1 |
Dear Bob, |
2 |
You obviously know you tech stuff. The number crunching I (we) do is |
3 |
huge floating point calculations used in drug design (quantum |
4 |
dynamics/quantum mechanics). At the moment the fastest results come |
5 |
from the Opterons with 1Gb ram. I'm not sure of their spec. But, we |
6 |
have found calculations to be faster on 64bit machines rather than 32. |
7 |
|
8 |
For imaging we use SGI Octane or Octane 2 however I have found recent |
9 |
Nvidia graphics cards to be as good, here at least I stick to crt SGI |
10 |
monitors; they are cheap and perform. Also, Irix (SGI OS) is awful. |
11 |
For weeks now I've been trying to install ssh. It WON'T. I was |
12 |
debating last night whether to install Gentoo on the Octanes. My fear |
13 |
is, If I do: |
14 |
a. Sybyl, the MOST useful drug design software won't run and. |
15 |
b. I will have to reinstall Irix. I did it once. It's a farce. It |
16 |
took all day (no breaks). You even partition when it's installed. |
17 |
I think SG have had their day. We won't buy any more. |
18 |
Furthermore, the Opterons are faster and cost 1/10 the price. |
19 |
|
20 |
> |
21 |
> It depends. In general, looking at MTBF numbers, Itanium systems are |
22 |
> less reliable than most 2P to 4P x86 servers out there. But more reliable |
23 |
> than most RISC based servers, though IBM's Blue Gene blade servers may |
24 |
> have an equivlant record or even be better. (Note: Realibility when I |
25 |
> use it means 24x7 full time compute load, not shutting it down at night |
26 |
> nor the system standing idle.) |
27 |
> |
28 |
> As to performance, it kind of depends. The bone stock Intel chipset |
29 |
> Itaniums have fairly constrained FSB to memory bandwidth. Itanums where |
30 |
> companies have used their own chipsets - such as us (SGI) and higher end |
31 |
> HP, among others, can deliever some impressive bandwidth to memory and |
32 |
> I/O. Thus improving overall computing throughput and high sustained |
33 |
> compute power. |
34 |
> |
35 |
> And depending upon the problem set, an Itanium based system may be faster |
36 |
> than amd64 based systems. If your problem set is more integer based or |
37 |
> transaction based, the amd64 will perform better. If it's more FP based, |
38 |
> memory constrained, threaded, and I/O bound, then an Itainium based system |
39 |
> might be better. |
40 |
> |
41 |
> If you need more than 1 TB physical memory and up to 512 cpus in a single |
42 |
> system image, then amd64 solutions won't be there until HyperTransport III, |
43 |
> sometime next year, perhaps. |
44 |
> |
45 |
> Downside - you have to recompile your apps to get the best out of them. Any |
46 |
> x86 32bit code will run in an emulation layer, which can be very fast or |
47 |
> very slow. Setiathome, 32 bit client screamed on our Itanium's (ran it |
48 |
> on a 512P system). Oh, and you'll need Intel's compiler. |
49 |
> |
50 |
> On the AMD side, figure about 80% FP performance and 102% Integer performance |
51 |
> compared to an Itanium cpu. Memory bandwidth tends to match when AMDs are |
52 |
> used in 4P or greater configs. I/O bandwidth is a bit constained in some |
53 |
> instances, but seems pretty tolerable and really isn't much of an issue |
54 |
> until more than 4 PCI-X slots are needed. But a lot of that is based |
55 |
> on which chipset is being used. And to get the est out of your apps, you'll |
56 |
> need to compile them with the PathScale compiler. |
57 |
> |
58 |
> If you're interested in an in-office/deskside style system. the AMD solution |
59 |
> is probably a better match, especially if your application needs only 1 or |
60 |
> 2 Gfx pipes, and fits in less than 16 GB memory. |
61 |
> |
62 |
> If you need department scientifc serving, then benchmarking your apps on each |
63 |
> platform is the better way to find which fits better. |
64 |
> |
65 |
> If you need serious visualizaton ability, with 3 or more - up to 16, Gfx |
66 |
> pipes, then there isn't much choice, no AMD vendor has produced a system |
67 |
> capable of doing that, though with some special Nvidia cards, it's possible |
68 |
> to get 4 Gfx pipe on 2 PCIe x16 slots. That's not to say it works well, as it's |
69 |
> pretty much beta hardware and drivers at the moment. |
70 |
> |
71 |
> Bob |
72 |
> -- |
73 |
> - |
74 |
> QA Curmudgeon. |
75 |
> Wacky and bizarre testing(TM) performed while-U-wait. |
76 |
> - |
77 |
> See http://www.gnu.org/philosophy/no-word-attachments.html |
78 |
> - |
79 |
-- |
80 |
Dr Gavin Seddon |
81 |
School of Pharmacy and Pharmaceutical Sciences |
82 |
University of Manchester |
83 |
Oxford Road, Manchester |
84 |
M13 9PL, U.K. |
85 |
|
86 |
-- |
87 |
gentoo-amd64@g.o mailing list |