   |
|||
|
|||
|
:
News
: Archives
|
|
NVIDIA GeForce 8800 GT |
|
|
Revitalizing the Midrange |
|||
|
|
|||
|
by Josh Walrath |
|||
|
The standard 8800 GT is clocked at a 600 MHz core, 1500 MHz shader, and 900 MHz memory. Even with only 16 ROPS, the fillrate of the card is pretty hefty. The memory connection is moving right along at 57.6 GB/sec of bandwidth, which is slightly lower than the older 8800 GTS’s 64 GB/sec. NVIDIA has supposedly optimized the architecture to more adequately utilize available bandwidth, and the doubling of the Texture Address units is probably the biggest contributor towards efficiency. 112 steam processors running at 1500 MHz gives the card a lot of math ability. The 8800 Ultra had its shader units running at 1500 MHz also, but it featured the full 128 units. 512 MB of GDDR-3 are included in the board, which should take care of most current and upcoming titles. This should run everything at 1600x1200 with 4X AA enabled, but the extra bandwidth and memory sizes of the 8800 GTX and Ultra make gaming at 1920x1200 and higher with AA enabled a lot more comfortable.
The back of the card is pretty generic and sparse. Shocking. Due to the transition to 65 nm, the power requirements of the G92 are much lower than the older G80. Not only are the power requirements lower, but the thermal properties are much improved. The reference 8800 GT board only requires one PCI-E power connection, as well as a single slot cooling device. In shader heavy applications, the 8800 GT should get pretty close to the 8800 GTX in overall performance. Once AA is enabled at higher resolutions, the advantages the 8800 GTX has over the 8800 GT should show up really well. There is one application that I tested in particular which really shows the differences between the two cards.
Card Impressions After a year of handling 8800 GTS and GTX cards, this is certainly a much different experience. It is significantly smaller and lighter than either of the earlier cards. The single slot cooling is obviously the biggest difference, but this card can fit in much tighter places than its predecessors. The reference fan is quite small, but it is able to move adequate air to keep the card cooled within specifications. It should not be confused with a cool running card though. The fan stays very quiet, even in heavy gaming. Once the fan hits 50% speed, then it starts to be audible. Only in heavy testing was I able to hear it go up over 60% or so, but never up to 100% speed. The card runs quite hot. The GPU can go as high as 92C according to nTune at full load, and typically was around 60C at idle. After a round of testing, I attempted to uninstall the card from the bench, and it was literally too hot for me to handle. I had to wait about 5 minutes before I could comfortably hold the card. If the card had still been running, it would have likely cooled down faster due to the fan. The card I have is a reference board direct from NVIDIA, but the first major run of the product is always done directly for NVIDIA, and they then sell the completed cards to their AIB partners. Likely most of the AIBs will continue to use this design, while a few others will utilize the Coolermaster based cooler which features a larger fan but rotates slower. The AIBs typically slap on their own stickers over the cover. In the next month we should see more customized designs from the larger AIBs like Gigabyte and Asus which will feature their own PCB design and cooling solutions.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2007 PenStar Systems, LLC. |
|||
