Thoughts, HEXUS.awards and HEXUS.right2reply
Our disappointment last time round with Paxville Xeon is almost completely alleviated with this new generation of the Xeon platform, based around new core logic and a Core-based Xeon processor.The combination of dual dual-core Woodcrests at 3.0GHz, each on a 1333MHz bus of its own, with a higher performing memory controller and better base core logic means that, for the first time a couple of years, there's something other than Opteron worth considering for the high-end workstation market.
The 'wide' nature of the Core microarchitecture, and in particular the healthy boost in floating point execution resources, per core, means that Opteron is outclassed for the first time since its release by a competing processor, clock-for-clock. Pair that with Woodcrest's frugal nature, even at 3.0GHz, and it's hard to say much of a bad word about the brand new 51xx-sequence Xeon CPUs and their host platform.
FB-DIMM negates some of the power advantage of the processors, but at the speed we tested at, access latency and bandwidth are beginning to approach what's really needed to feed a pair of Core processors properly. Another healthy glug of memory bandwidth would see the hardware do even better.
It's not an across-the-board win for new Xeon, but in general the new CPU architecture shows off nicely, and Intel are making host platform changes that make sure it's able to work properly, as unhindered as possible.
In short, AMD have serious competition in the high-end workstation space with Core-based Xeon significantly upping the game compared to the last gen Xeon which AMD found easy to dominate.
And if you were to think to the future and ponder Intel's Kentsfield, which is just two Conroe processors under a shared heatspreader, sharing the bus to the MCH, you'd glimpse most of the performance right here with this eval of 5160s. While we can't say much and we whisper this to you in hushed tones, performance of a Kenty clocked in the 5160 range is scarily similar in places.
Armari Systems
The Magnetar X2, configured as we tested, is some £4800 inc. VAT. 8GiB of memory and the highest-end Xeon 5160s push the price up, and it seems there's somewhat of a price premium on the Supermicro X7DAE mainboard at the time of writing.Compare that to £4600 or so for the tested Gravistar XRS, but with Raptor disk and 8GiB versus 4GiB, including £1300 of Opteron 285 CPUs, and pound-for-pound the two systems trade heavy blows with each other.
Armari integrate a sweet workstation, and your author would seriously ponder the price difference to invest in a Magnetar X2 versus the Gravistar XRS, almost everything being equal and comparing Opteron 285 to Xeon 5160. My only complaint is that the Magnetar X2's cooling solution, where the CPU and memory area on the mainboard are shrouded off and cooled via the bank of fans, is overly noisy.
Given Xeon 5160's TDP, we imagine Armari stick with the Supermicro hardware there for reasons of FB-DIMM heat, which is somewhat of a shame. That'd be the only thing putting me off, were I spending close to £5000 on serious workstation hardware in my DCC and development pursuits.
The power of the Magnetar X2 with the quietness of the Gravistar XRS, please!
Summary
Intel gives Xeon its balls back (and scarily so in places, with the performance turnaround astronomical depending on the codes), and now it's AMD's turn to play catch up again. We can't wait to see what they have in store.HEXUS Awards
Intel Xeon 5160 Processor
Armari Magnetar X2
