Nvidia GT200b rumours and speculation thread

Fudo thinks GT200b (55nm shrink of GT200) could be with us within a few weeks of R700, which would probably put it in September:

http://www.fudzilla.com/index.php?option=com_content&task=view&id=8391&Itemid=1

Anyone who bought a GTX280 at full price is going to be pretty p*ssed off if that's true. :devilish:

Its an interesting theory about probing the market to check out the competition and one I definatly wouldn't put past NV.

I'm expecting G200b to be quite an improvement over G200. Not so much in performance (although it will likely be faster, possibly at several price points) but definatly in cost.

Then again, it all depends how good R700 is. From the sounds of it it will outperfom any G200b when its working right. But how often does it work right and what will it cost in comparison.....
 
And hardspell via chiphell (en.hardspell.com/doc/showcont.asp?news_id=3710) seems to be convinced, that nvidia's skipping an already taped-out chip (one might think), to go straight for the Ub0r-1337 GT300 with 16 Clusters.

Increased clocks, more SP's and yet much lower power draw? Ok its on a 45nm process but given that they are switching back to a 256bit memory interface it makes you wonder what else they are cutting out.

Back to 16 ROPs? No DP support?
 
I can't see any way a 45 or 40nm card could be imminent - TSMC just isn't that far along yet.

I'm not sure about "probing the market" but there have been quite persistent rumours that GT200 was about six months behind schedule. GT200 would have been a very sensible product 6 months ago (RV770 was merely a twinkle in ATI's eye at that point) and now would have been a sensible time to be releasing a 55nm refresh of GT200 at more or less the same time as issuing a 55nm refresh of G92 (G92b).
 
NordicHardware also expect the GT200b to be launched in September. It's supposed to be faster than the R700.

Unless its running at 1Ghz/2Ghz Core/Shaders I don't see how it could. Thats based on R700 being 50% faster on average than the GTX280.

They do only speculate that this card is GT200b so perhaps the above source is correct and they are skipping it altogether and going to a next gen GT300.
 
At what cost though, I wonder? If NV are producing a chip with even more clusters using the current 55nm process it would be (even more of a) monster! We're not expecting TSMC's 40nm process for some time yet are we?
 
At what cost though, I wonder? If NV are producing a chip with even more clusters using the current 55nm process it would be (even more of a) monster! We're not expecting TSMC's 40nm process for some time yet are we?

Unless they cut back in other areas.....
 
TSMC's 40nm process won't be slated before late Q1 / early Q2 2009.
 
45nm, GDDR5, directx 10.1?

this all seems highly unlikely to me for any Nv part this year.

45nm sounds like a stretch but not the rest.

TBH this is exactly the kind of response I would expect from NV, especially when they have had so muchtime to sit back and prepare uber GPU's.

I mean, if this is GT300, then based on the kind of evolution we all expected from G80, this is probably about right for this time frame.
 
45nm sounds like a stretch but not the rest.
Do you think they'd go "hmm GT200 isnt that big (on 55nm), we could prolly add a few more in there.."

At what size would they entertain such thoughts?

NV/4045 ~ 300mm2
G70 ~ 305 mm2
G71 ~ 196mm2
G80 ~ 484mm2
G92 ~ 330mm2
GT200 ~ 576mm2
 
Do you think they'd go "hmm GT200 isnt that big (on 55nm), we could prolly add a few more in there.."

At what size would they entertain such thoughts?

[original design]NV/4045 ~ 300mm2
[refresh, additional units (MADD, processors)]G70 ~ 305 mm2
[Shrink, additional units (blenders)]G71 ~ 196mm2[]
[original design]G80 ~ 484mm2
[shrink, additional units (32 TAs), removed units (1/3 mem interface/ROPs]G92 ~ 330mm2
[refresh, additional units (well, you know)]GT200 ~ 576mm2[]
[my bold]
Now - does anyone see a pattern? I'm having a hard time to do so.
 
Do you think they'd go "hmm GT200 isnt that big (on 55nm), we could prolly add a few more in there.."

At what size would they entertain such thoughts?

NV/4045 ~ 300mm2
G70 ~ 305 mm2
G71 ~ 196mm2
G80 ~ 484mm2
G92 ~ 330mm2
GT200 ~ 576mm2

GDDR5 on a 256bit interface would reduce the size of the chip, not increase it. And I don't see why adding DX10.1 support would take up much space.

More shaders obviously would but as I said above, they miight remove other elements. Fewer ROPs for example, or the removal of the DP unit from the shaders.
 
I don't think so. If this rumour would be true:
  • ATI would remain performance king until Q109
  • Nvidia wouldn't be able to decrease productions costs if they skip GT200b

ATI would still be perf king with the release of the GT200b, although the margin will be closed. Not too much of a satisfactory move though.

The GT200b IMHO would be a stealth improvement, GTX260/280+ maybe.
 
Back
Top