When is an overclock not an overclock? When you try doing it with a nVidia 7800.
Apparently, raising the clock speed of the card only works sometimes. As the chart in the linked article shows, the overclock “takes” to any significant degree only at certain speeds
The article says that:
“This data seems to indicate that each increase in the driver does increase the speed of something insignificant at every step. When moving up to one of the plateaus, it seems that a multiplier for something more important (like the pixel shader hardware) gets bumped up to the next discrete level.”
This would seem to be a reasonable surmise. Perhaps some of the video units work off a number of multipliers of the PCI-E bus. That could easily be checked by increasing the PCI-E bus speed a little and seeing if it affects the “clumps.”
Clipping Overclockers’ Wings?
If nVidia cards really do have a multiplier built into the GPU, this would provide nVidia with the means to limit overclocking. Essentially, this would be the equivalent of a multiplier lock like that found in CPUs.
This would have two benefits for nVidia. First, it could prevent crazy overclocks. More importantly, it provides a way to limit the overclockability of cheaper cards. No more cutting pipelines or having to making different versions of CPUs with fewer pipelines. Just program the GPU with lower multipliers.
If this is the case, overclockers would then be forced to overclock the PCI-E bus, which in all likelihood would mean overclockers would work harder and get less out of it.
True, the GPU on the 7800 doesn’t seem to be terribly constrained at the moment, but it gives nVidia the option to implement this down the road when deemed appropriate. It will be interesting to see if the other 7800 cards will “use” this potential feature.
Even if they don’t, though, this would be a tool ready to be used.
Be the first to comment