Cebit 2006: High end Motherboards and Server News
by Johan De Gelas on March 13, 2006 12:05 AM EST- Posted in
- Trade Shows
AMD
AMD showed the new socket-F Opteron, a CPU with a 1207-pin Organic Land Grid Array (LGA) interface. The socket-F Opteron will support DDR-2 667. The maximum power consumption of these CPUs seems to go up to 95 Watt (Standard) instead of 89 Watt, and 68 instead of 55 Watt for HE low voltage Opterons. On the positive side, the quad core Opterons will not exceed these power requirements. A low voltage (65 nm) quad core Opteron, due in 2007, will consume the same 68 Watt maximum.
AMD also took a stab at Intel's FB-DIMMs strategy. One of the biggest advantages that the Opteron has is the fact that each CPU has its own memory. This allowed a dual Opteron to use 8 DIMMs while the Xeons were limited to 4. Intel's Nocona platform supported DDR-2, which allowed also 8 DIMMs.
Now that AMD moves to Socket-F and DDR-2, the Opteron can use up to 8 DIMMs per CPU. Intel Counters with FB-DIMMs, which uses full duplex serial point-to-point technology.
This technology enables 16 DIMM slots on one board. However AMD points out that the Advanced Memory Buffer (where the serialisation takes place) consumes up to 6 Watt. AMD then claims that this means that Intel systems will consume almost 200 Watt more when equipped with 32 DIMMs of FB-DIMM instead of standard DDR-II.
In a more realistic situation, the memory subsystem of the Intel server with 8 to 16 DIMMs will consume about 24 to 64 Watt more than an Opteron based server and not 100 Watt as AMD's slide seems to indicate.
AMD showed the new socket-F Opteron, a CPU with a 1207-pin Organic Land Grid Array (LGA) interface. The socket-F Opteron will support DDR-2 667. The maximum power consumption of these CPUs seems to go up to 95 Watt (Standard) instead of 89 Watt, and 68 instead of 55 Watt for HE low voltage Opterons. On the positive side, the quad core Opterons will not exceed these power requirements. A low voltage (65 nm) quad core Opteron, due in 2007, will consume the same 68 Watt maximum.
AMD also took a stab at Intel's FB-DIMMs strategy. One of the biggest advantages that the Opteron has is the fact that each CPU has its own memory. This allowed a dual Opteron to use 8 DIMMs while the Xeons were limited to 4. Intel's Nocona platform supported DDR-2, which allowed also 8 DIMMs.
Now that AMD moves to Socket-F and DDR-2, the Opteron can use up to 8 DIMMs per CPU. Intel Counters with FB-DIMMs, which uses full duplex serial point-to-point technology.
This technology enables 16 DIMM slots on one board. However AMD points out that the Advanced Memory Buffer (where the serialisation takes place) consumes up to 6 Watt. AMD then claims that this means that Intel systems will consume almost 200 Watt more when equipped with 32 DIMMs of FB-DIMM instead of standard DDR-II.
In a more realistic situation, the memory subsystem of the Intel server with 8 to 16 DIMMs will consume about 24 to 64 Watt more than an Opteron based server and not 100 Watt as AMD's slide seems to indicate.
Facebook
Twitter
RSS
19 Comments
View All Comments
Genx87 - Wednesday, March 15, 2006 - link
Itanium's marketshare is most likely at the expense of PA-RISC which is what it is replacing on the HP side.IBM, Dell, or Sun arent interested.
Outside of HP, what big OEM is shipping enough Itanium machines to bother mentioning?
logeater - Tuesday, March 14, 2006 - link
Here's a quick tip: Olives liven up the most jejune of pasta dishes. I prefer Spanish myself, but even Greek Kalamata can give it that zesty flavour for your next party function or Sunday dinner. Be sure to wish and thoroughly pit the brown fruit before slicing them.Phiro - Tuesday, March 14, 2006 - link
And what's with all the bagging on server virtualization? I think Anandtech's viewpoint on this is too focused on a crappy product like Microsoft's Virtual Server.We use ESX 2.5 from VMWare where I work, and while management is still worried enough about the risks to run tier 1 services on VMWare, we run tons of tier 2 on VMWare and most of non-Production on VMWare. VMWare is poised to reengineer our disaster recovery systems as well.
