This is still a persistent problem with me on my Dell E6510 with i5 processor. 
Unlike some, I still had the problem after upgrading to natty. 
In an effort to fix it I upgraded to a version of the 2.6.39.3 kernel 
compiled with corei7 as native which some people said fixed the problem. 
It didn't work for me. 

OK - I reason. If the problem is modules loading out of order, then lets get rid of the whole module loading issue completely.
So I compiled a 2.6.39.3 kernel with all components involved actually compiled into the kernel and not merely compiled as modules. The results are ...  interesting.

$ dmesg | grep intel
[    3.321097] intel_idle: MWAIT substates: 0x1120
[    3.321099] intel_idle: v0.4 model 0x25
[    3.321100] intel_idle: lapic_timer_reliable_states 0xffffffff
[    3.546876] ACPI: acpi_idle yielding to intel_idle
[    3.994957] [drm:i915_init] *ERROR* drm/i915 can't work without intel_agp module!
[   14.078987] intel ips 0000:00:1f.6: CPU TDP doesn't match expected value (found 25, expected 29)
[   14.079002] intel ips 0000:00:1f.6: PCI INT C -> GSI 18 (level, low) -> IRQ 18
[   14.083154] intel ips 0000:00:1f.6: IPS driver initialized, MCP temp limit 90
[   20.185256] hda_intel: Disable MSI for Nvidia chipset
[   24.284019] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   29.284544] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   34.285007] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   39.285459] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   44.287802] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   48.411798] hda-intel: IRQ timing workaround is activated for card #1. Suggest a bigger bdl_pos_adj.
[   49.290758] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   54.293801] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   59.306754] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   64.309786] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   69.312729] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   74.315683] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   79.318700] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   84.321632] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   89.324653] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   94.327672] intel ips 0000:00:1f.6: failed to disable graphis turbo
[   99.330620] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  104.333595] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  109.336594] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  114.339546] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  119.342546] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  124.345544] intel ips 0000:00:1f.6: failed to disable graphis turbo
[  129.348552] intel ips 0000:00:1f.6: failed to disable graphis turbo

and it continues to generate messages of this nature forever. It seems to me therefore that this problem is a little deeper than just a problem of modules loading out of order. Even when you compile everything involved into the kernel itself, it still won't work.  The intel_agp module is compiled and therefore present in memory. The i915 module is compiled and therefore present in memory. And yet they STILL can't find each other.

Are we perhaps looking at something as simple as a spelling error which stops i915 from recognising intel_agp because its name is misspelled? After all if the person who coded this can manage to misspell "graphis turbo" perhaps they also misspelled more important things. 

Anyway I hope someone finds the information valuable.
Now I've got to go and roll back my kernel ...