I've been trying to decide between i5 and i7 and have just spent weeks looking at benchmark results comparing the two chipsets (X58 vs P55). From what I have seen, there is very little difference between the two. Granted some applications benefit slightly from hyperthreading but in many cases it seems you're talking about a few seconds over long renders.
If money is an object, I would not feel bad at all about going i5. And there is very little if any difference between pcie x16 and x8. I've just looked at dozens of different benches showing absolutely no difference at all, except at ridiculous resolutions like 3500x2200 type thing.
Obviously if you have the money, then go X58, but if you're not a major player and are looking for a very fast system, a P55 chipset with an i5 is damn fast.

I have found the differences are subtle too. I just learned my 1x raid card is transferring files at nearly 900mb per second peak, but then it setlles in at about 750-800mb per second. The drives are the bottleneck in my system, not the 1x card. So how much faster would an expensive controller card be on a 32 lane board? None. For the average person running a raid 0 configuration with two drives, etc., a 1 or 2x card would seem to be sufficient. If you are running a raid configuration with 10 drives in Raid 30 or whatever such as Harm, then you have no choice but to move up to the bigger card, in which case the 1366 would be a better choice.

Hey Harm,

Thanks for your input. On your recommendation, I decided to stay with the 1366 platform.

I ended up getting the GA-X58A-UD7 board with a core i7 920 D0 stepping chip, 6 gigs ram, Quadro FX 1800 video card, WD caviar black 1TB drive, Seasonic 750W power supply, Dell U2410 LCD screen.

When SSDs become affordable for human beings, I'll put one in as my system drive.

I've set up a dual boot system, XP Pro and Win 7 Pro 64 Bit. Avid will be on my XP partition until they qualify Win 7.

So far, my experience is that it's been touchy to set up, but now that it's pretty much good to go, it's great. Very fast.

I'm interested in overclocking, but have research to do. The bios is incredibly complex. The last time I built a computer, the bios was simple to understand. Boy, things have changed.

Although this motherboard is finicky, it's also bullet proof. If you do something wrong, it tries to boot a couple of times, then gives you a workable bios and tells you that you screwed up with your settings.

Thanks again. When I get a chance to do some encoding, I'll try to post some results.

Points 1 and 5 are just pure misperception. The P55 itself has eight PCI-e lanes, while the Socket 1156 CPU itself has 16 PCI-e lanes integrated on the CPU die. This actually gives a Socket 1156/P55-based system a theoretical 24 PCI-e lanes. Not all P55 motherboard implementations utilize all eight of the P55's PCI-e lanes (if you include any onboard PCI-e devices and non-graphics-capable PCI-e expansion slots combined).

And if the P55 chipset itself provides eight additional PCI-e lanes (for a theoretical total of 24 PCI-e lanes), the X58 chipset's ICH10R actually provides six additional PCI-e lanes on top of the 36 graphics-dedicated PCI-e lanes that the main part of the X58 chipset includes (for a theoretical total of 42 PCI-e lanes). However, as I stated a few times, the LGA1366 relies on a potentially latency-inducing external bus to communicate with the 36 PCI-e lanes that the X58 chipset contains.

And the two chipsets are exclusive to the sockets which they had been designed for. A Socket 1156 CPU cannot work on the X58 chipset. Nor can a Socket 1366 CPU work on the P55 chipset. This is because the Socket 1366 CPU lacks many features that have been integrated onto the Socket 1156 CPU's die (such as the integrated 16-lane PCI-e graphics controller), while the Socket 1156 CPU lacks a few of the features that are exclusive to the Socket 1366 CPU's die (the Socket 1366 setup relies almost entirely on external back-side buses, many of which the Socket 1156 CPU lacks, for communication to the connected expansion devices). And Socket 1366 is less efficient than Socket 1156 largely because the external back-side buses add additional latencies compared to the integrated controllers.

I looked at that graphic, and that's misleading. You forgot to look at the bottom of the graphic, which explains that the ICH10R that's used with the X58 chipset adds six PCI-e lanes on top of the X58 IOH's own 36 PCI-e lanes while the P55 chipset itself adds eight PCI-e lanes to the 16 that's integrated onto the LGA1156 CPU's die.

