The AMD Opteron 2214 is the Rev F version of the 275, the same chip we
bought last fiscal year in the yili systems. It was the sweet spot at that
time. The last Sun price I saw was $3464.80 for an 8x1GB DIMM config that
was only a single 250GB drive. I think a second 250GB drive should be on
the order of $195 or lower.
The pricing I have seen so far from both Microway and Dell (have asked Sun
for this info) would seem to indicate that the 2216 is the sweet spot now.
My rough analysis (would need more checking) of the CPU pricing from both
Microway and Dell is something like this:
Model Frequency Microway Dell list
Price Price
2210 1.8GHz $135.00 $244.44
2212 2.0 $147.50 $293.00
2212HE 2.0 n/a $342.00
2214 2.2 $198.64 n/a
2216 2.4 $217.50 $366.67
2218 2.6 $272.69 $413.85
2220SE 2.8 $325.36 $471.79
So I think the 2216 may be a better spot. (And of course the power
consumption/heat dissipation is the same 95 watts for everything except the
65 watt 2212HE and the 119 watt 2220SE. And with variations on fans, power
supplies, and mother boards, even the 30 watt savings of the 2212HE might
disappear.)
This purchase, like any other, needs to include pricing for racks, for
cables (2 Ethernets plus serial console plus hub/switch/rack for IPMI plus
serial concentrators for console lines), plus the price for Cisco Ethernet
1Gb ports.
If we can coordinate this buy with KIPAC's buy, you should get better
pricing all around.
On Thu, 30 Nov 2006 09:08:43 -0800 Chuck Boeheim wrote:
>
> Ok, here is what I'm planning to ask Sun to quote for the
> initial Atlas farm:
>
> 3 18TB fileservers: V240 + 3 tray 6140 (1 controller + 2 expansion)
> 64 Compute nodes: X2200M2, 2x2214, 8GB (8x1GB DIMMS), 2x250GB drive
>
> The storage should come to around $152K and the compute nodes
> to around $223K. That comes about $25K short of $400K, which
> we can either hold back to fix miscalculations (e.g., too much
> storage per server, have to buy more fileservers to spread
> the load), or we can buy 5 or 6 more batch nodes.
>
> Any last-minute course corrections before I go ahead with this?
>
> Chuck Boeheim
> [log in to unmask]
>
>
>
