There
are usually a certain percentage of parts during any manufacturing
process that don't perform up to specifications. In the semiconductor
industry, many of those parts will end up being speed-binned and set
aside for use in less stringent applications.
NVIDIA
started mass production of its GF100 GPU using their new Fermi
architecture on TSMC's 40nm
process in January. Yields weren't great, but they were good
enough for NVIDIA to ship their GeForce
GTX 480 and 470 graphics cards in April. Their primary
competitor, ATI, had been shipping their own DirectX 11 5870 cards
for over
six months.
Faced
with a surplus of chips and a lack of a mainstream performance
product, NVIDIA is now launching the GeForce GTX 465 for $279. The
primary difference on the graphics card is the use of a 256-bit
interface and 1024MB of GDDR5.
However,
the card itself uses the same cooling fan and PCB as the rest of the
GTX 400 series, which means that NVIDIA's cost still remains very
high. The die size of the GF100 chip is massive, measuring 23mm x
23mm for a total of 529 square millimeters. By comparison,
theCypress chips used in the Radeon 5800 and 5900 series
cards come in at a more moderate 334mm², making the GF100 almost 60%
larger.
This is set to be a very strong launch for
NVIDIA, with over 100,000 units expected to be ready for retail from
add-in-board partners such as ASUS, EVGA, Galaxy, MSI, Palit,
PNY, Zotac and others.
The GF100 chip on the GTX 465 has the
same core and shader clock speeds as on the GTX 470, but a reduction
in CUDA cores similar to the GTX480M means that it has a much lower
number of stream processors, texture units, and ROPs. One of the four
Graphics Processing Clusters has been disabled, which means that
rasterization also takes a 25% hit.
Even
though a large part of the GF100 is disabled, power consumption and
heat generation aren't affected proportionately. The GTX 465 is rated
for a Thermal Design Power of 200 watts, just 15 watts less than the
GTX 470.
The specifications for the GTX 465 are similar to
the GTX
480M mobile GPU announced last weekwhich will first appear in the
middle of June. NVIDIA is trying to launch as many GF100 parts as
possible before ATI can ready its second generation DirectX 11 GPUs,
codenamed Southern
Islands.
ATI
had been planning a 32nm Northern
Islands series
of GPUs for launch this summer, but that is currently being
redesigned for the 28nm
process node due to problems at TSMC.
|
GeForce GTX 465
|
GeForce GTX 470
|
GeForce GTX 480
|
|
Graphics Processing Clusters |
3 |
4 |
4 |
|
Streaming Multiprocessors |
11 |
14 |
15 |
|
CUDA Cores |
352 |
448 |
480 |
|
Texture Units |
44 |
56 |
60 |
|
ROP Units |
32 |
40 |
48 |
|
Graphics Clock
(Fixed Function Units) |
607 MHz |
607 MHz |
700 MHz |
|
Processor Clock (CUDA Cores) |
1215 MHz |
1215 MHz |
1401 MHz |
|
Memory Clock
(Clock rate / Data rate) |
802 MHz / 3208 MHz |
837 MHz / 3348 MHz |
924 MHz / 3696 MHz |
|
Total Video Memory |
1024 MB |
1280 MB |
1536 MB |
|
Memory Interface |
256-bit |
320-bit |
384-bit |
|
Total Memory Bandwidth |
102.6 GB/s |
133.9 GB/s |
177.4 GB/s |
|
Texture Filtering Rate (Bilinear) |
26.7 GigaTexels/sec |
34.0 GigaTexels/sec |
42.0 GigaTexels/sec |
|
Fabrication Process |
40 nm |
40 nm |
40 nm |
|
Connectors |
2 x Dual-Link DVI-I
1 x Mini HDMI |
2 x Dual-Link DVI-I
1 x Mini HDMI |
2 x Dual-Link DVI-I
1 x Mini HDMI |
|
Form Factor |
Dual Slot |
Dual Slot |
Dual Slot |
|
Power Connectors |
2 x 6-pin |
2 x 6-pin |
1 x 6-pin, 1 x 8-pin |
|
Max Board Power (TDP) |
200 Watts |
215 Watts |
250 Watts |
|
Recommended PSU |
550 Watts |
550 Watts |
600 Watts |
|
GPU Thermal Threshold |
105° C |
105° C |
105° C |
|
MSRP |
$279 USD |
$349 USD |
$499 USD |
