|
|
Back to UserFriendly Strip Comments Index
|
High end server mb | by Kickstart | 2004-10-22 13:47:16 |
|
Never used them personally, but a lot of people | by PsychoI3oy | 2006-11-19 12:55:59 |
|
Opteron vs. Xeon | by Kickstart | 2004-10-22 14:00:23 |
| AFAIK, |
by rorajoey |
2006-11-19 12:55:59 |
the Opterons are 32-bit compatible, so they'll run a 32-bit OS efficiently until such time as a 64-bit OS is released; and when that happens, you'll already have the infrastructure to support the new 64-bit OS.
And here's a news item on xBit Labs, dated 2/16/04:
Today AMD cut down prices of Opteron 800 series processors by around 50% – an unprecedented move no one would expect from a server CPU maker. Pricing of AMD Opteron 200 series is aligned with Intel Xeon product line, while the costs of Opteron 100 series remained on the previous level.
And this from extremetech.com, regarding HyperTransport implementation:
Coherent links are used to communicate cache-coherency information between multiple processors in an MP system configuration that share data...The 200-series will have 3 HyperTransport links, but 2 are non-coherent, and 1 is coherent. The 800-series will have 3 coherent HyperTransport links.
Multi-processor systems using Opteron CPUs include a bank of memory for each processor, while the Xeon processors are designed to share a single pool of RAM. According to the extremetech article quoted above:
Bandwidth, also, is higher than a bus-based system. Even using an 800 MHz FSB, as in a future Xeon version of Intel's Canterwood chipset and a 3.0GHz P4 (when it works correctly), with a shared memory system and four processors attached to the bus, you can get 6.4GB/sec of memory bandwidth to be shared among the four processors. In a 4P Opteron system, each processor-to-memory local link is 6.4GB/sec aggregate (or 3.2GB/sec each way) and multiplied times four processors is 25GB/sec aggregate bandwidth to memory.
|
|
[ Reply ] |
|
|
[Todays Cartoon Discussion]
[News Index]
|
|
