Nvidia GT300 core: Speculation

Or is it 32KB per 16-wide SM (two in a cluster) for a grand total of 512 SPs in 16 clusters and 1024KB array?!

 

What's he talking about? An SM previously referred to each independent 8-wide SIMD of which there are 30 in GT200. He's referring to each SM as being 32-wide and only having 16 of them per chip. Is it safe to assume he's mucking up the terminology, the architecture really has changed, or he's making up the whole thing?

 

It really has changed. I can't say (well I could) if Theo's right or not, but GF100 is not terribly GT200-like in places. All will be revealed later today anyway, not long to go now.

 

http://www.fudzilla.com/content/view/15741/1/

Yep, so Fuad says as well. JHH will give us the business during his keynote at 1pm EST. Delays aside, it's good to know we'll have something new to dissect over the next few months

 

My present next year

 

I hate next year

 

Yea, today's keynote is a key point for me
Will they give the final specs? Will there be actual working silicon? Will they mention a launch date?

I feel like I'm in a black hole in terms of upgrading right now. nVidia doesn't have DX11 yet, and not sure when they'll have it, or how good it's going to be.... AMD doesn't have OpenCL and DirectCompute yet, and no clue about when they'll deliver that.
I'm only going to upgrade to a card that does DX11 AND OpenCL/DirectCompute, out of the box (my 8800GTS already does DX10 and OpenCL/DirectCompute, would be pointless to get a HD5850 now if I still need to use my 8800GTS for software development, which is my primary use for the card).
Will be interesting to see who's first to offer me what I want.

 

I just hope Nvidia isn't trying to emulate Intel and demo hardware a year before it hits retail

 

You know I was just thinking about this, and the info black hole GPU makers try to exist in.

It's not so bad for Nvidia fanboys that they have nothing out, it's that they dont have the slightest clue even when to expect something.

I dont see why the GPU makers in a situation like this wouldn't say something like "well we hope to have our next GPU's out by Dec" or something to that effect.

I could see a lot of people then going, "ahh well, I'll wait til Dec then".

Maybe I'm wrong, maybe more people would go "Dec, thats crazy, well I was gonna wait but I'm buying ATI now!"

But I dont think thats the case.

 

A year isn't good enough *cough cough cough*

 

Just because JHH will show it doesn't necessarily mean we'll get to see some nice details.

 

Because if you are unable to fulfill this promise you'll get burned much harder than if you don't say anything at all.

 

Problem is that if you are late you will be crucified, and you also cannibalise sales of your own current products if your potential customers are waiting for your next big thing. If you need an extra respin, that's pretty much another quarter.

Is this the first big leak? Sounds impressive on paper, though I do wonder if all the focus on GPGPU means that the gaming side of things will be taking a backseat.

I can't see how this is going to be easy or cheap to build, which might account for the rumours of paper launch now, soft launch in Dec with hardware late in Q1 10, along with yield issues. It's a monstrous, and interesting new direction, if it lives up to what Nvidia intends and can be delivered at a reasonable price. It's a different product that might not have to compete with traditional GPUs in price or market. It doesn't look like they've abandoned the drive to ever bigger, more complex chips at the top end.

 

If we're talking about Q1 2010, then it's only couple of months left until ATI refresh/shrink comes out - might as well wait for that!

 

WTF "native C, C++, Fortran etc" means?!
NV were able to make ANSI C++ compiler for the new chip?

 

Or that's what they intend to have at some point in the future. It looks like that instead of making their own CPU, they made their GPU very CPU-like, and then gave it standard CPU-like interfaces for people to program it.

 

Assuming ATI doesn't need the refresh just to catch up to "Fermi" Btw, who's going to have 32nm by mid-2010? I thought all current roadmaps were pointing at Q410 for the first prodcuction runs?

 

Shouldn't this change as they start targeting the HPC audience like they are at GTC? It's probably not a good idea to keep those guys in the dark until stuff ends up at Newegg. They have long lead times for their purchasing decisions so if you want to make it into their budget they need to know what you have coming.

My team just put in our reqs for the 2010 hardware budget and it's based on upcoming Beckton based servers. We obviously couldn't do that if we didn't have advance notice from the vendor. I presume the same dynamic would apply to Tesla.

 

Also:
"In a nutshell, this is just baseline information about what nVidia is going to introduce in the next couple of weeks"

Hopefully in a few hours we'll know if this article is true or not. If it is, that's excellent news.

 

"Introduce" doesn't mean ship. As was pointed out further up this page, they could do an Intel and be talking about this tech a year ahead of shipping products. That kind of advance info would be very unusual for Nvidia to send out, but I guess they have to combat 5800 with something.