nV goes for high frequency with 256 shaders for GF104

MarcVenice

Moderator Emeritus <br>
Apr 2, 2007
5,664
0
0
Wasn't all of this allready known?

What they probably mean, is that the CUDA transistors that are now pretty much dedicated to double precision, are now also able to aid in rendering graphics. I'd take this with a grain of salt though...
 

jones377

Senior member
May 2, 2004
460
64
91
Wasn't all of this allready known?

What they probably mean, is that the CUDA transistors that are now pretty much dedicated to double precision, are now also able to aid in rendering graphics. I'd take this with a grain of salt though...

More likely they streamlined the CUDA cores to use less transistors by not supporting high performance double precision the way GF100 does. Didn't ATI do the same with their 57xx line?
 

ViRGE

Elite Member, Moderator Emeritus
Oct 9, 1999
31,516
167
106
L2 cache is tied to the ROPs in Fermi's overall design. So if it has fewer ROPs (likely) then it will have less L2 cache. This gives us the following two possibilities:

32 ROPs: 256bit memory bus, 512KB L2
24 ROPs: 192bit memory bus, 384KB L2
 

Paratus

Lifer
Jun 4, 2004
17,522
15,566
146
Having not read the article, I predict it will be priced between the two closest performing ATI cards. It will be biased though closer to the higher performing card and everyoje will complain it needs to be $20 - $40 cheaper.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Having not read the article, I predict it will be priced between the two closest performing ATI cards. It will be biased though closer to the higher performing card and everyoje will complain it needs to be $20 - $40 cheaper.

Yes but will this be true because of price fixing?
 

happy medium

Lifer
Jun 8, 2003
14,387
480
126

With 256 shader's it's a gtx 480 with the full 512 shaders cut in half.
Half of a 512 sp gtx 480 is 5830/5850 performance?

They must have cut more off this chip somewhere else or the gtx 465 preview was bogus.

The last rumers I heard had the gtx 465 at about 90% of a 5850 performance @ 279$ and this gtx 460 equalling a 5830 for about 220$, gts 450/440 slightly faster then the 5770/5750 for a little more money.

It just fits in the plan, fill the gaps with price/performance that matches.
 

Qbah

Diamond Member
Oct 18, 2005
3,754
10
81
This reminds me soo much of the last two times - remember those "optimized" shaders for G92b or GT200? One of the reasons back then nVidia changed the 8800GTS 512 to 9800GTX was that they said "we optimized the shaders" - and all they did was up the clocks... 8800GTS 512s were OCing sky high anyway...

Same for the GTX2xx family - they also said those shaders are somehow better - yeah, well... we all know how that turned out. A 9800GX2 with 256 shaders was faster than a 240 shader GTX280 D:

I'd say they're trying the same line again...
 

SHAQ

Senior member
Aug 5, 2002
738
0
76
Only 768mb of memory. This is the mainstream version. No chance of it touching a 470 much less a 480.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Uhmm... That is not going to happen.

What about a dual GPU card with 3.0 GB total (1.5 GB on each bus)?

As far as "input lag" goes maybe Nvidia could eventually develop some type of SFR to go along with this?
 

Skurge

Diamond Member
Aug 17, 2009
5,195
1
71
according to that article it has 86GB/s of bandwidth, Its gonna have a hard time competing with a 5830 at 1920x1080, which is pretty common now.

Since nV's GDDR5 controller isn't as good as ATI's I hate to see how much bandwidth the 450 and 440 are gonna have. I'm sure they will use 128bit and 96bit buses respectively.