Over the last few days me and a friend have been building his machine and going thru some GPU's etc and I've been pondering the differences in Core Clock speeds.
I had been using a PNY GTS 250 1gb with a 738 mHz core clock and have now upgraded to his old Palit 1gb Sonic GTX 460 which has a Core Clock of 700. He is also now using GTX 580's in SLI and the Core Clock I believe was about 712. Now, I also looked at the GTS 550 that's supposed to be out and it has a CC of 900. Why are the lower cards showing higher speeds and is this really effecting the abilities on the higher cards? I'm not sure I follow the need for the difference in speeds.
You cant just compare Core clocks like that its meaningless. Its what I would call an Apples to Oranges comparison.
You have there 3 totally different architectures from 2 different companies. Each builds GPU's in a totally different way. Even the basic building blocks of the GPU's SP's (Stream processors) for ATI/AMD and these days CC's (Cuda cores) are totally different.
You need to do lots of research/read reviews.and ask lots of questions on Forums before you can get even a basic understanding of how to best compare GPU's.
Techpowerup do decent reviews. Actually if you go to one of the main review threads you will get lots of links to reputable sites.
And this is just why I posted a question on the GPU FORUMS asking why there was such a drastic difference in the core clocks, particularly with the higher end cards being lower in value. Your response doesn't narrow that particular question down in any way..
Stock speeds are determined by the manufacturer to place the product within a certain niche within a marketing hierarchy. The architecture of a card determines it's general placement within the hierarchy, while the clock speed fine tunes that placement.
More expensive, higher performing cards generally have more graphics rendering power determined by the number of pipelines or shader cores. The speed of the graphics chip works within these architectural limitations.
So the chip architecture determines the general performance level, i.e. the GTX 580 has 512 rendering pipelines, while the GTX 460 has 336 rendering pipelines. No matter how high you clock the 460, it will not be enough to overcome the advantage in rendering pipelines/shader units. A GTX 460 at 772 mhz will still deliver much less performance than a GTX 580 at 772 mhz.
The reason the GTX 460 is clocked at 675 mhz, at reference settings, is to place it below the performance of the GTX 465 and GTX 470. However, we know that a GTX 460 can run very well at 850 mhz or more. By increasing the reference clock speed of the GTX 460, it would created unwanted competition with the GTX 460's more expensive siblings, and confuse the marketing hierarchy. Who would buy a GTX 470 if they knew a less expensive GTX 460 was just as capable. On the other hand, by arbitrarily lowering the reference speed of the GTX 460, Nvidia erred too much on the conservative side, which allowed the AMD 6850 and 6870 to score wins when benchmarked against the 460. In retrospect, I believe that Nvidia would have been wiser to shoot for a reference speed of around 750 mhz.
So basically, even tho the 250 might have a higher C.C. then the 460, say, it would be something like bus width, with the 250 pushing more info PER LANE then the 460, but the 460 has more lanes to push the info on, allowing the transfer of more data quicker?
And sorry bout my previous post if it came out sounding rude, was not my intention if it did, just had a blonde prego in my ear so I may have been a bit snappy lol..
ok, ill go back to carrying buckets of water analogy that someone here used before. Imagine a bucket of water is data to be processed. a man has a bucket that he can carry to another container "twice per minute" (thats your MHZ). Now you have Man2 who can do this only "once per minute" but his bucket can carry 3x as much water. So even though the the first man can carry more buckets per minute (more mhz) he is getting less work done than the guy who can only carry 1 bucket per minute (less mhz) as he can carry 3x more each cycle. So you see more MHZ doesnt necessarily mean more performance. Higher end GPU's can do more work per cycle.