You know, I keep saying this and I seem to have to remind everyone of this when people start hating on BD and PD. The simple fact is there is less and less single-threaded software now and software companies tend (if they're smart,) to make software to be multithreaded when it needs that extra compute power. If AMD increases the IPC on PD, you're not just increasing it on 4 cores, but rather 8. I would like to see an Intel chip for the same price as the 8120 that can encode video, play a video game, and be updating Windows and still have some kick to do something else as well. Intel's HyperThreading is nice, but it doesn't scale well and only does so good on selective workloads since it's only used unused portions of the CPU where BD/PD has dedicated extra hardware to running those extra threads so scaling is much more linear on multi-threaded workloads.
Is AMD slower than Intel thread for thread? YES! Only an idiot would try to dispute that because AMD's IPC isn't up to par and quite frankly neither is AMD's IMC (now, it used to be good but they haven't changed a whole lot to it, also having a SB-E with quad-channel memory, it won't help you until you start running the CPU (all cores,) over 50% and even that is dependent on the workloads.)
Is AMD, as a multi-threaded platform, better than Intel? With SB, sure was. Only problem is the majority of users don't use that kind of software and it isn't widely available for most tasks.
As I see it the following will happen with PD gets released in comparison to IVB. IPC will be improved but will still trail IVB and maybe even still SB. Clocks on PD will be increased for the FX processors and power consumption will be moderated with the use of RCM. However where the IPC improvement will really shine is multi-threaded workloads (once again,) since that IPC improvement will be across
all logical threads and not just the physical modules. With that said I once again believe we will see AMD demolishing multi-threaded workloads and being just good enough on single-threaded workloads. So a mixture of IPC improvements and clock speed bumps I think we should see a decent product.
I think a lot of people need to realize that AMD and Intel have two different goals in mind with their CPUs and I think AMD has the right idea even if it isn't proving to be the right one as it stands right now. Don't get me wrong, I like both AMD and Intel as companies, it's why I've been bopping between the two for the last several years, but right now Intel has the crown so I went with SB-E. AMD is making their architecture so it will scale nicely, Intel on the other hand is still squeezing performance out of their same architecture, which isn't a bad move, but I bet you that Intel will find that there will come a point where you can only improve the architecture so much. Keep in mind that a BD module is only like, what, 20% larger than a Phenom II core? As far as raw performance for the size of the module's die size, that's pretty impressive and if AMD keeps going that route we could see CPUs with a lot more cores and a lot more multi-threaded horsepower while Intel is still leading single-threaded tasks.
Additionally, for the cost, I would jump on an Interlagos 16-core CPU for servers rather than an 8-core Xeon, mainly because the 8-core Xeons run really hot and don't have as much kick for server applications and costs half as much.
I just thought that pointing out both Intel and AMD's strong points would be better than saying what each of them sucks as doing because honestly, they're both good chips, just Intel does some things better than AMD and AMD does some things better than Intel, simple as that.