rp440-8 (L2000)vs rp7400 (N4000)

[Karl Myers_1]

I am trying to find some performance specs (specint/tpm) and performance data, that I can compare these two machines with.
Can anyone help please.

[Joseph Loo]

hi,

refer to this for rp7400:

http://www.abtechsys.com/hp_servers/rp7400.html

and rp4440:

http://www.abtechsys.com/hp_servers/rp4440.html

regards.

[Joseph Loo]

another website:

http://www.alimartech.com/9000_servers.htm

regards.

[DCE]

One more site....

http://www.tpc.org

[Jeff Schussele]

Hi Karl,

Well to put it simply, An L-class is simply *half* an N=class.
That doesn't mean an N-class is twice as fast.
It means an N-class has twice the I/O capability as an L-class.
I'd say that an N-class has probably a 70% higher I/O throughput.
Now, an N-class can also have twice the CPU count as a rp5470 - same rule applies....

Rgds,
Jeff

[Ted Buis]

You are confused. The rp4440-8 is much different than the L2000 (a.k.a. rp5450). For starters, the L2000 had only 4 CPU cores while the rp4440-8 can have 4 Dual core processors or 8 processor cores. The max clock rate on the L2000 was 540MHz, while the rp4440-8 is either 800MHz or 1GHz with either the 32MB L2 cache with the pa-8800 or the 64MB L2 cache of the pa-8900. The bandwidth to memory on the L2000 was only 1.3GB/sec while on the rp4440-8 it is 6.4GB/sec for the processor bus to the zx1 chipset and 12.8 GB/sec to memory with up to 4GB/sec for I/O. The rp7400 was much more capable, but 750MHz was the fastest CPU possible, and earlier N4000 could be as slow as 360MHz, with a max of 8 CPUs (up to 8.6GB/sec memory if all memory carriers are installed). The rp7400 had more independent PCI buses (10) than the rp4440-8 (6) for I/O, but the were the slower PCI-4X slots rather than PCI-X. For most applications, I would expect the fastest rp4440-8 to outperform the fastest rp7400 by 10% to 30% assuming you were not memory or CPU bottlenecked. The rp4440-8 supports twice the RAM of the rp7400, so that of course raises its potential a bit if you need large memory.

[Ted Buis]

Oops, I'm wrong on one point, the rp4440-8 (128GB) can have up to 4X the memory of the rp7400 (32GB).

[Joshua Scott]

Estimated OLTP rates:

rp4440 up to 176,000
rp7400 up to 80,000

Josh

[Ted Buis]

Some of those tpm estimates are higher on the rp4440-8 due to its having much greater memory, better disk systems and newer software releases than were available with the rp7400 estimates were done. I would stick with my more conservative estimates for equal configurations. The web site for SPEC is simply www.spec.org and more specifically for the CPU tests at:
http://www.spec.org/cpu2000/results/

You want to look at the SPEC_rate numbers since these are multi-processor systems.
For the N4000 with 8 cores at 552MHz processors you will see the results at:
http://www.spec.org/cpu2000/results/res2000q1/cpu2000-20000313-00047.html
for the rp4440-8 with 8 cores at 1000MHz processors you will see the results at:
http://www.spec.org/cpu2000/results/res2004q1/cpu2000-20040126-02753.html
The result are roughly 33 for the N4000 to 73 for the rp4440-8, but remember that the clockrate almost doubled and the newer system has the 32MB L2 cache.
The rp7400 with 8 cores at 750MHz got almost 47. http://www.spec.org/cpu2000/results/res2001q4/cpu2000-20010911-00895.html

Since SPECint_rate does no disk or network I/O, it is more likely to be a best case senario. If the disk and network are the same for both systems, then your real application performance will not improve as much. Naturally, this can change if you have new disk and faster networks.