55-nm vs 65-nm GTX 260

To test the impact the smaller 55-nm manufacturing process has on power and cooling, we booted up all three GTX 260 GPU variants we’ve received over the last 10 months: our original 192-shader GeForce GTX card is an NVIDIA reference board we received waaaay back in late May of last year (read: it’s a very, very early engineering sample), while the 216-shader card represented include cards from BFG, XFX, and Zotac (we averaged the readings from the three cards), and finally, the 55-nm BFG and EVGA cards we’ve included in this article, whose readings were also averaged. Let’s look at the results shall we?

Power consumption

Notes

No surprises here. As expected the 55-nm GTX 200b delivered the lowest power consumption of all the GTX 260 cards we tested, running an impressive 16W lower at idle and 5W lower at load than the 65-nm 216-shader GeForce GTX 260. Interestingly enough, the 192-shader reference board consumed the most juice. We think this is largely due to the early nature of its design. As time passed and NVIDIA’s GPU manufacturing line matured, TSMC and NVIDIA likely found ways to reduce the power consumption of the chip, which is reflected in our benchmark results: the original GTX 260 card we received was a reference board produced months before the 216-shader cards we received back in the fall of last year.

Temps

Notes

Thanks to its lower power consumption, the 55-nm GeForce GTX 260 managed to run three degrees cooler than the 65-nm cards did in our testing, however under load the 55-nm boards ran about two degrees warmer on average than the 65-nm cards did. Surprised? So were we.

Our best guess is that the aluminum backplate NVIDIA uses on the 65-nm GTX 260 may be playing a role here. You see, NVIDIA’s original GTX 260 reference board design called for a completely enclosed cooling solution. The entire underside of the GTX 260’s PCB was cooled by an aluminum plate.

For their 55-nm GT200b GPUs however, NVIDIA has elected to remove this aluminum plate, most likely in an effort to help reduce production costs of the card. Instead, the PCB on the bottom of the card is completely exposed to air. This provides better ventilation to the PCB itself, but this leaves one critical hotspot on the card completely uncooled: the area directly underneath the GPU.

55-nm GT200b top, 65-nm GT200 bottom, note the missing backplate on the 55-nm card

As anyone with a high-end GPU can tell you, the area directly underneath the GPU can get pretty toasty, especially after extended gaming sessions. This is why some users have elected to attach heatsinks to the bottom of their cards. This helps disperse heat off this area, resulting in cooler GPU temps.

Since the 55-nm GTX 260 no longer has the aluminum backplate underneath the GPU, our theory is this is preventing the 55-nm GPU from running as cool as its 65-nm cousin does under load.

The 192-shader GTX 260 delivered higher temps than the other GTX 260 cards, but this shouldn’t be too surprising considering the power results we saw above. Again, this is probably due to the early nature of the card we received. NVIDIA and TSMC have had months to refine the manufacturing process of the GTX 200 GPU since the card was originally launched last summer.

Another item missing from 55-nm cards are covers for the SLI connector and audio connector. This is another cost-cutting move by NVIDIA

Before we move on to the next page, one thing we should note about the 55-nm GTX 200b cards we tested is how quickly they can cool down from peak temps. While it would take ten minutes or more for the 65-nm GPUs to cool down to their lowest idle temps, the 55-nm cards could get there in half the amount of time. This may or may not be important to you, depending on how you use your computer, but nevertheless, it’s an interesting observation we noted while watching the GPU temps rise and fall over time.