Printable View
The other night, I was talking to my fellow sharky about my particularly annoying geforce 3 ti200. It seems, when I overclock the core/mem HIGHER than 238/538 scores get LOWER. Obviously, this makes very little sense. If any of you have any ideas as to why, please post, because this simply confuses the heck out of me!
What drivers are you using chipset/vid card? And, what mobo are you using?
Beyond a certain point you may get errors in the processing of information. These errors may cause information to require processing more than once, thus reducing efficiency.
If you are basing this only on 3DMark scores, I wouldn't. I would bench your programs and see what they say. Unless your actual game programs reduce in score, then you have nothing to worry about.
It may also have a lot to do with the increasing latencies as the memory frequencies climb.
But, I have noticed that trend. It was most notable in the original Radeon series of cards. They maxed out near the 200/200 mark, and anything over, even 204/204 would produce lower scores, by a significant margin somtimes.
There just seems to be that "sweet spot" in a chip where it just likes to run. Anything more/less and it doesn't perform as well as it could.
I got this on mine when I overclocked my Geforce 3, I actually started to get lower scores. After some research I found out it has something to do with the extra heat being generated that actually slows the chip down, or something like that. In any case it's safe to assume that your card is under too much stress anyway and should be clocked lower.
See guys, all those theories I have heard! There has to be a more scientific reason for lowering speeds. The "Sweet spot" theory Is not what I want to hear. Anyone REALLY know? Un4given's seems to make sense, BUT if there were errors, they wouldnt be corrected, they would be shown onscreen! There is (to my knowledge) no error correction in video cards. As for mellojoes response, I dont know how that could be revelant. Look at PC memory, if you overclock it, it doesnt hit a "sweet spot" and then slow down, it keeps going faster until it hits a wall and wont OC any farther. Here are my specs
epox 4g4a+
2.26 p4
256mb sammy ctl
visiontek ti200 with crystal orb
etc etc etc
with todays technology we dont have time to correct the last mistake they dont include error correction of any kind on todays video cards... they would just move on to the next frame and most likely overlook the last error hence "artifacts" that we see when we oc too high, you might be getting to the point right before it starts arifacting and it is just generating too much heat (like Sunday Ironfoot said) to run at top efficientcy.Quote:
Originally posted by OCking
See guys, all those theories I have heard! There has to be a more scientific reason for lowering speeds. The "Sweet spot" theory Is not what I want to hear. Anyone REALLY know? Un4given's seems to make sense, BUT if there were errors, they wouldnt be corrected, they would be shown onscreen! There is (to my knowledge) no error correction in video cards. As for mellojoes response, I dont know how that could be revelant. Look at PC memory, if you overclock it, it doesnt hit a "sweet spot" and then slow down, it keeps going faster until it hits a wall and wont OC any farther. Here are my specs
epox 4g4a+
2.26 p4
256mb sammy ctl
visiontek ti200 with crystal orb
etc etc etc
Maybe the signaling paths where data travels through was never designed to teolerate such speeds. Depends on how they build them I guess. I think I'm totally off.