Nvidia GT300 core: Speculation
- Thread starter Shtal
- Start date
- Status
Freak'n Big Panda
Regular
NV will definitely have some DX11 arch ready in 2nd half '09. As to whether it will be based on G80 design or not is up in the air. Although I would be surprised if it was. I expect NV will try to optimize more for perf/mm^2 for cost reasons.
Last edited by a moderator:
G80 already had better performance/watt/mm^2 than G71, despite the absolute die size increase.
So 3 years on from the release of the first DX10 architecture we are to expect the first DX11 architecture? Makes me wonder what the point of DX10 (let alone 10.1) was... Clearly developer uptake is practically non-existent, as evidenced by a lack of DX10/10.1 titles.
Freak'n Big Panda
Regular
DX9 took a long long time to gain a foothold too. After every computer has a DX10 card game devs will start to target DX10 when developing games. They just don't want to cut down their user base or invest money on features that the majority of their customers will not see.
Yeah I expect I similar change with NVs DX11 arch. It should be a fairly big change from the G80 derived design that is seen in GT200.
Yeah I expect I similar change with NVs DX11 arch. It should be a fairly big change from the G80 derived design that is seen in GT200.
Well, DirectX11 won't take off any faster than DX10 did. We'll be seeing DX10 titles long after DX11 launches, just as we saw DX9 titles long after DX10 launched.
I'm guessing you meant "definitely" - but I like this version better.
Freak'n Big Panda
Regular
LoL yes I did. Embarrassing typo! corrected.
With such a short time between API releases and a constantly shifting target (10.0->10.1), I'd imagine most devs will skip 10.x altogether.
Only if, going by the current rumours, certain DX10 hardware of today already is capable of to-be DX11 features/satisfies requirements. Developers will skip a generation if the next one's hardware has sufficient market share.
With the popularity of the 8800GT and with ATI and NV having their little price wars with the HD4 and GT2x0 the current generation could be an important inflection point in market share that slows/reduces the adoption of the next generation. If the rumour turns out to be false developers could be faced with market share picture composed mostly of DX10.x hardware even well after DX11 is introduced.
All good points, perhaps I'm going about this wrong. Installed-base does mean a lot to developers...
Which rumours? I never heard this before.
armchair_architect
Newcomer
Similar? I can't think of many changes of similar magnitude to going unified. I'd expect a more evolutionary change, along the lines of NV10->NV20, NV20->NV30, or NV30->NV40. All of those were still big, but not "change everything" big.
Freak'n Big Panda
Regular
Yeah very true, the jump to a unified arch is probably on a different scale. But I do think there is a possibly that NV will try to really rework their arch for a higher number of flops per mm^2. Maybe, maybe not. But I think there's a solid chance.
A few things might get in the way of this though.
If the smaller branch granularity ends up paying off with future applications they will not feel a need to switch. Also if multichip solutions end up requiring too much software support to be feasible they will not feel a need to switch. Or rather they will continue to address the high end with huge monolithic designs.
But the fact remains that RV770 does what G200 does with 1/2 the mm2 and I'm sure NV is feeling the heat because of it.
Only possible cards would be Radeon HD3/4k series, IMO. Or at least I haven't heard of anyone else having a tesselator (HD2k, yes, but apparently 3k/4k got improved tesselator over it or something? at least I remember reading something along those lines, and 2k doesn't support all of 10.1 either so it's out)
Well, DX9 (with all it's different incarnations) has been the longest lasting DX so far, 4 years or so, DX10 (i'd count 10.1 as part of it, similar to how the DX9 HLSL etc profile updates were) is lasting over 3 years, less than 4, but not too far anyway, it seems.
We're only a year-and-a-half into DX10's life-cycle and they've already announced its successor, and updated the API once. Seems too soon given how big of a change DX10 was.
Bit over, it was released in november 2006 after all, but anyway, announcing something now doesn't mean anything, heck, whats becoming Windows 7 was talked even before XP was released, Blackcomb project (or what's left of it anyway, things changed on the way too many times and Vista came along the way to stir the pot too) by the info I've gotten is what will be Vienna aka Windows 7, but it's still not out
Point being, announcing something is irrelevant, release date is relevant, hell, I can't even remember when OGL3.0 was originally announced
It will be sometime in 2010 when we'll get DX11, most likely.
Sure, to corporations. How many corporations buy OSes to play games?
The retail/OEM/SI release was end of January, 07. Which would place the release almost precisely a-year-and-a-half ago, but we're picking nits now
I realize DX11 isn't knocking on the doorstep or anything, it just feels weird to me to be talking about it so soon in such a formal fashion/venue.
Here's another rumor:
http://www.techpowerup.com/66660/Sp..._Big_Bang_II__from_NVIDIA_this_September.html
The "Big Bang II" (when was the "I" ?
), whatever it really is, may show up this September, according to "Mr. Blurry Slide" here.
http://www.techpowerup.com/66660/Sp..._Big_Bang_II__from_NVIDIA_this_September.html
The "Big Bang II" (when was the "I" ?
), whatever it really is, may show up this September, according to "Mr. Blurry Slide" here. - Status