NVIDIA Maxwell Speculation Thread

For what i understand this is only a chip like a GK 107 ( GM107-108 ) who will be annonced.. not a GM100 or 104 sadly.


Wynix said:
"booked up with orders from industry giants including"
It's not a full list, but Nvidia to be in that list and AMD to be excluded is not a good sign.
Click to expand...

The list look mostly done about Mobile devices, and low chips .. On the same same side of bad things, the article is using 3 different interviews coming from TSMC chairman and directors, and then mix it with " industry sources " ( that you can allready take as rumor ).
 
Last edited by a moderator:
Interesting. Very interesting.

HW accelerated coherence even without CPU vendor exposing the cc protocol. I wonder how that works.

Pre-maxwell they only offer coherence before kernel launch and finish. Will they offer coherence for an in flight kernel?
 
http://videocardz.com/48478/new-roadmap-confirms-mobile-geforce-800m-series-heading-february

SfOTv2u.png


GM104=GTX880M?
 
It seems that we have double confirmation on the GT 840M.

I don't think (just from the PDF) that there's any reason why at least some of them can't be Maxwell parts, although I wouldn't be surprised if they're all Kepler rebadges/refreshes. Maybe there will be mobile Maxwell parts in February and desktop Maxwell parts in March?

If the 880M uses a Maxwell chip, then the memory size suggests a 256-bit bus. It could be the supposed GK104 successor that boxleitnerb has claimed will show up around February (I don't think he's said anything about bus widths). Such a part could be around 770-780-level performance on the desktop, and given that the 780 Ti was recently released I doubt there will be a higher-performing Maxwell part until H2 2014.
 
iMacmatician said:
It seems that we have double confirmation on the GT 840M.

I don't think (just from the PDF) that there's any reason why at least some of them can't be Maxwell parts, although I wouldn't be surprised if they're all Kepler rebadges/refreshes. Maybe there will be mobile Maxwell parts in February and desktop Maxwell parts in March?

If the 880M uses a Maxwell chip, then the memory size suggests a 256-bit bus. It could be the supposed GK104 successor that boxleitnerb has claimed will show up around February (I don't think he's said anything about bus widths). Such a part could be around 770-780-level performance on the desktop, and given that the 780 Ti was recently released I doubt there will be a higher-performing Maxwell part until H2 2014.
Click to expand...

I really doubt about the +8Gb ... if this mean the card have 8gb of GDDR dedicated and not in SLI....
 
lanek said:
I really doubt about the +8Gb ... if this mean the card have 8gb of GDDR dedicated and not in SLI....
Click to expand...

GTX 680M and 780M had 16 memory chips. With new 4Gbit devices, they are able to do 8GiB.

btw.
NVIDIA_DEV.1198 = "NVIDIA N15E-GX-A2"
NVIDIA_DEV.1199 = "NVIDIA N15E-GT-A2"
NVIDIA_DEV.119A = "NVIDIA N15P-GX-B-A2"
NVIDIA_DEV.119E = "NVIDIA GeForce GTX 780M"
Click to expand...
http://forums.laptopvideo2go.com/topic/30350-inf-v4132/
So this probably a high amount of GK104 rebrands. But GTX 860M (? 1152SPs@700MHz + 4GiB@192-Bit ?) could be a bigger increase over 760M (768SPs, 128-Bit).
 
UniversalTruth said:
TSMC only promises bright future with even 16 nm chips by the end of 2014 but where are the 20 nm? When?
Click to expand...


TSMC's 16 is similar to their 20, just with upgraded transistors. It's more like the difference between 28HPL and 28HPM than it is like a new node. It'll be interesting to see which products ship on 20 versus which products skip 20 and go straight to 16.
 
AnarchX said:
GTX 680M and 780M had 16 memory chips. With new 4Gbit devices, they are able to do 8GiB.
Click to expand...
You are quite right that it's technically possible. I don't see much point though in adding more memory just for bragging rights especially since this makes the product slightly worse (due to slightly higher power consumption).
 
mczak said:
You are quite right that it's technically possible. I don't see much point though in adding more memory just for bragging rights especially since this makes the product slightly worse (due to slightly higher power consumption).
Click to expand...

Like that would have stopped companies making those cards ever before?
512MB GF FX 5200
2x 512MB dual GF 6200
1GB DDR2 HD 4650
2GB DDR2 9600 GT
and the list could go on and on and on
 
Kaotik said:
Like that would have stopped companies making those cards ever before?
512MB GF FX 5200
2x 512MB dual GF 6200
1GB DDR2 HD 4650
2GB DDR2 9600 GT
and the list could go on and on and on
Click to expand...
Note that on the desktop at least the minimally increased power consumption doesn't matter at all. Plus, you didn't actually get higher power consumption neither really as these solutions didn't need more memory chips to achieve this. gddr5 clamshell configuration though has some higher power draw (nowhere near twice as high but still). And if they used ddr3 instead of gddr5 (or some similar change with older memory) it actually made the cards cheaper if they had more memory. That is certainly not the case if you just use twice the gddr5 memory.
Though you're quite right they might be willing to push this purely for marketing reasons.
 
mczak said:
Note that on the desktop at least the minimally increased power consumption doesn't matter at all. Plus, you didn't actually get higher power consumption neither really as these solutions didn't need more memory chips to achieve this. gddr5 clamshell configuration though has some higher power draw (nowhere near twice as high but still). And if they used ddr3 instead of gddr5 (or some similar change with older memory) it actually made the cards cheaper if they had more memory. That is certainly not the case if you just use twice the gddr5 memory.
Though you're quite right they might be willing to push this purely for marketing reasons.
Click to expand...

Wasn't there a GDDR3L spec going around awhile ago that was suppose to be like 90-95% of the performance but 70% of the power?
Edit- Aha! Found it. http://www.micron.com/~/media/Docum...eet/DRAM/DDR3/ddr3l_4Gb_graphics_addendum.pdf
 
Last edited by a moderator:
Esrever said:
AMD isn't making a 20nm ARM chip before those listed. GPUs [probably have nothing to do with that list.
Click to expand...

This original quote:

Industry sources said TSMC's 20nm production capacity has been booked up with orders from industry giants including Apple Inc., Qualcomm Inc., Xilinx, Altera, Supermicro, NVIDIA, MediaTek and Broadcom Corp.

http://focustaiwan.tw/news/atod/201312050035.aspx
Click to expand...
makes no mention that the "20nm production capacity" is only for ARM SOCs. Large die sizes such as GPUs and FPGAs will take up a lot of that 20nm production capacity and AMD is not mentioned as one of those who procured early 20nm production capacity from TSMC.
 
You must log in or register to reply here.

Similar threads

T
GPU Stability testing for Nvidia cards - PowerSweep
Replies
0
Views
451
T2098
T
D
Speculation and Rumors: Nvidia Blackwell ...
45 46 47
Replies
928
Views
104K
Kaotik
Kaotik
Rys
  • Sticky
Hello again
3 4 5
Replies
91
Views
12K
homerdog
homerdog
M
AMD made a mistake letting NVIDIA drive Ray Tracing APIs
5 6 7
Replies
122
Views
20K
MfA
M
xpea
Nvidia Hopper Speculation, Rumours and Discussion
7 8 9
Replies
177
Views
38K
Deleted member 2197
D
Back
Top