NVIDIA GT200 Rumours & Speculation Thread

LOL, there is no double precision support in any NVidia GPU released so far. That's precisely why this feature is such a good indication of NVidia's failure to deliver.

Jawed

 

I sense you're slightly biased in this...maybe...a bit?

 

I'm not even sure what definition of "executing well" that Jawed is using. He seems to be focusing on meeting internal schedules. That's fine but in isolation pretty meaningless. If you're late internally but still earlier than the competition or significantly faster than the competition most people will perceive that as good execution.

 

In the end, I'm not sure it's even relevant. There has been little, if any proof, for quite a while, that nV is unable to counteract most things ATi does-and ultimately that's what matters. Fantasizing about how internal release schedules are missed/met is...erm...useless?Not in the least because we don't really know how those schedules actually look.

Do they need a part that crushes the 8800GTX? Ehm, I guess we'll find out by seeing how the new(ish) 9 series high-end propositions do.

 

Probably not well. the problem is it's not an attractive upgrade for all the 8800GT/GTX/GTS512/Ultra owners out there, as it should just be a little faster.

I guess they get new customers+everybody who has a GTS 640/320 or lower (it's an attractive upgrade from there), though.

 

That was precisely my point. If they do poorly it means that the market pretty much reached saturation at this current level of performance/features/whatnot, and in order to keep on selling you have to come up with something new(IGPs, lowest of the low-end money making parts are ignored in the above, BTW). If they do OKish, it means that the need for a new part wasn't so dire after all, as there were still buyers for this type of product.

This says nothing about whether or not nV actually met it's internal schedule, BTW.

 

Yes, I'm the late nazi, always on the lookout for stuff that's late.

Jawed

 

I am with you on this one. They are DEFFINETLY not delivering.

 

I'm not sure where the bias is in wanting new tech sooner rather than later--or, to not read even that much into Jawed's post, in noting the obvious (that nV expected [June '07] an earlier [Nov '07] release [either of soft-enabling a hidden h/w feature or of a new GPU]).

Which would be fine from the business perspective, but this is the tech forum.

 

Lol if we go by that then everything is late since we always want everything "now" and we invariably get it later than that

 

Why want something new from nV only?The 3870 isn't exactly monstrously new either, and the base architecture for it(the R6xx) was arguably less innovative than the G8x one. Jawed simply likes red better, just like how some ppl like green better and that's that, and that was what the biased comment was making reference to. And as we all are subjective and as such we're prone to take sides, it was hardly a criticism, merely a statement of fact.

Of course everyone would like new toys yesterday...preferably DX11 capable. Sadly, it seems we aren't going to be getting em soon.

 

While I understand where you trying to get at, I'm personally in general more interested in a good combination of IQ and performance as the highest priority. Compliance is a lot lower on my list. However if I'd have the dilemma between two different GPUs with roughly the same IP/performance ratio, I'd of course go for the highest compliance yet that under certain presuppositions too.

What good would an average user a D3D11 compliant GPU today do anyway? If it would be by X times faster than today's GPUs then the majority would run out and buy it for its performance and not for its (for the time being) useless compliance.

------------------------------------------

As for the ever repeating nonsense of scratched or delayed designs, IHVs release their products when they're ready. The answer I guess will come when we find out the die size of GT200 and it doesn't take that much then to see if the result would had been possible earlier or not. Unless someone of course is as naive and believes that IHVs add or deduct hundreds of square millimeters with a blink of an eye or that they sit on ready design and enjoy losing precious income.

Why is it so important anyway? I mean we've heard rumours also that AMD postponed a design until they can use 45nm and RV770 is supposedly a design that didn't take according to rumours again a huge amount in R&D. If that helps AMD to avoid delays of the past, means outstanding execution as with RV670 and the result is able to corner more than just a bit of today's G9x sollutions, why should anyone care in the end how goal Y has been reached?

2007 has been one of the most boring years ever; it's about damn time that things heat up now that AMD seems to be on an uptake again.

 

Yeah, I see what you're saying Morgoth, but at least Jawed has (as far as I can recall) always been an equal opportunity late nazi. NV has just been better at managing perceptions and general expectations of new product since nv3x and has obviously been able to leverage superior execution and performance to its advantage.

I think NV has also been quite lucky recently: wasn't the DP Q4 2007 boast made in a quarterly CC? How NV hasn't been pinged about this and their very tardy MCP execution by an analyst, I don't know!

 

You said it

?????

Interesting wait and see what happens in June.

 

Valid point, but % of buyers in that range would be miniscule

True

they already have that

They have that already and same as above

 

I included the DX11 thing simply as an example of extreme forward push, having a part supporting the API before the API is even announced. I could just as well have said:a card that runs Crysis(as that seems to be the current favorite) with 8X AA at 2560x1600 with Very High settings.

 

lol I was not sure, I thought I read that somewhere, but I admit it's been a while.

 

As I've already said in this thread, "GT200" might be a new codename relating to the revised design we'll see this summer/autumn, replacing the codename for the GPU that went AWOL last November.

For what it's worth, the "missing G100" codename isn't something I'm attaching any importance to. GT200 could be smoke'n'mirrors for G100. It could be something made up by a forumite in Taiwan. As far as I'm concerned it's just a place-holder for the name of NVidia's first double-precision GPU.

What is quite clear, regardless of codename, is that NVidia's initial design for a double-precision GPU has been significantly revised (functionality, including selected double-precision functions, has been dropped). Whether that re-design caused the delay or is an opportunity arising from the delay we dunno.

Jawed

 

GT200 would be cheaper to produce then the GX2 without a doubt (2PCBs and 2 rather large dies can't be too too cheap after all) and it would perform better and could be priced higher. Now why wouldn't NV want to release it again? They would just increase their margins while dominating the competition to an even greater degree. They could price the GT200 above the GX2 while they clear out old inventory and then adjust the price later down the road. There's just no reason not to do it. IHV's always release their parts the moment they are ready.