amount of room for an integrated solution is substantially more restricted than that of a discrete solution. A single Nvidia processor takes up more space than a quad core processor, a North and South bridge, etc. They are mammoth. And there is no way around it.
ATI needs to first finalise ruining the driver for this on 64bit OS's before they release hardware, can't have people running modern OS's left and right can we now? Where's the pain in that? Next you'll suggest they start to respect customers, that's just not done in this filed
.
yeah tell ATI to make drivers that arent a total clusterfuck and then well talk.
do you remember the GPU vs CPU debate? the thing is, no one can compare between them as they process different types of code (scalar and vector i think). Therefore no one can compare which company will release the better CPU/GPU.
Besides Nvidia will probably use a mixture of the i7 architecture and the new GTX300 series to try create a new GPU/CPU multi-purpose processor. I can however predict that it won't run very well as it is new technology that has yet to be perfected. we will see it lagging behind in performance and power usage but advance in terms of size (uses less space as there doesn't need to be a GPU)
my guess is that anyone who thinks ATIs drivers such either a) havent had an ATI card in years or b) runs linux
because I have never had a problem with my ATI cards in windows. in ubuntu on the other hand, ati drivers blow.
I think Random and QuantumPhysics are 14 year old idiots with no grasp of what any of this really means.
Better luck next time guys, TIME FOR BED
The only safe prediction is that both Intel and NVidia will continue to release products and Engadget readers will continue to vomit verbiage all over topics that they do not understand.
As for me, I plan to sit at the sidelines broadcasting my critical commentaries which add nothing to the discussion and serve only to illustrate how much of a cock I am.
Regardless, any competition has got to be good for the consumer, eh?
so this one is gonna cost $1000 and take up 3 slots.
@thatrotierkid
Try installing a 64bit windows and then one of the last 4 ATI drivers and say again they don't suck, wait you can't because your system would bluescreen at boot, never mind then
FIRST!
............ to type a good response...... This looks cool..... needs an 'X' after that 'GT' though. haha.
Last
...........to be picked for a middle school kickball game
well at least they're being consistent with the naming scheme
GT300 is the chip name. The card will most likely be called the GTX 380 (for the high end part).
..... you are all aware that I was making fun of people who do the "FIRST!" thing...... right?
Your advanced humor is understood by few.
Your advanced humor sucks.
stop trying to copy me >> look at facial recognition post http://www.engadget.com/2009/04/25/lip-reading-software-can-identify-multiple-languages-has-big-pl/
except that wasn't a good response...
Woah.
Yeah, if I can finally offload some software rendering threads onto this beast, that'll be a big WHOA on my side too!
Nvidia will reign forever.
And even when they don't, their drivers should still be at least 10x better than AMD/ATIs.
Ati graphics drivers have always been better than Nvidia in my experience. Nvidia drivers always cause problems part of the reason I will probably buy ATI in the future.
That's how it goes, you get frustrated by idiotic driver issues on nvidia and go ATI, then get frustrated with the same crap with ATI until the memory of nvidia fades a bit and go nvidia again, rinse and repeat ad infinitum.
It's very similar to politics and voting I guess.
Last time nVidia "reigned" we had 350+ dollar cards, that were freaking Pieces of S^%@, until ATi finally pulled up their pants, and immediately the cards price were driven down. Competition weeds out the weak.
actually thats BS, ati cards have way better dual GPU support.
and what they lose in driver support they make up in ridiculously powerful hardware.
Nvidia isn't immune to the odd stability blip with new drivers but ever since AMD bought ATI their drivers have been buggy as hell.
Life - Graphics by NVIDIA
I don't know what you guys are on about. I had over one years worth of BSOD with my nvidia 8800 gts, from the driver failing every hour to a total system crash on a daily basis. Replaced card, did not fix it. New drivers fixed one problem created another. Yet never actually addressing the previous issues.
Switched to 4870 ATI, have not had a single driver crash or BSOD in 7 months now. Basically since replacing the card. 64 bit - Vista.
So what exactly does this mean?
In essence if you can run everything through a single processor significant power reduction. It also means that we could be seeing companies focusing on 1 single chip rather than 2. It really paves the way for new chips setups for laptops and other mobile devices.