Thus, you have a theoretical maximum of 42 PCI-e lanes in an LGA1366/X58 system or 24 PCI-e lanes in an LGA1156/P55 system (not counting the integrated PCI-e Gigabit LAN). The 36 (for the LGA1366) or 16 (for the LGA1156) PCI-e lanes that you quoted are graphics-slot-exclusive, and cannot be used at all by any other PCI-e slot. Any PCI-e 1x slots included on an X58 or a P55 motherboard are actually run off of a separate chip which works together with the chipset and CPU (integrated into the P55 chipset itself on LGA1156 motherboards or in an ICH10R ICH on LGA1366/X58 motherboards). These i7 systems are equipped with a mixture of PCI-e 2.0 and PCI-e 1.1 buses (the graphics-dedicated slots are all PCI-e 2.0 while any 1x or 4x PCI-e slots are still PCI-e 1.1), unlike the earlier systems (such as the typical Core2 system on a P35 chipset motherboard) which share all of their PCI-e lanes between graphics and other peripherals. (For the record, the P35 chipset provides only 20 shared PCI-e lanes, which leaves only four PCI-e lanes open for expansion with a single graphics card running in full x16 mode.)

Because of this, the Socket 1156/P55 system actually has more PCI-e lanes (outside of what the X58's IOH or the LGA1156 CPU natively provides) for additional expansion capabilities compared to the LGA1366/X58 system (eight versus six). Remember, the LGA 1366 CPU itself provides no PCI-e lanes; it instead relies on an external bus connection to the X58 IOH for everything except the memory controller (which in all i7 and i5 CPUs is integrated on-die). However, as I stated in my previous post, not all motherboards make full use of the maximum number of available PCI-e lanes. After all, what good are 36 PCI-e lanes if you can't use any of them at all for anything besides graphics card(s)?

On the other hand, the compatibility of non-graphics PCI-e cards with graphics-dedicated PCI-e slots (or lack thereof) is a matter of debate - and trial-and-error. Some might work, but others are allergic to the higher clock speeds of the graphics-dedicated PCI-e lanes.

And please, do not take my post as favoring one over the other. I was only trying to clarify any differences between the two.

With all that said, the i7-860 and i7-920 systems are roughly equivalent in cost to one another at present (largely due to the memory prices which are at present significantly higher than they were just three months ago - and the higher memory prices mitigate much of the difference in price between a typical P55 motherboard and a typical X58 motherboard). However, the 860 is the better value if only stock-speed performance is considered. But the 920 does better at overclocking.

The above quote is especially true if an Intel-brand motherboard is thrown into the mix. Admittedly, the Intel-branded P55 motherboard that's sold at MC is overpriced for a microATX motherboard with no overclocking or voltage tweaking capability at all whatsoever. That mobo, plus an i7-860 CPU currently costs over there exactly the same amount of money as the same brand's X58 motherboard with overclocking and voltage tweaking capabilities plus an i7-920 CPU (combo-wise). Under this circumstance, I went for the LGA1366/X58 platform which can be updated to support the forthcoming six-core CPUs. The ultimate difference in price between the two ended up boiling down solely to the cost of the memory itself, at the rate of about $45 for each 2GB of DDR3-1600 memory (as purchased in a matched multi-channel kit).

The i7-920 cannot run memory at anything faster than DDR3-1066 speed with the CPU running at stock speed. That CPU must be overclocked in order to run memory at even DDR3-1333, let alone DDR3-1600, speed.

Alright, I am trying to clear up some of the misconceptions made by myself and others:

It is true that the X58/ICH10R combo has a total of 36 PCI-e 2.0 lanes plus six PCI-e 1.0 lanes. It is also true that the P55 has eight PCI-e 2.0 lanes that are restricted to PCI-e 1.0 bandwidth - in addition to the 16 PCI-e lanes on the 1156 CPU's die. And I was wrong about the PCI-e 2.0 lanes being graphics-only. They can be used by any device. With 36 available PCI-e 2.0 lanes, any USB 3.0 and SATA 6 Gbps controllers on an X58 motherboard will still allow you to have one or two graphics cards running at the full PCI-e 2.0 x16 bandwidth without the need for a safety mode. The 1156 CPU has only 16 full-bandwidth PCI-e 2.0 lanes, forcing the need for a Safety Mode on the controllers which "steals" four of the eight bandwidth-restricted lanes on the P55's PCH (or a PCI-e bridge chip that provides four additional PCI-e 2.0 lanes) and a Turbo Mode which restricts the main 1156 CPU's PCI-e lanes to eight available lanes.

After 10 weeks from last post it seems funny that this particular thread has been resurrected. The platforms each have their advantages. The practical differences are so slight they are almost not even worth a debate, IMO.

However, in this case the debate is between a person and themselves, which I suppose is interesting enough by itself.

Yes, but my most recent reply corrected some of the mistakes I made earlier in the thread.