Articles: Graphics
 

Bookmark and Share

(5) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 ]

As far as we know, the GeForce 7800/7900 chips feature two MIO interfaces with the bandwidth of 1GB/s each. However, one of them is only enabled when we have a quad SLI system.

The top PCB of the GeForce 7950 GX2 looks emptier. The power circuit here is simpler than on the bottom PCB and includes fewer elements, although it is this PCB that receives external power (it carries a standard 6-pin PCI Express power connector for that). The difference may probably be due to the lack of necessity to power the PCI Express bridge chip.

Another difference from the GeForce 7900 GX2 is that the top PCB now carries a TV output and DVI-I connectors for monitors.

A curious fact, this graphics processor is dated the 13th week of the year and so belongs to another batch than the bottom one. This doesn’t matter much, of course, since the chips have the same revision and are absolutely identical. Both the GPUs of the GeForce 7950 GX2 card are clocked at 500MHz and this frequency is the same for all the subunits (pixel and vertex processors, ROPs and TMUs). So the GeForce 7950 GX2 doesn’t differ from the GeForce 7900 GX2 in this respect.

Considering that it’s impossible to mount a massive (i.e. efficient) cooler on this two-storied card, it’s reasonable to keep the GPU clock rate low, even though the G71 is not a very hot chip by itself. Nvidia puts its stake on the combined power of the two processors rather than on their frequency. We’ll see soon if Nvidia’s right here.

According to Nvidia, the peak power consumption of the GeForce 7950 GX2 is 143W, but this number looks a bit exaggerated. As we know, the standard GeForce 7900 GT consumes about 48W under load. The GeForce 7950 GX2’s GPUs work at a 50MHz higher frequency, but its memory frequency is lower compared with the GeForce 7900 GT’s (600MHz against 660MHz). The new card has more memory chips, so we suppose it consumes about 110-120W under load, comparable with the consumption of the power-hungry ATI Radeon X1900 XTX.

On the reverse side of the PCB there is only the other part of the connector that connects the two PCBs of the GeForce 7950 GX2 and a few power elements, including an Intersil ISL6560 controller which controls the GPU power supply. The memory power supply is managed by an Intersil ISL6549 chip. The GeForce 7950 GX2’s PCBs have in fact identical power circuits: the +12V voltage coming to the external connector of the top PCB is transferred to the bottom one via dedicated pins of the connecting strip.

The left part of this strip transfers power (wide contacts) as well as MIO interface signals (ordinary contacts). The right part is dedicated to PCI Express contacts. Note that the GeForce 7950 GX2 uses only one such connecting strip whereas the GeForce 7900 GX2 had two (because the GeForce 7900 GX2 was the basis of the Nvidia Quadro FX 5400 X2 model that supported four DVI outputs).

Unlike on GeForce 7900 GTX or GeForce 7900 GX2, the memory chips are placed in the shape of the letter L here. This helped make the PCB smaller but the wiring is somewhat more intricate due to the necessity to keep the lines between the GPU and memory the same size. You can see this clearly in the photo.

Nvidia’s new solution employs Samsung K4J52324QC-BC14 memory that works at 1.8V voltage and has a rated frequency of 700 (1400) MHz. Each PCB of a GeForce 7950 GX2 is equipped with eight GDDR3 chips, so the total amount of memory is 1GB (16 x 512Mb). However, you should be aware that 3D applications will only have 512 megabytes of graphics memory at their disposal even if you’ve configured a quad SLI subsystem with two GeForce 7950 GX2 cards. So, it’s not exactly correct to say that these graphics subsystem offer 1 or 2GB of graphics memory.

The GeForce 7950 GX2 clocks its memory at the same frequency as the GeForce 7900 GX2 does: 600 (1200) MHz. Such a low clock-rate means low memory bandwidth, and this may mean lower performance if you are using high display resolutions, especially with full-screen antialiasing. The low memory frequency was among the reasons why the four-processor Nvidia quad-SLI configuration did poorly in extreme graphics modes in our preview (for details see our article called Quadtet: Nvidia GeForce 7900 Quad SLI Performance Unveiled).

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 ]

Discussion

Comments currently: 5
Discussion started: 08/28/06 06:06:05 AM
Latest comment: 08/13/07 08:50:18 PM

View comments

Add your Comment