Benchmark Comparison: 486/66-VLB PC - 16Mb , 256k cache [EIDE]
Noname 166Mhz PCI - 32Mb , 256k cache [NCR SCSI]
Notes: Cache obviously makes a difference on memory operations.
In general, the alpha memory subsystem must be brain damaged to give
same performance on double the clock speed.
Alpha beats PC on some arithmetic functions but not by much. In general
overall performance looks pretty iffy.
Disk benchmarks look suspect, but 'dd' backs up the results to a certain
extent.
The PC is over twice as fast at block transfers.
[dd if=/dev/sda2 of=/dev/null bs=<size> count=<count>]
Summary: What's all the fuss about; this noname board seems worse than useless.
buy a P130! they're half the price and faster!
Apart from that I have a fully functional network system with web server
telnet, ping etc & so forth that appears pretty stable.
Nice job by Linus & Co, but the message from DEC now seems to be, "now you've
spend the time getting it working you need to spend 3 grand to upgrade
to a board that fits the advertised benchmarks".
[my colleague has an identical that gives almost equally poor results on NT]
Help! There I must've done something wrong somewhere ????????!!!!!!!!
[Ratio - large number indicates the slower machine]
Alpha PC
Test Rating / Time (u+s) Rating / Time (u+s) Ratio
Dhrystones per Second: 85714 32450 1:2.6
Memory Access:
Array Size: 512 bytes 0.033 0.17 1:5
Array Size: 1024 bytes 0.050 0.17 1:3.4
Array Size: 2048 bytes 0.050 0.18 1:3.4
Array Size: 8192 bytes 0.050 0.17 1:3.4
Array Size: 16384 bytes 0.067 0.18 1:3.7
Random Memory Access
Array Size: 512 bytes 0.21 0.21 1:1
Array Size: 1024 bytes 0.21 0.21 1:1
Array Size: 2048 bytes 0.21 0.20 1:1
Array Size: 8192 bytes 0.25 0.21 1:1
Array Size: 16384 bytes 0.30 0.21 1.4:1
Arithmetic Test (type = arithoh) 0.02 0.01 0.08 0.02 1:3.3
Arithmetic Test (type = register) 1.12 0.01 1.16 0.02 1:1
Arithmetic Test (type = short) 1.22 0.01 1.33 0.02 1:1
Arithmetic Test (type = int) 1.11 0.02 1.16 0.02 1:1
Arithmetic Test (type = long) 0.99 0.01 1.17 0.02 1:1.2
Arithmetic Test (type = float) 0.38 0.02 1.98 0.02 1:5
Arithmetic Test (type = double) 0.56 0.01 1.98 0.02 1:3.5
System Call Overhead Test 0.02 0.09 0.01 0.12 1:1.3
Pipe Throughput Test: read & write 0.01 0.26 0.01 0.09 2.7:1
Pipe-based Context Switching Test 0.00 0.05 0.00 0.05 1:1
Process Creation Test 0.02 0.97 0.01 0.52 1.8:1
Execl Throughput Test 0.05 1.72 0.04 1.01 1.7:1
Filesystem Throughput Test
File Size: 50 (1024 byte) blocks 0.00(user) 0.65(system) 78k 0
File Size: 100 (1024 byte) blocks 0.00(user) 0.70(system) 146k 10M/sec
File Size: 200 (1024 byte) blocks 0.00(user) 0.80(system) 256k 6.6M/sec
File Size: 400 (1024 byte) blocks 0.01(user) 1.01(system) 405k 5M/sec
File Size: 800 (1024 byte) blocks 0.00(user) 1.45(system) 564k 4.7M/sec
File Size: 1600 (1024 byte) blocks 0.03(user) 2.32(system) 706k 4.5M/sec
Failed to get working;
xntpd
gated
....
+--------------------------------------------------------+
| Gareth Bult - Frontier Internet Services Limited. |
| Sent from the Office: 0171 242 3383 (gareth@ftech.net) |
+--------------------------------------------------------+