From the article: "NVIDIA could be right on the cusp of delivering a single chip that can handle tasks that were typically separated for the CPU and GPU, and we needn't tell you just how much your life could change should it become a reality"
Thus, a single chip to handle both. No offense, but it didn't require anything more than crude comprehension.
It just says in florid language that nvidia will release a DX11 compatible card soonish.
It means, you buy your cpu, you buy your gpu, and you buy a tesla to do everything else.
These are about $2000 at the moment i think, and they take up 2 x16 ports, and have no I/O
A Tesla is actually $1699 MSRP, but you can get them for less.
Tested some, they're on the order of 500x faster for certain tasks than a Core 2 Duo. The GTX295 (each GPU) has the same 500x performance increase, for programs using less than 900MB of space. Remember tho the Tesla has an ungodly 4GB of memory onboard for calculations.
All this probably is just a GPU that can handle more data types for G-GPU a bit better than current solutions.
To run x86 code would mean signing away your soul to Intel and there is no way in hell Intel is giving Nvidia x86 rights when the two companies are on poor terms.
Also the whole hype of or GP-GPU is getting silly, there are barely any uses for consumers and the most common one (media encoders) are still inferior to CPU based solutions (e.g x264)
Intel cant refuse letting Nvidia to use x86
That's like saying Microsoft refuse to let hp use windows
Intel tried in the past to stop nvidia and they have failed
@Dr House,
No. It is nothing like your examples at all. The point is that Nvidia can't create a chip that directly processes x86 code, because Intel hold patents and won't license it to ANYONE, except of coarse AMD, because of cross license agreements stemming from a lawsuit settlement eons ago. Nvidia woult have to acquire the license some other way, like buying another company that already holds license agreements with Intel, like Via. (also the result of cross license agreements stemming from a lawsuit, I believe, but I'm not sure about that).
glenn s:
It has held been that the instruction set of a chip is not copyrightable and the x86 instruction set (pre-SSE) is not patented.
Also, in your licensing stuff, I think you're confusing arguments over interface (FSB) licenses confused with the instruction set.
Yeah, the replacing a CPU thing is just BS from Engadget because they don't understand CPUs. For one thing, this is still running on the PCI-E bus, which is far too slow to be used as a general purpose bus. Also, with AMD chips since the Athlon 64 and Intel since Nehalem, the memory controller is on the CPU die, so as of right now this card would still have to go through the CPU to access main memory.
Furthermore, this isn't x86 (should be obvious for a site claiming to be a gadget blog), and it'll be a while yet until Intel licenses x86 to someone other than AMD. If nVidia is making a CPU, it's almost certainly a RISC CPU, and probably an ARM variant, and it's definitely not taking on Intel in any way, shape, or form outside of perhaps challenging Atom, which, on an unrelated note, is a shitty CPU.
Seriously, this is still so far from a CPU it's not even funny how bad the engadget summary is. Also, what the hell is a cGPU? Sure, this will be nice for the same areas that GPGPU stuff is helpful for already, i.e. massively parallel floating point number crunching, but it's not going to expand the flexibility of GPGPU by much, if at all.
@why not the LS2LS7?
Then it's settled, we'll make a chip without and of the SSE instructions. Even without a GPU it's sure to be a top performer...
owwwwhhh come one man!, give Badaboom or ATI Video Transcoder a spin, they transcode video at 35 fps vs 5-10 fps on fastest CPU and you can still use your CPU for other activities, plainly brilliant, cheers to them both. I still recall having to leave transcoding task overnight!!!, this is the future man
Didn't we move away from one chip handling both graphics and normal processing for a reason? Are we really at a point now where that reason isn't true anymore?
We never moved a away or one chip for all. Both Intel and AMD are set to bring one chip for all solutions the next couple of years or so. There will be a long time before on chip solutions will provide graphics good enough for mid gaming or above. Unless AMD work heavily with ATI. Intel chips will be media accelerator or future larrabee tech.
At no point did anybody say the next FPU generation from either ATI or nvidia would run x86 code, it dies the FPU stuff if it's suitable, and that's it.