NVIDIA GeForce GTX 460 - GF104 breaks cover
Monday 12th July 2010, 10:23:00 PM, written by Rys
NVIDIA's latest DX11 hardware hasn't had an easy life since it was born back in March. Roundly criticised for being late, hot, hard to find, power hungry and expensive, GF100-powered GeForces continue to struggle to overcome those downsides with their obvious performance. Those enthusiasts and fans won over by the potential framerates and pixel quality available on a modern high-end GeForce will have enjoyed those facets of their purchase, but it seems the rest of the world has yet to be won over.
The Californian graphics behemoth have finally done something about it, and it appears they've done it in style. It doesn't hit the heady heights of performance available in the highest end GeForce SKUs, but it gets close, and does so without getting too hot, noisy or hard on your electricity bill or your balance sheet. For $199 or $229 depending on the variant you choose, NVIDIA have placed their new GeForce GTX 460s right where ATI probably didn't want them, sniffing around Radeon 5830 and 5850 and spoiling for a fight. Their own GTX 465 and GTX 470 products should be looking over their shoulders too.
GF104 powers the new hardware and brings to the table a refreshed shader multiprocessor architecture and scheduler, marrying a minor abundance of those with cut-down amounts of the supporting logic you'll find in a GF100, to create a scaled piece of silicon that seems to hit the spot. It's not that small -- 1.95B transistors will see to that, even fabbed at 40nm -- but it packs a punch.
Headline figures of 384 ALUs, 8 quad ROPs, 64 filtered texture samples per clock, a peak triangle rate of 2 per clock and 256-bit memory interface are all part of the GF104 experience, putting it within striking distance of cut down GF100 products.
The Tech Report have a full, flowing analysis, and Scott picks out the highlights of the architecture, marrying those details with his take on how it affects performance in the real-world.
The executive summary is a pair of new, attractive GeForce DX11 products that warrant a much bigger chance at your hard-earned than GeForces 465, 470 and 480 ever have done.
The Californian graphics behemoth have finally done something about it, and it appears they've done it in style. It doesn't hit the heady heights of performance available in the highest end GeForce SKUs, but it gets close, and does so without getting too hot, noisy or hard on your electricity bill or your balance sheet. For $199 or $229 depending on the variant you choose, NVIDIA have placed their new GeForce GTX 460s right where ATI probably didn't want them, sniffing around Radeon 5830 and 5850 and spoiling for a fight. Their own GTX 465 and GTX 470 products should be looking over their shoulders too.
GF104 powers the new hardware and brings to the table a refreshed shader multiprocessor architecture and scheduler, marrying a minor abundance of those with cut-down amounts of the supporting logic you'll find in a GF100, to create a scaled piece of silicon that seems to hit the spot. It's not that small -- 1.95B transistors will see to that, even fabbed at 40nm -- but it packs a punch.
Headline figures of 384 ALUs, 8 quad ROPs, 64 filtered texture samples per clock, a peak triangle rate of 2 per clock and 256-bit memory interface are all part of the GF104 experience, putting it within striking distance of cut down GF100 products.
The Tech Report have a full, flowing analysis, and Scott picks out the highlights of the architecture, marrying those details with his take on how it affects performance in the real-world.
The executive summary is a pair of new, attractive GeForce DX11 products that warrant a much bigger chance at your hard-earned than GeForces 465, 470 and 480 ever have done.
Tagging
nvidia ± tech, report, gf104, geforce, 460, fermi, 199
Latest Thread Comments (244 total)
Add your comment in the forums
Posted by trinibwoy on Wednesday, 01-Sep-10 12:22:58 UTC 
Posted by Kaotik on Wednesday, 01-Sep-10 12:52:22 UTC
Posted by Dave Baumann on Wednesday, 01-Sep-10 16:15:14 UTC
Posted by Babel-17 on Wednesday, 01-Sep-10 23:41:57 UTC
Posted by onethreehill on Thursday, 02-Sep-10 04:17:07 UTC
Posted by Alexko on Friday, 03-Sep-10 14:53:30 UTC
Posted by Erinyes on Friday, 03-Sep-10 18:15:05 UTC
Posted by swaaye on Friday, 03-Sep-10 20:34:07 UTC
Posted by Babel-17 on Friday, 03-Sep-10 22:07:19 UTC
Posted by onethreehill on Saturday, 04-Sep-10 00:32:18 UTC
Add your comment in the forums
Related nvidia News
NVIDIA's Parallel Nsight finally released
PhysX87, ancient tragedy in 5 acts by RWT
So long, Chris, and thanks for all the fish
NVIDIA GF100 graphics architecture details
NVIDIA Fermi: new GPU architecture, starting with GF100
NVIDIA release OpenCL GPU drivers for Linux and Windows
NVIDIA GeForce GTX 275 at $250 to fight HD 4890
A look at NVIDIA's SLI Multi-OS and new Quadros
Ahead Nero gets CUDA support for video encoding
G92b renamed again, this time for notebooks
PhysX87, ancient tragedy in 5 acts by RWT
So long, Chris, and thanks for all the fish
NVIDIA GF100 graphics architecture details
NVIDIA Fermi: new GPU architecture, starting with GF100
NVIDIA release OpenCL GPU drivers for Linux and Windows
NVIDIA GeForce GTX 275 at $250 to fight HD 4890
A look at NVIDIA's SLI Multi-OS and new Quadros
Ahead Nero gets CUDA support for video encoding
G92b renamed again, this time for notebooks
I think Dave is implying and others are chiming in with, is that disabling units in 5850 has less to do with yields and more to do with product differentiation.