Jaskanwar Singh
Aspiring Novelist
NVIDIA GeForce 600 Series: 28nm, OEM-only Fermi Refresh? - Bright Side Of News*
Rumor : NVIDIA GeForce 600 Series: 28nm, OEM-only Fermi Refresh?
- Thread starter Jaskanwar Singh
- Start date
-
- Tags
- geforce nvidia parts production series
coderunknown
Retired Forum Mod
or maybe we'll have something like:
GT 610/620/625 for OEM (old architecture + die shrink).
GT 630/640/650/.... for retail (new architecture + die shrink).
cause if you look into the OEM business, almost every company just offers a lowend crap in their machine. so for the selected highend offering, they may continue with 5-series cards or Radeon GPU till real 6-series shows its face.
GT 610/620/625 for OEM (old architecture + die shrink).
GT 630/640/650/.... for retail (new architecture + die shrink).
cause if you look into the OEM business, almost every company just offers a lowend crap in their machine. so for the selected highend offering, they may continue with 5-series cards or Radeon GPU till real 6-series shows its face.
JSingh: It makes sense, because fermi gen 1 was...lets face it...a disaster. Fresh tech on unproven manufacturing process. Even Fermi gen #2 was a semi-disappointing, a disappointing 590 and 550 Ti. Even the GT 545 had two flavours, with the OEM one being better than retail. The competing southern island chips were better.
So now they do not want to risk it again by giving Northern Islands more ammunition to shoot at Kepler.
In fact Kepler and Southern islands will have different manufacturing processes on the 28nm node.
So now they do not want to risk it again by giving Northern Islands more ammunition to shoot at Kepler.
In fact Kepler and Southern islands will have different manufacturing processes on the 28nm node.
Last edited:
J
Joker
Guest
historically...ati has been the company which likes to jump on new fabrication processes whereas nvidia focuses more on architecture. its a good decision by them to have a pipe cleaner...although if hd 7000/gcn turns out to be good...this move i.e. delay in kepler parts might backfire.
Actually after the ill-fated Geforce FX, nVIDIA is sort of shy to jump on new fabrication process. Generally AMD tests newer process with older generation cards (like the 4770), but this time they are jumping straight to newer generation. In any case, the delay of Kepler will give them breathing space for sure.
Can you explain the fermi anomaly ?(according to what u said, Nvidia had other chips in 40nm.GT 210,220 and 240 dont count, as they were low end and do not reflect the performance of the main architecture)
"Actually after the ill-fated Geforce FX, nVIDIA is sort of shy to jump on new fabrication process."
Isnt that what they did with fermi?
Isnt that what they did with fermi?
What I was referring was jumping prior to ATI/AMD, with reference to what Joker has said in post #4. In any case, newer fabrication process are just part of life, no company can bypass it forever, but nVIDIA is taking it slower as compared AMD. If you remember, HD6000 was supposed to come in 28nm, and it was sort of last-minute change from AMD's part.
6 series was coming in 32nm. TSMC scrapped it. so it came in 40nm.
AcceleratorX
Youngling
One of the announcements said that both 600 series and 700 series will be on 28nm.
I think Fermi cost NVIDIA a lot of money to develop and given how the GTX 465/470/480 were filled with leaking transistors causing heat buildup, I believe they wanted to recover some of the costs, which is why the 500 series was introduced.
The 550 Ti had a new mixed density memory configuration which allowed it's 192-bit bus to feature 1GB memory (it would otherwise have been 768MB or 1.25GB or something like that). I think NVIDIA probably wants to make good on this new mixed memory density controller, which they can easily utilize in various OEM products in the form of the 600 series.
Think about it - make a refresh of low end 500 series, bump up the clocks, disable or enable 64 extra bits of memory bus, use mixed density controller to achieve 1GB, 2GB, 3GB - whatever. This will allow them to adapt the GPU to the OEM's requirements. This is probably the only reason the GT600 series will be there for - to get the OEM win.
The real refresh will most likely occupy either the GTX 700 series, or will form the GTS and GTX lineups of the 600 series.
(The bad point: Potentially we will see PC vendors selling computers with "GeForce GT 640 3GB 64-bit", and likely this will fool quite a few people).
I think Fermi cost NVIDIA a lot of money to develop and given how the GTX 465/470/480 were filled with leaking transistors causing heat buildup, I believe they wanted to recover some of the costs, which is why the 500 series was introduced.
The 550 Ti had a new mixed density memory configuration which allowed it's 192-bit bus to feature 1GB memory (it would otherwise have been 768MB or 1.25GB or something like that). I think NVIDIA probably wants to make good on this new mixed memory density controller, which they can easily utilize in various OEM products in the form of the 600 series.
Think about it - make a refresh of low end 500 series, bump up the clocks, disable or enable 64 extra bits of memory bus, use mixed density controller to achieve 1GB, 2GB, 3GB - whatever. This will allow them to adapt the GPU to the OEM's requirements. This is probably the only reason the GT600 series will be there for - to get the OEM win.
The real refresh will most likely occupy either the GTX 700 series, or will form the GTS and GTX lineups of the 600 series.
(The bad point: Potentially we will see PC vendors selling computers with "GeForce GT 640 3GB 64-bit", and likely this will fool quite a few people).
comp@ddict
EXIT: DATA Junkyard
No high end Kepler anytime soon??
o.o.0.
o.o.0.