It totally depends on the type of work. I have done a lot of testing back when the Opteron 6100s first came out. Depending on the type of workload I use, I can make either of them appear to be the winner. Intel shines in some areas, AMD shines in others. For most applications, the lower-priced AMDs are much more than adequate, and in virtualization, they appear to be the more powerful of the two choices right now. I will qualify that by saying that I have not tested the new 10 core Intels.
You can pretty much prove or disprove anything with enough stats and graphs 
Allen Beddingfield
Systems Engineer
The University of Alabama
Shane wrote:
scowse wrote:
There is truth in that article from Eric.
"...VMware had to do something. Otherwise, the massive scale-up potential for servers would result in fewer hosts with fewer sockets -- and fewer licensing dollars for VMware."
Like I and many others have stated before we dont disagree with the new model, we disagree with the ludicrous entitlement amounts. Had they implemented this model and limitations 3 or 4 years ago it would have made sense but not today.
I do disagree with the new licensing model. The goal at VMWare should be to completely eliminate the competition, by making a rock solid product which is worth buying. By overpricing it, they simply open the door the competition.
A windows 2008 license costs me roughly $600 each-this helps offset the cost of writing the software (which represents millions of lines of code). A vmware license costs me 4 times this, per socket, and represents substancially less development work. In my mind, is it the best product on the market-absolutely, but it's time for VMWare to drop the EMC (lube required) licensing model, and start charging their customers more fairly for what they are getting.
(In other words, sure it works great but it was a heckuva lot easier to write than windows so let's price it accordingly!)
As a first time poster to VMTN, I'd quite sad that I'm spending my first message on this licensing topic. However, while I don't agree with the current licensing standpoint of vRAM, I wanted to put in my $.02 about licensing per CORE and why I don't think it will work (which is why VMware needs to find the best licensing scheme which doesn't penalize current customers but allows them to get a piece of pie as their customers flourish).
The definition of a core is not a guaranteed science today nor in the future. There are processors being developed with multi-core capacity that is devoted to and excelling at executing specific types of instructions. There are also chips being researched (and produced) that are comprised of hundreds of "cores" that allow themselves to be reprogrammed at a moments need. Let's look at an over-simplefied example; if a cpu had four full cores designed to executed standard instructions but two additional cores designed to only handle complex floating point operations, would that be a quad core of six core CPU? Why should customers pay for cores that software can't completely run on but nevertheless gain a bonus from those two special "cores" under certain computational processes? If I have a 500 core chip that reprograms itself into six independent execution pipelines, should I pay for all 500 cores or only 6?
The future of CPU's is currently undergoing some interesting research and proof of concepts. It's absurd to think these advances won't spill over into our x86 world at some point in the future. What do we do then? Wait for VMware to create yet another licensing model?
Unfortunately, it's much easier to charge on a per physical device basis. The physical nature of a server won't instanteously change without some forethought being put into it. I understand that VMware is looking towards the future as hardware surpasses the current generation's capabilities, but they are doing it at the wrong time. I only hope that they can resolve this problem and get their customers excited about the new features rather than spending time in test labs looking for alternatives. I personally believe that a per-socket base cost along with separate but incremental physical RAM packs were the best way to go. That way, every server has an associated base license for the product and VMware gets a chunk of the cash as the server "scales-up". If people want to overcommit their resources, then let them take the chance of performance degradation. And for those people thinking that will reduce VMware's devotion to their advanced memory utlization mechanics, then so be it! As a company they need to survive to bring you the technology you want or need... honestly, who wants a customer that demands more but doesn't offer to pay for it?
.... and also, VMware may be first to the punch here, but don't think the other vendors aren't watching to see if they too can find a better model that nets them more profit as server hardware advances. They too need the cash to fund their own research initiatives, especially if they plan to go from market follower to market leader - or find themselves in that position in the future.
We all need to find what's best for our own institutions, but remember that AFFORDABLE quality saves money, time and resources in the long run compared to purchasing today's special of the week. VMware needs to ensure they are affordable!
P.S. Oh yeah, the notion that a Hyper-V farm is free is absurd. Even if you create a farm, it's useless without the Microsoft licenses you need to buy to run Microsoft guests; unless you plan on running free linux guests on Hyper-V, but that too is absurd... lol 
However, if you are strictly a Microsoft customer with datacenter licensed per socket, I can agree that Hyper-V is worth a look if you are considering alternatives.