All the signs are pointing in the right direction and AMD is ahead of their schedule. But when I consider comparing this to intel's 2010 line up. I think it falls a little short. We'll have to see what happens when 16 hyperthreaded xeons go up against 16 real thread AMD's at the end of 2010.
the only thing I think we be can be sure of is AMD retaking the budget server market share that they lost to core2 based xeon servers.
The real potential greenfield for AMD's new memory-rich, high core density servers (2P and 4P) is in desktop virtualization and server based computing. As this market is expected to boom in 2010/2011 AMD's mix of core count and "cheap memory" put Maranello platforms on the top of the VDI price-performance heap.
Considering the net savings in energy proposed by VDI migrations, the 4P Maranello systems should find their sweet-spot in that arena. Overall system prices will be determined by chipset integration costs, unit CPU and memory costs. Besides dense memory and cores, the chipset platform for Maranello is a simple, well known entity with partners having lots of lead time to market. This will make it difficult for EX to challenge in value.
In the broader server market - increasingly driven by virtualization forces - AMD offers a strong value proposition if it can hit targeted performance and price points. With DDR3 memory coming down, AMD appears to have the initiative (for a change) at a time when companies are loosening budgets and looking to enter into a new round of vendor affinity. As AMD and Intel's directions begin to diverge again, cashing in on this opportunity to win market share is crucial for AMD and that requires execution: something they've being steadily getting better at since Shanghai.
I just deployed a new 16-core compute server after checking everything in the market I could find. And it is with a 4-socket Nvidia board, 4 AMD 8380 CPUs, 16 GB of RAM. The motherboard is very upgradable. We can either go to Istanbul Opterons to upgrade to 24 faster cores or add a daughter board to increase sockets to 8. Intel was nowhere close to this configuration/price/upgradability. In the last three years this is the third time we ended up buying AMD. The previous configurations were even bigger systems. I especially like AMD platform due to its uncomplicated, straightforward configuration and upgrade path.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2010
I have enjoyed this run of four complimentary articles. Thank you.
I think AMD will do fine relative to the hyperthreaded xeon processors. True cores scale so much better. Hyperthreading gives ~10-20% performance gain, but there are also places where your overall throughput drops because of cache contention or other issues. Hyperthreading is not the silver bullet that some people paint it to be.
Thrax
🐌Austin, TX Icrontian
Comments
:celebrate:celebrate
Great piece, Rob.
the only thing I think we be can be sure of is AMD retaking the budget server market share that they lost to core2 based xeon servers.
http://solori.wordpress.com/2009/12/30/quick-take-year-end-dram-price-follow-up-thoughts-on-2010/
Considering the net savings in energy proposed by VDI migrations, the 4P Maranello systems should find their sweet-spot in that arena. Overall system prices will be determined by chipset integration costs, unit CPU and memory costs. Besides dense memory and cores, the chipset platform for Maranello is a simple, well known entity with partners having lots of lead time to market. This will make it difficult for EX to challenge in value.
In the broader server market - increasingly driven by virtualization forces - AMD offers a strong value proposition if it can hit targeted performance and price points. With DDR3 memory coming down, AMD appears to have the initiative (for a change) at a time when companies are loosening budgets and looking to enter into a new round of vendor affinity. As AMD and Intel's directions begin to diverge again, cashing in on this opportunity to win market share is crucial for AMD and that requires execution: something they've being steadily getting better at since Shanghai.