I've a external 4Mhz Xtal.
here's my test:
#include <avr/io.h>

void delay_ms(unsigned short ms) {
unsigned short outer1, outer2;
outer1 = 200;
while (outer1) {
outer2 = 1000;
while (outer2) {
while ( ms ) { ms--; asm volatile("nop"); }
outer2--;
}
outer1--;
}
}

int main(void){
unsigned short duration=5000;
//SET OUTPUT PINS
DDRD |= (1 << PD6);
DDRD |= (1 << PD5);
DDRD |= (1 << PD4);

while (1) {
PORTD |= (1 << PD6);
delay_ms(duration);
PORTD &= ~(1 << PD6);
PORTD |= (1 << PD4);
delay_ms(duration);
PORTD &= ~(1 << PD4);
}
return 0;
}

And makefile:
bkw@snigel:~/tmp/code$ cat Makefile
MCU=atmega16
F_CPU=4000000
F_OSC=4000000
CC=avr-gcc
OBJCOPY=avr-objcopy
# optimize for size:
CFLAGS=-g -mmcu=$(MCU) -Wall -Wstrict-prototypes -Os -mcall-prologues
all: test.hex
test.hex : test.out
$(OBJCOPY) -R .eeprom -O ihex test.out test.hex
test.out : test.o
$(CC) $(CFLAGS) -o test.out -Wl,-Map,test.map test.o
test.o : test.c
$(CC) $(CFLAGS) -Os -c test.c
#erase the AVR before it is programmed
load: test.hex
clean:
rm -f *.o *.map *.out