I read the article. I sincerely hope AMD will bounce back in shape... I believe that the 8 core cant be wrong. You see I believe the more physical cores a chip has, the better it will be once the devs and software are effectively using the cores.

I think, and I may be wrong, that the current software (and it may even apply to Linux) are not even using all 8 cores efficiently.

I wouldnt be surprised to see the Linux kernel being modified soon to effectively use all cores to their maximum potential, and when that happens, the 8 core chips will be ahead of anything else with 4 & 6 cores... The 8 core chip might just be ahead of its time.

After all,the Bulldozer chip is a consumer chip right? I understand most consumer applications are not even capable to use more than 2 cores, and most games dont even see all cores. I think the Bulldozer is just for now a "misunderstood" chip. same applies for Intel optimized benchmarking applications that are used to test AMD chips. The results will never be the same as if it was an intel chip being tested.

Again, Im not a CPU expert.

I understand what you are saying. I however dont want to run 12 conputers at the same time. Electricity costs will be over board and Im not sure my gf will like the idea of a geekier geekdom that what i already have ;)

Plus, I may very well use the rig for other CPU/GPU intensive tasks like converting movies and fluid dynamics and other FEA tasks. With a "farm" I imagine I could always do these tasks but how complicated?...

I am not sure to understand you. Are you saying that server mobo's are not normally as upgradeable as consumer boards? In other words, will I be able to upgrade 2 or 3 generations of consumer grade CPU's while I miht only be able to upgrade once for server's CPU?

If I got your point, then why are they advertising server stuff time proof with long term reliability and comaptibility? Most companies with hundreds of Xeon or Opteron based servers are not upgrading often, but that doesnt mean that these mobos are not capable of running newer CPU's? As long as the socket is the same it should work?

Am I wrong??

Found this in a discussion on xtreme systems website. This is what's coming. I don't know what this would cost but it is food for thought if you want to wait.





That's approximately the same PPD of 3 2600k SBs running at 4.5Ghz on about 25-30% less power (assuming 400w draw is accurate) not to mention less heat and less things to break.

An i5-2400 is 4/4 according to Intel, so 4 would be 16/16.
All the 22nm CPU's will be much more power efficient (about 50% more efficient than SB), not just the server cores. So imagine an i5-2400 with a 35 to 45W TDP, if you like.
Clock for clock we are likely to only see around a 5 or 6% increase in performance, and clocks are only set to modestly rise, so processor cores will only be around 10% faster. The chipset and motherboards will also be more efficient; 1.35V DDR3 for example, and faster clocks.
It's also likely that we will see 8 core and perhaps 10 core processors for desktops within a year, and with HT.
The fact is that the high end GPU's (or equivalently priced GPU's to that of a CPU) will do far more work (10 to 60 times more on a GPU implementation of HCC). So 2 single socket systems would be much less expensive than a quad skt server system, and would better facilitate GPU's; the chipsets typically support one x16, 2 x8, or 2 x8+ 1 x4 / 1 1x16. This is important because high end GPU's cannot be fully utilized at x4 (at x4 PCIE2 I experienced a 14% decrease in performance on a GTX470 and a GTX580 saw a 20% decrease). 1X is awful. The next (pending) generations of GPU (HD7000 and GTX600 series) will tax even the 2 x8 PCIE3.0 bandwidths (which are equivalent to x16 PCIE2.0 slot bandwidths). As for power efficiency on the GPU, they use more power but do much more than a CPU, and the forthcoming generations will be 28nm; more efficient and more powerful than present GPU's. Thus, two boards will be better than one.