Last September, we had the opportunity to take a closer look at the Promise NS4300N. We were quite taken with the unit’s build quality, its wide-ranging support for various network protocols, and its clean and well-organized configuration interface. The unit's data transfer rates were a mixed bag, though.
While the NS4300N acquitted itself quite well in office productivity testing, its transfer rates were only average in our multimedia benchmarks. Luckily, Promise hasn’t just been sitting idle since then, and is presenting the successor to the NS4300N, the NS4600.
More RAM, New CPU
On the outside, the NS4600 appears to be virtually identical to its predecessor, but Promise has made a lot of changes under the hood. For example, the new model sports 256MB of RAM, twice as much as its predecessor. Promise also replaces the 400 MHz MPC 8343 Freescale CPU with Intel’s first x86-based System-on-a-Chip (SoC) since 1994, the EP80579 (code named Tolapai), running at 600 MHz.
This second improvement gives us a couple of points to look at when assessing Promise’s new NAS. First, how does it stack up against the Freescale architecture that came before, and second, how well does Intel’s latest SoC perform? Will it turn out that embedded solutions will remain the domain of vendors like Freescale (Motorola), AMD, and Marvell? Or does Intel have a chance here?
XOR Calculations on the CPU
The embedded solutions manufactured by the aforementioned companies all offer fairly high performance coupled with low power consumption. This is significant because a NAS appliance's data transfer rates are influenced to a large degree by the processor it employs, since most NAS devices use host-based RAID solutions that rely on the system’s CPU.
Simply put, instead of using dedicated hardware to processor parity calculations (XOR bits) used in certain RAID configurations and to distribute the data across the various disks, these tasks have to be handled by the NAS server’s CPU. As a result, any NAS device’s performance greatly depends on the hardware architecture at its core.
Tolapai: Attacking AMD‘s Geode?
Intel designed its EP80579 SoC for use in telecommunications systems, VPN, firewall appliances, VoIP gateways, and storage solutions like the one we're looking at today. These are all areas which have traditionally been covered by companies catering to the embedded market and, increasingly, by AMD. One such NAS device that is powered by AMD’s Geode processor is the Thecus N4100 Pro. The introduction of the EP80679 means that together with its Atom, Core 2, Xeon, and Itanium processors, Intel now offers a product for almost any class of device.
Over the next few pages, we’ll take a look at the feature set and bundle of the Promise NS4600, in addition to analyzing how Intel’s embedded EP80579 performs.

TH people, please do some tests in iSCSI mode =)
Can you please be more specific in your configuration setup?
Can you please be more specific in your configuration setup?
what's the price of this file server?
Would it make more sense to buy an atom Dual core system as file server? Since 25MB/s transfer rates are nothing to boast about!
At least the Atom processor may not be as low in power requirement, but runs Linux AND Windows, can playback 720p video, and get data transfer speeds far higher than the ones mentioned in the benchmark, probably in the likes of 150MB/s(as a wild guess), or perhaps even higher...
For $300, one could basically get a singlecore Atom 270 + 2GB of RAM, and load the files into RAM for faster upload speeds, probably close to the maximum possible through a gigabit lan connection;
(or am I understanding incorrectly why this NS4600 was made?
Depending on the filesystem employed and the client accessing the files, NAS devices show varying results. An iSCSI setup would be much more level, compared to higher level FS/NFS performance tests.
you do realize that most laptops cannot do 4 disk raid 5 right?
Were the hard drives in the 4100 and the 4300 the same? Why no iSCSI? Can the eSATA port be used to automatically back up the whole array to another disk (say having a pair and keeping one off site in case of a fire or other catastrophic event, obviously swapping ofsite adn local disks periodically).
How does this system compare to some some of the home brew RAID boxes that THG has set up, both in price and preformance?
TOO MANY UNANSWERED YET OBVIOUS QUESTIONS!!! Tom's used to be soo much better than this! What is going on over there? Where have all your good writers/reviewers gone (or are you being told to dumb things down by BOM)?
Yep, you can run the game from the drive!
Yeah, I wanted a dedicated NAS type file server, because at the time my gaming computer served that purpose and ran all the time. So I went to craigslist and went out and got a $10 Pentium 3 computer, slapped a PCI SATA card in it and added a 1 TB drive and was good to go.
First thank you for your valuable feedback. Please let me answer a few questions.
Test setup: To lower the impact on transfer rates we are always using the same computer and harddrives for our NAS device tests. You can have a look at the hardware details on page three of this review. There is only one exception: If the NAS device comes preconfigured with harddrives we are using these instead of our Samsung HD321KJ drives.
Slow network performance between Samba and Vista: There has been an issue on both sides, Samba and Vista that led to extremely slow transferrates (around 5MBit/s) when copying files from the NAS device to the Vista machine. According to my present knowledge these issues have been addressed with Service Pack 1 for Vista and the release of Samba 3.0.28a.
So the network configuration used is the standard Vista network configuration as you can find it after a fresh install of the OS. The reason why we do this is that we believe that the average user shouldn't have to fiddle with the network configuration to get satisfying transferrates. A customer expects that the device is running hassle-free out of the box. If it isn't working we think that the manufacturer has to come up with a solution and not the customer.
Jumbo packtes: Some NAS devices support jumbo packets, some don't. As we want to have comparable results we have to go without jumbo packets. Moreover there is no official standard for jumbo packets so it can have a negative impact on data transferrates in certain network setups. But you are right, we shoud have mentioned this in the review.
Stripe size: Almost all NAS devices intended to be used in a SOHO or home network don't allow a selection of the stripe size when setting up the RAID configuration for reasons of simplification. This is also true for the NS4600 and could have been mentioned in the review.
iSCSI tests: The reason why we haven't tested the NS4600 as an iSCSI target is plain simple: It doesn't support this feature.
Greetings from Germany,
Marcel