How come nobody does clock for clock comparos between card generations?

skadebo

[H]ard|Gawd
Joined
Feb 27, 2005
Messages
1,126
I'm just curious to see how much better new gen architecture is compared to old architecture to see what kind of performance has been gained that is not attributable to plain increases in memory speed or frequency increases due to die-shrinks of the GPU.

I know they may not be directly comparable, because Nvidia GPUs use a separate shader clock, but that can be assumed to be part of the new architecture.
 
Because it doesn't show anything. Example:

7900GTX Core Clock: 650mhz
8800GTX Core Clock: 575mhz

Core clock speed decreased, performance went through the roof.
 
Core clock apples to apples comparisons mean nothing when the architecture is completely different.
 
While true, I think he's just a bit curious. Such comparisons have been done before, though not between different generations, and they're interesting if not really valuable.

I could see doing this kind of comparison with, say, the 2900 XT and the 1900 XTX, but not between the 7900 GTX and the 8800 GTX. The discoupled shader clock would make any kind of comparison somewhat meaningless (and would you scale it if you underclocked the core?).
 
GPU's aren't like CPU's. You can't really compare them like one CPU core to another CPU core.
 
GPU's aren't like CPU's. You can't really compare them like one CPU core to another CPU core.


ahh its the same for cpus too :p

Core 2 Vs Pentium D? ---> HUGE increase, a 2.4ghz Core 2 outperforms a 3.6ghz p4
AMD X2 Vs pentium D? same, AMD x2 3800 @ 2ghz out performs a 3ghz pentium d
Core 2 Vs Amd X2? also core 2 has a 400mhz advantage clock for clock(3ghz AMD X2 ~ 2.6Ghz core 2)

but it serves no purpose with vcards because vcards change so often where as cpu archectures stay a bit longer =)
Throughout the life of the pentium 4, we saw how many diff gen of vcards? Geforce 3, 4, FX, 6, and 7, all during netburst. also 2 amd archectures too, AMD Athlon and Athlon 64 during that time span. ^^
 
Benchmarks are the closest you'll get to a clock to clock comparison. Its X amount of work in Y amount of time. That is how you can at least get a feel for the architecture.
 
I could see doing a comparison of a stock last gen card compared to a stock this-gen card, to see how far we've come in terms of performance since the last gen. But to try and clock the cards the same won't really prove anything since the architectures are massively different between fixed-pipeline and unified architecture GPUs.
 
Back
Top