Going down the road, the global manager in ESX 3.0 looks like an absolutely killer feature and we will definitely rearchitect our environments to take advantage of it. We're already seeing tremendous cost savings in hardware with ESX 2.5, 3.0 will only increase that margin.
If Anandtech can't follow what the market leader in server virtualization is doing, nor are they able to get the product working correctly, they might need to take a clue from what the rest of the world is doing. Normally you guys are pretty in tune with trends - I think you're way out in the rain on this one.
Stolly - Wednesday, March 15, 2006 - link
Totally agree. That assesment of virtualisation displays a lack of real world experience. We have implemented ESX at a customer who have collapsed a 40 server system in less than 10. They have some 2 node clusters with one node being real hardware and one node being inside ESX, its the hardware nodes that remain a problem. Servers inside ESX are NOT more prone to problems.Plus, hardware migrations are a thing of the past. Using vmotion they can move a running server from one ESX server to another with 0 downtime, the users do not even notice. A multi week phased hardware migration can now be done in minutes. Thats the power of server virtualisation, and i'm suprised that Anandtech is not conversant with the latest state of the art.
JustAnAverageGuy - Monday, March 13, 2006 - link
"no less than" = "up to""No less than" implies that that it is a minimum amount.
"Up to" implies that the value given is a maximum.
I doubt a server requires a minimum of 128GB of RAM. :)
However, another excellent article, as always, Johan.
- JaAG
DSaum - Monday, March 13, 2006 - link
One the basis of "a few vague benchmarks", you state "Montecito is not only a vast improvement compared to Madison when it comes to running typical database applications, but also the platform has simplified quite a bit too." What happened to your objectivity? LOLdexvx - Tuesday, March 14, 2006 - link
Here's the HP briefing of the Moniceto:http://www.hp.sk/mediaservis/prezentacie/pdf/7_Mon...">http://www.hp.sk/mediaservis/prezentaci..._Monteci...
But lets roll over the basics:
Madison: 1-1.5Ghz, 32KB L1, 256KB L2D, 9MB L3
Moniceto: Dual Core 1.6Ghz+ with HT, 32KB L1, 1MB L2I and 256KB L2D, 2x 12MB L3
It does not take a genius to come to the conclusion that Moniceto will be a LOT more performance oriented.
JohanAnandtech - Tuesday, March 14, 2006 - link
Maybe because it is very obvious? 1 MB L2 instead of 256 KB L2, two cores versus one, 4 threads versus 1...that is more than enough to call the montecito a vast improvement over Madison in database and other enterprise applications.Would you need benchmarks to know that a clovertown which has twice the cores of Woodcrest, but the same architecture, is a vast improvement in these kind of benches?
FreshPrince - Monday, March 13, 2006 - link
imagine the fps you'd get from that beast... :DSAS really isn't that impressive yet...
the enclosures I've seen are mostly 12 drive external cases...which can't do much.
The SAS white paper I've read described a much more scalable solution, and you can't find those enclosures anywhere yet...
I'll stick to my NexSAN SATABeast.... :D
SCSI backplane + 42, 500GB, SATA 3.0GB/S = 21TB raw in a 4U device.
Until they come out with something equally impressive with SAS, don't bore me anymore ;)
cornfedone - Monday, March 13, 2006 - link
From the crap Asus has shipped in the past three years they can't even deliver a properly functioning mainstream mobo, let alone a high-end product. Their SLI, An8, ATI 480/580 mobos are all riddled with voltage, BIOS and memory issues that Asus can't or won't fix. Their days are numbered.